Pending author invites (due to Substack limit): , , .

TLDR: I asked 23 founders, consultants, and creators who use AI every day to name the one task they keep fully human, and what breaks when they don’t. Their answers landed on five things you are already better at than AI: the decision, the conversation, the first move, the sign-off, and protecting your own skill. Almost all of them hand AI more work than they keep. None of them handed over the start or the end of their process. AI takes the middle, and your core sits at the edges. Subscribe for part two, where the same 23 people share the tests they use to decide.

I build AI systems for people whose expertise is their product. The first thing I look for is the part of their process they won’t give up.

That part decides how everything else gets built.

Everyone has one. I call it a core. It’s the piece of the work you identify with most, the one with the most emotional investment. Hand it over and the work stops feeling like yours.

This core is what you shouldn’t automate, because it’s you. And by virtue of being you, you’re better at it than AI, no matter how smart the model gets.

That has been my working principle for a while now. But is this the same for everyone? I’m just one person, no matter how deep into AI I am. I wanted to validate this idea among people in my network, people who are far more experienced, technical, and smarter than I am at many things.

And I wanted to present you with their answers, so you can apply it to your own business.

This is for you if you’re building an online business or a brand with AI. You’ve felt the push to automate everything, but that doesn’t mean you should. Where do you hold the line? How do you determine? Here’s what 23 AI-powered creators say.

The more you automate, the more you lose

If you don’t know what not to automate, everything AI touches turns generic. You have no core.

Like a hydra, this problem can rear its many ugly heads in different parts of your business.

In marketing, you publish more than you ever have, and less of it lands. Every article feels like it’s building trust. Every article is quietly burning it.

What about in your branding? A brand is what someone recognizes before they read your name. But if you don’t know who you are, what you protect, it becomes generic. Generic output wipes that out. You get volume, efficiency, yet nobody remembers who made it.

Then lead acquisition. The DMs cool off. Offer posts stop converting. Prospects can’t tell you from someone charging half your rate, so they choose on price.

You have a business, and a core offer you spent ten years or more building. You’re token-maxxing everything you can. You’re moving faster than you ever have.

And you’re losing trust, losing leads, and losing money.

Why? Because you’re automating everything. Even the things you’re already inherently better at than AI. You didn’t know what not to automate.

These 23 people did. Every one of them automates more than you do. The difference is that they know where their line goes, and why it sits there.

The five things you’re better at than AI, according to 23 AI-powered creators

There are things you shouldn’t automate, because you don’t have to. You’re already better at them than AI is.

Finding them takes introspection. You have to know every inch of your own process, decide where AI fits into it, then decide where you hold the line.

That can take years.

It won’t for you. These five are what AI power users decided not to automate. Read them, use them to inform your own process, and skip the years of trial and error the rest of us paid for.

1. Deciding what you’ll answer for

About half the group named a version of this. What gets built, what gets published, what’s true, what matters.

Nobody complained about the writing. They complained about accountability.

A model returns the most probable continuation of what came before. Probable is not correct, and it is nowhere near committed. Ask it to make a call and you get the average of every similar call it has absorbed. Averages hold no opinions, and no risk.

Anita L. leads AI transformation for Microsoft Azure’s hardware infrastructure division. At Amazon she secured the GPU capacity that let her team launch the company’s first public generative AI app on schedule, because she was the only person in an advertising org tracking hardware supply. Nobody else had a reason to ask. Her line: AI “doesn’t know which questions matter.”

AI answers the question you bring it. It can’t tell you which one was worth bringing.

Elena Calvillo of Prompt-Led Product ships products end to end with prompts and audits other people’s AI products, so she spends half her week watching builders automate the wrong thing. “AI says yes to everything,” she wrote. “It has never watched a user rage-quit your onboarding. It has never felt the silence after a launch nobody cared about.” It carries no downside. It never pays for being wrong, so nothing in it wants to say no. Her verdict: “the fastest builder I’ve ever worked with. It’s just a terrible product manager.”

Ileana Marcut, product strategist and co-founder of Creative Glue Lab, keeps direction with her team, because AI has “no responsibility or ownership.” You end up with “a direction nobody committed to.”

You won’t see that in the output. You’ll see it three weeks later, when nobody follows through, because nobody chose.

Pinkie, of AI Meets Girlboss, spent a decade in brand strategy before building brand systems with AI. Her version of the same limit: AI “can generate a hundred directions all day long, but it has no taste about which one is actually you.”

If you can’t name who answers for the outcome, don’t let AI produce it.

2. Connecting with another human

DMs. Comments. Client emails. Sales replies. Six people drew their line around talking to other humans.

This category runs on different logic than everything else in the survey, and it’s the one people misjudge most.

Everywhere else, automation costs you quality. Here it doesn’t. An AI-drafted reply can be more articulate than anything you’d write at 11pm. The words come out fine.

The words were never carrying the value. A message carries the fact that a specific person spent attention on a specific other person. Automation strips that out and leaves the sentences standing, which is why the loss is invisible from your side of the screen.

It isn’t invisible from theirs. Claudia Faith of Level Up With AI answers every DM herself. “The moment a reply sounds even slightly auto-generated, the person feels it.”

Detection is asymmetric, and that’s the part that should worry you. One caught message doesn’t cost you one message. It reprices every message you’ve ever sent that person, because they go back and re-read the history.

Which makes a detected template worse than silence. Not replying costs you an opportunity. Getting caught costs you the relationship, and the ones before it.

Yana, who builds Substack businesses with AI and automations, frames it as identity rather than workload. “It’s like asking someone else talk instead of you - it’s about identity, not just a task.”

Anfernee of Solopreneur Code is the only person here who can name what this cost him in revenue. AI-drafted offer posts go “generic and ‘salesy’ again, which is the exact feeling that killed my sales.”

Jonas Braadbaart goes furthest. Twelve years building AI products, ML engineer through to CTO, and he keeps AI out of internal strategy memos. That’s the document most operators automate first.

3. Setting the direction of your project

The question. The noticing. The raw dump before anything gets shaped. Five people refuse to hand any of that over.

AI is extraordinary at generating options and structurally unable to rank them for you. It can rank by convention, by frequency, by what reads as reasonable. Ranking by what matters takes a stake in the outcome, and it holds none.

So volume works against you. Twenty directions isn’t more useful than three when you have no basis for choosing. It’s just more directions to evaluate against a rapidly-depleting resource — your energy.

ClariSynth, who writes about creating alongside AI, compressed this into six words. “Possibilities are not the same as direction.”

Ordering matters more than people realize. Set direction first and AI amplifies your judgment. Set it second and AI has already amplified something else, then makes that something else look coherent. A well-paved road is harder to abandon than a rough one.

David of ENG Sales found his line by failing at it. He tried letting AI outline first, then writing from that draft. It never held. “I always feel like I’m trying to stitch it together.”

The prose was fine. The direction wasn’t his, and editing downstream doesn’t recover that.

Mia Kiraki, an AI narrative architect, keeps the first move: which idea to chase, what feels strange or frustrating, whether something is worth keeping. After that AI gets everything. “I don’t use AI to give me something to care about.”

Wyndo of AI Maker still writes his own ideas and early drafts by hand, beacuse “AI doesn’t know my experience or things I feel.”

Joel Salinas, an SMB AI strategy consultant, draws it widest of anyone. His human-only list is “creative tasks, innovating, ideating.” AI gets execution inside closed parameters.

4. Weighing what it costs someone else

This is about taking responsibility for real-world impact.

AI evaluates artifacts in isolation. Risk doesn’t live in the artifact. It lives in the context around it: where the thing sits, who can reach it, what they’d already need, who reads the result and what they do next. A model sees the pattern. It can’t see the blast radius.

Christopher Ijams of ToxSec is a security engineer at AWS with access to some of the strongest security AI in production. He uses it for scale. Scanning, vulnerability discovery, triage across huge codebases.

Then he stops. Models produce false positives and run hot on severity. Something that looks critical alone might be medium risk once you know where it sits, and what an attacker would already need to reach it.

His reason for keeping the final call is sharp. “If I throw high-severity tickets at teams every time an AI scanner gets excited, I burn trust. And once teams stop trusting the security findings, even the real high-severity issues become harder to get fixed.”

Follow that one step further.

Being wrong about impact doesn’t only cost you the wrong output. It costs you the channel. Once your findings get discounted on arrival, your correct findings stop working too, and accuracy alone won’t win them back.

Almost nobody prices that in. It’s why the analysis can be delegated and the sign-off can’t. The analysis is about the thing. The sign-off is about the people who receive it.

Sam Illingworth of The Slow AI holds the same line on publishing. He checks every source quoted, and makes sure anything posted “was done by me and not automated.”

5. Protecting your own skill

Only 2 of 23 people gave this answer, but it affects 100% of AI users.

It’s protecting your own mastery.

Every task you hand over is a rep you don’t take. That sounds obvious until you notice which reps AI takes first. It’s best at exactly the parts that are hard, and hard is where skill comes from.

The trap is that output holds steady while capability drains. Nothing looks wrong, because the work still ships. You find out years later, on the day the model is confidently wrong and you no longer have the judgment to catch it.

Nika Kotláriková, a marketer of ten years who writes Business and Marketing, built a Chrome extension and wouldn’t let Claude write it.

She made it teach instead. Claude explains what needs to happen, then asks her to write the line. When she gets it wrong, it gives hints rather than answers. Then she explains the code back in her own words, to prove she understood it.

“My goal isn’t to get things done as quickly as possible. My goal is to learn, expand my skills, and increase my cognitive capacity.”

She is slower than almost everyone reading this. But a few years like this and she would master the skill, leveraging AI in ways that the purely AI-reliant can’t comprehend.

Jose Antonio Morales of Automato consults on digital transformation and has run a coworking community in Slovenia for twelve years. And … he’s stepping back on purpose. He deleted AI tools he had already adopted. The Notes optimizers. The article-review formulas. Gone.

“Now, what I do is to develop tools that coach me to bring the best of me to my work.”

Same technology, opposite direction. Tools that do the work, or tools that make you better at the work. Almost nobody picks on purpose, and most default to the first, to their own ruin.

What happens when you outsource your core?

Six people described the same breakage, in almost the same words, without ever comparing notes.

Sidenote: of everything that can break with AI (and there is many), this is the one to watch, because it never looks like damage. Every other failure announces itself. Bad output looks bad. You catch it, you fix it, you move on.

This one improves the work while removing the reason it existed.

The harm is this: A model optimizes toward the middle of everything it has seen, which means it sands off whatever made your version specific. The sharp number. The failure you admitted to. The angle only you would take, because only you paid for it.

What comes back is more balanced, better organized, more defensible. And interchangeable.

Daniel Ionescu spent more than twenty years in journalism and publishing before building Millennial Masters. When he lets AI past his editorial judgment, “everything starts becoming more reasonable, balanced and polished. The draft can look better while becoming less mine.”

Dheeraj Sharma of GenAI Unplugged spent twenty years building enterprise software and now runs his content pipeline on an AI operating system he built himself. “The output reads fine and says nothing.” It picks the safest angle and “drops the specific number or the failure that made the post worth writing in the first place.”

Zeng of Picaisso sees the same thing in images. “Polished yet interchangeable, and the work stops feeling like mine.”

Quality goes up. Distinctiveness goes down.

That’s why a quality check won’t catch it. Your draft passes. It reads well. It’s just no longer worth reading, and you’re the only person who can tell.

AI takes the middle. You take the edges.

Put all 23 answers side by side, and you can see the same shape. About half protected the beginning of their process. About half protected the end.

Nobody protected the middle. Execution.

The middle of any process is the drafting, building, searching, formatting, generating options, running the pass again. Execution is where AI is strongest, because execution has a defined target and a checkable result.

The edges are where the target gets set and the result gets judged. Both needs a stake in the outcome, and is uncheckable by AI. It needs you.

So the line was never about how much you automate. These 23 automate more than we do. It’s about position, the thing nobody can copy off you.

Listen to how similar they sound.

Pinkie of AI Meets Girlboss hands over “everything after the decision gets made.” Mia hands over “pretty much everything around that first move.” Dan Cucolea, co-founder and sole engineer behind an AI-native SaaS used by thousands of people, hands over “everything that leads up to that final decision.” Christopher lets AI do the first pass, and “it just does not get the last word.” Andrea Chiarelli, a management and strategy consultant, lets AI “map the evidence, explore ideas, dig deep into stuff,” then refuses to let it “decide what is important and why.”

Five people, all saying the same thing.

Find your own edge

Know yourself first. Not the job title, but what people actually come to you for. The thing you identify the most as. The change you want to do in the world. Your mission.

Then map your process. Every step, including the ones that happen in your head. Find the step that only worked because you were the one doing it. Not the step you enjoy. The one that would have come out worse in anyone else’s hands.

That’s your core. Designate it, then automate everything around it.

This is also why Katharine Gallagher of Learn Grow Monetize wouldn’t name just a single task. What she protects is “my professional, accredited, experienced lens.” Her judgement. Once you know where your core sits, it stops being a step and becomes the thing you bring to all of them.

Dan said it better than I can. “I want AI to help me explore more possibilities and execute faster, while I remain responsible for deciding what is worth putting into the world.”

You can have that today. You just have to know which part is yours.

The 23 creators I invited had far more to say than I had room for here, and their answers gave me a stack of usable tests for finding your own line.

That’s part two. It goes out next week, so Subscribe to be notified.

P.S. Twenty-three people gave their time to this, and the thinking in it is theirs as much as mine. Big thanks to everyone who participated. To our dear readers, subscribe to get part two, and tell us in the comments where your core sits. We’d love to know.

If you want to build an online presence for your business, you’ve probably thought about LinkedIn.

We’ve all heard the pitch. It’s the professional network. It’s where the CEOs are. Post there and clients will find you. And to be fair, some of that holds up. LinkedIn has about 1.3 billion members. And roughly 80% of all business leads that come from social media come from LinkedIn.

Personally, though, I use it as a discovery engine. I post top of funnel material there and I send people back to this newsletter. I haven’t cracked the code for it yet, but I do believe there’s potential in it.

Plus, I believe that Unpromptability means showing up where your readers might be. This doesn’t mean being tied to just one platform.

For a lot of experts, that place is LinkedIn.

Substack made a decision about AI, and I wondered what LinkedIn was doing

In July, Substack added an AI detector. A reader can scan any post and get a number showing how much of it a machine wrote.

I wrote about it twice, here and here. In both articles, my conclusions come down to one line: A scan can see the tool you used, it can’t see you.

The feed argued about it for weeks. Scores, false positives, whether writers were about to get dragged into the town square for using Claude.

Then I got curious about the platforms outside. LinkedIn has the same flood of AI content … and it handled it in almost the opposite way. Same problem, two answers, and neither platform is done deciding.

So I asked Nika Kotlarikova, one of the best LinkedIn creators I know with a thriving 10k follower community on Substack. Here are her thoughts.

AI on LinkedIn: helpful tool or hidden risk?

We can all agree that AI has made our lives easier in many ways. (Content creation, research, new inventions.)

But AI also comes with its downsides.

I feel this especially as a creator on LinkedIn, where I have my largest community.

In today’s edition, we’ll look at how LinkedIn views artificial intelligence and how to use it without sabotaging your attempts to grow your personal brand with AI.

I’ve divided this article into three parts:

1. The situation on LinkedIn from a creator’s perspective

2. LinkedIn’s stance on AI

3. How I personally use AI

The situation on LinkedIn from a creator’s perspective

If you’re also seeing phrases like “It’s not X, it’s Y,” “Here’s something nobody talks about,” “Agree?”, “Here’s why,” and “The best part?” more and more often in your feed, you can probably guess what’s happening.

We’re all starting to sound the same because people are using AI to create their posts.

Some take it a step further and use AI to generate their images, too.

But the worst part comes when people start using AI to write their comments.

Most of the time, it’s just a paraphrase of the original post, an agreement with the author, followed by a long wall of text. Sometimes, it also comes with too many emojis and tagging the author.

It feels impersonal, and real communication starts to disappear.

And we’re not the only ones noticing it.

LinkedIn is noticing it, too.

LinkedIn’s stance on AI

Every social platform benefits from keeping as many people on the platform for as long as possible. The more attention it gets, the more valuable its advertising space becomes to advertisers.

But if the platform is filled with AI-generated posts, followed by AI-generated comments from other AI-assisted creators, it starts to resemble the dead internet theory.

Who will be watching the ads?
Who will be buying from advertisers?

AI? Maybe in the future.

Still, LinkedIn has taken steps to limit the role of AI on the platform. For example:

• LinkedIn systems can detect automated activity, and excessive activity can result in restrictions on your account. Sometimes, even manual activity can trigger LinkedIn’s AI detection systems, which is why you should monitor how much and how frequently you engage.

• LinkedIn’s newer algorithms, 360 Brew, can also detect AI-generated patterns and may choose not to prioritise this type of content.

• LinkedIn has even introduced a feature for reporting AI-generated spam.

On the other hand, this can be counterproductive, since people can sometimes mislabel content as AI-generated when it isn’t.

There’s also a certain irony to LinkedIn’s approach. The platform itself offers automated AI-generated reply suggestions to users, while its own technical support is, at least to some extent, AI-powered too.

So what should we actually do with AI in our content creation process?

How I personally use AI

I don’t want to sound like an anti-AI movement, but in my opinion, we should complement AI and not leave everything to it. Yes, it can save a lot of time and our brain activity, so we can potentially gain time to learn other, more important things. But how many people will actually do that? Probably not very many.

My goal is to use AI in a way that I can learn something, increase my cognitive capacity and at the same time not lose critical and creative thinking.

So, when using AI to create LinkedIn posts, try to think first about how you would structure the post yourself.

Then, when you ask AI to improve it, analyse why its version is better. What exactly makes the improvement? Try to understand.

I wouldn’t call these hacks, but this is how I personally use artificial intelligence.

• Most of the time, I write my own draft first and then ask AI to polish it.

• I also ask Claude to identify which parts of my writing contain typical AI phrases. If something sounds too AI-generated to me, I replace it with my own version.

• I use Grammarly AI to fix grammar and phrasing.

• I also sometimes ask AI to rate my post and give me feedback on how it would evaluate it.

• For visuals, I use AI to create an initial concept for the graphic I want. Then, I try to recreate or adapt it in Figma, since AI-generated visuals can sometimes get a CR label.

AI and LinkedIn: It’s complicated

As you can see, the relationship between AI and LinkedIn is quite complicated. But I’m sure you can find your own way by finding a reasonable compromise.

You can use AI to make many things easier, but don’t forget to remain human and to actually connect with people who want to make new friends.

The internet has given us the ability to connect across continents. You never know when those connections might eventually bring us together in the real, offline world.

Wishing you all the best on that LinkedIn journey!

Yours,

Nika

Write down your own rules before the platforms write them for you

Look at what Nika does. She writes her own draft first. Then she asks AI to improve it. Then she studies why the new version is better (or not). She takes AI image concepts and rebuilds them by hand in Figma.

She’s not using AI to pass some scan. She’s using AI to protect her thinking. That’s why it still works when LinkedIn changes its algorithm again, and when Substack changes its mind about Pangram.

Platforms keep moving.

You have to hold the line. Your standard has to hold still.

So do this today: Write down where AI is allowed in your work, where it isn’t, and why. Write for as long as you can, as specifically as you can. When the next platform policy lands, you’ll already have your own answer ready.

Then, just keep showing up as a person.

The parts of your work that prompts could never reproduce are the parts worth protecting.

Your judgment. Your taste. The way you sound when you talk to a person you actually want to help.

I usually write about that in the context of writing. Today it shows up in sales … which is nearer to writing than most people know, anyway.

David Roy has spent fifteen years in B2B sales and helped build revenue past $100 million. He now works with technical founders who inherited the sales job by accident and have been running it on instinct ever since.

He’s going to show you why the voice on your page is your first sales call, three moves that build trust whether you are writing an email or sitting across a table, and where AI belongs in that chain.

David’s first message to me was a good example of his expertise. He engaged with my work, got to know me, and showed up as a person before pitching for anything. Fifteen years of selling, and the thing he led with was a connection.

The lessons he’s written here, he practices for his own outreach. He’s established his argument’s credibility in the DMs, before he wrote it down here.

Read it below.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can replicate. This week’s piece comes from , who has sold for a living for fifteen years.

For the foundation underneath it, read How to Train AI on Your Brand Voice.

Writing is selling

You wrote the pitch yourself.

Then you handed it to AI to make it better.

It came back cleaner, smoother, and sounding like no one.

That’s the trap. Used one way, AI sharpens your voice. Used the lazy way, it sands your voice off until you sound like everyone else who took the same shortcut. Clean grammar. Balanced sentences. The frictionless tone of a machine that has never once been nervous about losing a deal.

The customers you are reaching out to can feel it. When everyone sounds polished, polish stops being a signal. It becomes noise or worse an indicator of AI. The one thing that cuts through noise is the one thing AI can’t fake for you….YOU, showing up as a person.

I’ve spent 15 years in B2B sales, and these days I help technical founders sell without feeling like they’re wearing a costume. James and I come at the same problem from two doors. He helps you write in your real voice. I help you sell in it. Turns out they were never two problems.

Here’s the part you might not call selling. Every guest-post pitch, every “can I write for your audience,” every DM to someone you want as a client is a sale. You’re asking a human to trust you with their time, their inbox, or their readers.

Sales is all about problem solving with another human. To do that you have to communicate with that other person on a human level. Although AI is getting better it’s still very easy to pick up on AI generated communication. So you have to learn to write and sell in your voice. The people you’re pitching want to engage with you, not your AI.

Today James and I going to walk through that sales process and why selling in your voice is so important.

The buyers can smell the script

Think about the last time you got a message clearly written by a machine wearing your prospect’s name. You felt it in the first lines. Something was just…off. We had a version of this long before AI. Every call center script we’ve all hung up on. AI just spread it faster and dressed it up nicer.

That same instinct runs the other way when you’re the one pitching.

When I first started using AI, I was pitching other writers for collabs and guest posts. I wanted connections to my framework and a way to connect to their audience. I found a lot of creators to reach out to and I wanted to reach out to all of them, fast. So I did the lazy thing. Drop the writer’s name and their newsletter into AI, say “here’s what I want to pitch them,” paste the reply, send. Repeat.

Two of them wrote back with the same question:

“Did you just ask AI to come up with these ideas?”

The rest went into a black hole. No reply, no signal. I’ll never know for sure, but I’d bet at least one of those silences was the same thought, unspoken.

What stung is that I did it to writers. People who live and die by voice. I’d skipped the only thing that mattered to them: had I actually read their work, and did I have anything worth bringing to their readers? I had no idea because I did not to any of the work. I was still in funnel mode. Volume over connection.

The pitches that landed were the opposite. Every one started with me connecting to the writer and their audience first, before I ever floated an idea. That happened when I stopped cold pitching with AI alone. Instead I started spending the time to understand the work, then writing to it.

Because pitching a guest post is a sale. You’re asking someone to trust you with the audience they built. Nobody hands that over because your AI wrote a tidy paragraph. Don’t get me wrong, I still use AI at times to help with my pitches, but I make sure I spend the time to understand the value add to their audience before I send over any ideas.

The pattern is always the same. The moment I sound like a script, the other person starts performing back at me. Polite (most of the time 😁) nodding along, then gone. The moment I say the true, slightly awkward thing (I loved this piece, here’s where I’d push back, here’s what I could add for your readers), they lean in. Now there’s a person in the room.

Same voice writes and closes

By the time a buyer talks to you, they’ve already done most of the deciding. Consider this:

Most B2B buyers work through 60 to 70 percent of their research before they ever reach out.

They’ve read your site, your posts, your emails. They’ve built a picture of you from your words alone.

So the voice on the page is your first sales call. If your writing sounds like a committee, you’ve told them who to expect before you open your mouth. If it sounds like generic AI-generated copy, they don’t know what to expect and will likely move on to the next. And if they do click through, when the real you shows up on the call sounding like a different person, the difference becomes risk in the buyers mind.

The founders who close fastest collapse that difference. The email sounds like them. The call sounds like the email. The follow-up sounds like the call. One voice, all the way through.

Trust is the absence of a gap between what you wrote and who you are.

Be authentic in writing and selling

Authenticity gets treated like a vibe. In sales it’s a set of moves you can run, the same moves whether you’re writing a post or sitting across from a buyer. Here are the three that have closed the most for me.

Start with their problem

The instinct is to lead with what you do your features, but don’t. Instead open with the problem your reader is living in, in their words, before you mention your solution. It works in a cold email and it works in the first five minutes of a call.

A local HVAC company was hesitant to even book a call about their sales process. They knew it needed work. They just weren’t sure I was the right person to fix it. So instead of pitching how I’d help, I answered that doubt with value. Their CRM, ServiceTitan, was new to them, so I pulled three insights they could use to improve their sales process, whether we worked together or not. That’s leading with their problem: solve the real hesitation before you ask for the deal. We’ve been working through their sales process ever since.

Say the awkward true thing

The sentence you want to soften is usually the one that earns trust. The price is high. This part won’t fit your setup. That timeline is optimistic. Say it plainly, in the doc and on the call. I learned this by getting it wrong.

Early on, I was shadowing a senior rep on a call to learn the product. He talked the entire time. All the engineer wanted was a rough price for his planning. I could have cut in: “You’re just after a ballpark, right? Here it is. When you’re closer to kicking off, let’s touch base.” I stayed quiet instead. We lost the sale, and any shot at the next one.

The true, slightly awkward line is yours to write. Don’t be afraid of it. We’re all thinking it in the moment anyway.

Follow up with something useful

“Just checking in” is the most sent and most ignored follow-up there is. Every one teaches the buyer you have nothing to add. Trade it for something specific: the example they’ll need next, the answer to a question they haven’t asked yet, the one line that moves their thinking.

A founder I know runs a VA agency that helps hair salons book clients. She wanted a real sales process so she could hire a rep who’d actually succeed. We talked, I followed up but then… quiet. I followed up. Still nothing. So instead of nudging again, I took what I’d learned in our discovery call and built her a sales process outline. I did the value work first, then sent it. That got us back on a call to finalize it. Another ‘just following up’ email and I’d probably still be waiting.

Add value first. That’s the judgment call AI can’t make for you, because only you know what your customer actually needs.

Let AI do the non-human work

None of this means abandoning AI. I use it every day. The move is knowing which parts of the job are human and which parts aren’t.

AI is excellent at the non-human work. Pulling research on a prospect. Structuring a messy draft. Turning your voice memo into a first pass. Catching the typo you’d have sent at 11pm. I use it for all of these things regularly. But I stay in the loop when connecting with the other human I’m trying to solve a problem with.

AI falls apart on the human work. The diagnosis of what your buyer is actually afraid of. The one specific line only you would write, because only you sat in that meeting. The decision to say the hard true thing instead of the smooth safe one.

So keep the chain honest: AI drafts the scaffolding → you write the sentence that could only come from you → you have the actual conversation.

The human stays in the loop at every point where trust gets built. That’s the whole game.

Your real voice is the close

Take the last pitch you sent: a cold email, a collab ask, a DM to a writer whose audience you want. Read it out loud.

Any line you would never say to that person’s face, cut it. Rewrite it the way you’d actually say it, sitting across the table.

If it now sounds a little more like you and a little less like everyone, you’re doing it right.

AI can draft the words. It can’t be the person. And the person is what the buyer came for.

Show up as yourself in the writing, and you won’t have to become someone else on the call.

Sales is just problem solving with another human.

Trust is the absence of a gap between what you wrote and who you are

I keep returning to this line from David. I think about that gap more than most people would expect, because my work can widen it.

Every system I build for a client moves some piece of their thinking out of their hands and into a machine. Handled carelessly, it puts a stranger’s voice in their mouth and they don’t hear it happening. Handled well, it clears out the work that was never theirs to do and leaves them more room to be the person their buyers came for.

The gap never announces itself. Nobody writes back to tell you the email sounded like a template. They just stop replying, and you spend the next quarter wondering why the pipeline went quiet.

So do what David said. Read the last thing you sent out loud. Cut any line you would not say to that person’s face.

David writes ENG Sales at engsales.substack.com. If you defaulted into running sales and want a read on where it’s breaking, his free Flywheel Assessment is the place to start. Subscribe to him. He earned it with this one.

Then look at your own side of the chain. Most of the gap David describes comes from handing AI the work that was supposed to stay yours. The AI Readiness Diagnostic takes five minutes and tells you which parts of your work a machine can carry and which parts it will quietly ruin.

Your voice is the asset. Build the system around it.

PS. If you want to see which parts of your own work AI can safely carry, take the free AI Readiness Diagnostic.

I recently heard a story. Someone talked about it on a podcast.

In a big baseball game, you can measure someone’s popularity by how loud the crowd cheers when the camera pans to them and they get noticed.

Who do you think got the loudest cheers? Not the TV stars. Not the influencers.

It was the livestreamers.

Now, I don’t know if the story is a fact. But it makes a lot of sense. Livestreamers show their lives. They spend all day, sometimes multiple days, just interacting with people. Most importantly, they show the real world. With close to zero filters. People feel like they’re right there.

Unpromptable exists to bring you ideas to be more differentiated. Not just to stand out, but to be irreplaceable by AI.

One of the best ways to do that today is to do real things.

In the age of AI, reality is the moat. What you do in the physical world is what makes you irreplaceable in the digital one.

New here? I’m James. I help founders and experts build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do, by turning your expertise into systems no one else can copy.

For the foundation this piece builds on, read The Only Moat Left In The AI Age.

Reality is the moat

The internet is not scarce anymore.

An SEO firm called Graphite has been tracking this. They sampled tens of thousands of English-language web articles and found that AI-written pieces passed the halfway mark in late 2025, and have sat right around that line ever since.

Half of everything new. That is the room you publish into now.

Two men worked out what happens next about fifty years ago, and neither of them was thinking about the internet.

Peacocks and differentiation

In 1975, a biologist named Amotz Zahavi got stuck on a question about peacocks. He called his answer the handicap principle. The tail makes no sense. It is heavy. It costs a lot of food and energy to grow. It slows the bird down and makes it easier for predators to spot. Evolution should have deleted it.

Zahavi’s answer was that the tail works precisely because it is a burden.

A peahen cannot look inside a peacock and check whether he is healthy. All she has is what he shows her. And a sick, badly fed peacock cannot grow a huge tail and survive carrying it around. The strong one can.

The tail works as a receipt for health. The cost is what makes it true.

College degrees as signals

Two years earlier, an economist named Michael Spence had reached the same place from the other direction. He was thinking about hiring, and the work he did on it later won him a Nobel Prize.

An employer cannot open your head and measure how capable you are. They get a resume and an hour of conversation. What they can see is a degree, and a degree works as a signal for a reason that has nothing to do with what you studied. Finishing one is harder for a person who lacks the ability than for a person who has it. The cost sorts people, so the employer trusts it.

Both men land on the same rule, and it governs every market where the buyer cannot see inside the seller:

A signal only carries information when it is expensive to fake.

Differentiation today

The reverse matters more right now. Once a signal gets cheap enough for anyone to send, it turns into noise. Everybody sends it, so it stops separating anyone from anyone, and people quietly start looking for a different signal.

Writing was your tail.

Twelve hundred clear, organized words about your field used to prove you had spent ten years in your field, because nobody outside it could produce them. That price has dropped to about four minutes, and your reader knows it. The signal still goes out. It stopped meaning anything.

What still costs something is everything that happens off the screen. It cannot be generated, and it leaves evidence behind.

Ordinary people are already moving this direction without being told to. Eventbrite reported a jump in block parties, picnics, and neighborhood gatherings through 2026. Forbes called community building the hottest trend of the year, driven by run clubs, supper clubs, and other third places people are building for themselves.

Nobody instructed them. They just got tired of screens.

What does this mean for you?

You have fifteen years of judgment and a website that reads like every other website in your category.

You and your closest competitor describe the same expertise, with the same tool, in the same clean voice. Your buyer opens both pages and finds no reason to pick one.

They cannot see the difference between an advisor with a decade of pattern recognition and someone who wrote a good page about having one. Credentials stopped doing that job sometime in the last two years.

So the raw material for your best work stops coming from your reading list. It starts coming from your calendar.

That change works in your favor. A generic operator can copy your homepage in a morning. They cannot follow you into a room. You already have the decade behind you, and the room is where people finally get to see it.

Three ways to start.

Connect with real people

Community is one of the five foundations for a reason.

Someone with a prompt can rebuild your framework in an afternoon. They cannot rebuild the four weeks you spent arguing with a peer until the two of you landed on an idea neither would have written alone.

Pick five people whose work you actually read. The algorithm will not keep them in front of you, so this part has to be deliberate. Do something with each one that requires a shared calendar. A call. A working document. A piece with both names on it. A comment is not that. Neither is a compliment in the DMs.

I have three co-authored pieces in flight right now, each with a different builder, each on a subject I would not have chosen alone. I had to build a habit tracker to make myself do this consistently. Every one produced something I could never have prompted my way into, because it came out of disagreement.

Do uncommon things

Anyone can be busy. The moat comes from spending real hours on something specific and slightly unreasonable, pointed in a direction only your mission would justify.

Look at your mission and find the version of it that takes months instead of minutes. Build the tool nobody asked for. Commit to something you cannot possibly finish this quarter.

The cost is the whole point. A prompt cannot pay it.

Mine is a book. I’m writing a book on unpromptability, built out of case studies, teardowns, and systems I put in front of real people first. It will take most of a year. That is the appeal. A year of stacked evidence is not something anyone catches up to by typing faster.

Show yourself

Your presence is the one signal a model still cannot produce. Your face. Your voice. Your hands on the actual work.

It costs a different currency than the other two. You pay in discomfort.

I am bad at this.

DMs are fine. Video calls are fine. Put me on a podcast and I can talk for an hour without flinching. Posting a picture of myself is somehow worse than all of those combined. I have spent a long time writing about differentiation while quietly skipping the most effective version of it.

The discomfort is the evidence. It works for the same reason it is hard. Cost is what makes a signal believable, and this one gets charged in awkwardness instead of hours.

Start smaller than you think you should. Your face in the frame at all. A two-minute voice note with no editing. Your actual screen while you work through a real problem.

I’m working on it... Slowly … And perhaps, badly.

Do it for the differentiation

Doing real things so you have something to post is a little weird, and readers can smell it. The consultant who joined the run club for the content fools nobody.

Order fixes it. Do the thing because it serves your mission. Then talk about it because someone out there needs to know it is possible.

The test is simple. If you would not still do it with the camera off, skip it. There is plenty of real work waiting.

What comes back to you compounds. Every hour spent in the physical world turns into proof and material at the same time. The proof feeds your offer. The material feeds your writing. And the people you meet along the way feed both.

Reality is the moat

Everyone in your market has the same models, the same prompts, and close to the same words. The gap between you and your nearest competitor stopped being a writing gap two years ago.

What is left is what you did this week that took time, cost something, and happened somewhere a machine could not go.

Go do something real. Then talk about it where your people are listening.

P.S. If you want to see which parts of your expertise are still doing the work of proving you know your field, and which parts stopped, the AI Readiness Diagnostic takes five minutes.

This newsletter goes back to one idea: Your twenty years of knowing what works should run inside systems you own, instead of leaking into tools that make you sound like everyone else with the same $20 subscription.

I called this idea Unpromptable. Being impossible to replace by AI.

One of the four pillars of Unpromptability is building assets that compound. An asset is something you build once that keeps working after you stop touching it. It builds trust and spreads your ideas, even when you’re not there. It also helps you earn an income.

And if you build it the right way, each asset makes the next one easier to build.

That’s why I brought in Timo Mason. He runs Write Your Way to Wealth and sells digital products for a living. His method for turning a freebie into a paid offer is the cleanest I’ve seen. His funnel have made 8,000 sales so far.

Timo’s half makes your product effective. My half makes it yours, so nobody can rebuild it the Friday after they buy it. What survives that test is the part carrying your judgment. Your edge cases. The things you refuse to do.

You’ll leave with not just any product, but a product that’s specifically from you, for your audience. Irreplaceable. Let’s go.

New here? I'm James. I help founders and creators build AI systems that make them impossible to replace.

My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can replicate. This piece is co-written with , who builds and sells digital products for a living.

For the principle underneath all of it, start with The Only Moat Left In The AI Age.

The first step is to make something

All of this assumes you already have a freebie people are downloading. If you don’t, that’s the thing to build this week.

My first product took two hours, cost nothing, and was a Google Doc full of prompts I’d tested over months of writing with AI. People downloaded it. Some of them donated, which added up to $300. Two of them later booked bigger projects with me.

The doc was never the point. It gave people something to hold, and holding something is what turns a reader into someone who trusts you.

The bar is lower than you think. A checklist, a spreadsheet, a document that answers one question you keep getting asked in DMs. Pick the narrowest problem you can find, answer it properly, and ship it before it’s pretty.

I walked through the full build in Create a digital product with AI. Timo also wrote about his method for this, and you can find it in his newsletter. Timo also wrote about his method for this. You can find it in his newsletter.

Timo’s framework: From freebie to product funnel

Your freebie is working. It’s bringing in subscribers and people are getting value from it.

Now it’s time to turn that attention into revenue.

The mistake most creators make is building a random paid product that has nothing to do with the freebie that attracted their audience in the first place. Instead of creating disconnected product islands, you want to build a product ecosystem where each product naturally leads to the next.

In this article, I’ll show you how to turn a freebie into your first low-ticket funnel, the optimizations that increase conversions, and the principles behind building a product ecosystem that compounds over time.

If you don’t have a freebie yet, read this article first.

The “Problem → Solution” Framework

Your Target Persona has a problem, you hand them the solution in your freebie, and solving that first problem automatically creates the next one. Your low-ticket product is the solution to that next problem.

Problem → solution → problem → solution…

My example:

I built a free GPT that helps people ideate viral content, it worked well, people used it, and it solved their problem of having no ideas.

But solving that ideation problem created a new one immediately. They now had to write the actual content.

So I put a viral content writing tool directly in front of them right after, and it converted really well. :)

This “Problem → Solution” Framework is your foundation for a successful digital product funnel, but to maximize the conversion rate, there are 3 optimizations you should have in place…

But before that, here’s James.

Use AI to find the new problem, and make it yours

As you’re finding that new problem, you need to think about another dimension -- one that will take your product from “digital product” to “branded asset”“ and that is your differentiation.

The boost is checking which direction your readers already asked you about. You have that on record. It’s in the questions people sent after downloading the freebie, the replies you never filed, the comments under the post that promoted it.

Think through Timo’s framework first, then run the ones below, and build where the two agree.

Prompt 1: The evidence sweep

What it does: turns your scattered audience evidence into a ranked list of problems your readers actually hit, in their own words.

Before you run it, gather: the freebie itself, every question anyone asked you about it, any refund or unsubscribe reasons, and the last ten conversations you had with a customer or client.

1. KEY RULE

You are reading evidence, not generating ideas. Every problem you surface must trace back to a specific line in the material I give you. If you cannot quote the source, do not include it.

2. ROLE

You are an analyst who reads raw customer material and finds the patterns underneath it. You do not brainstorm, speculate, or fill gaps with what usually happens in this kind of business.

3. TASK

First, ask me to paste my material and label each source (freebie contents, questions received, comments, refund reasons, customer conversations).

Then:

- Pull every point of friction you find, quoted word for word as   the person wrote it.

- Group the quotes by the underlying problem, not by topic or   wording.

- For each group, state what the person was trying to do at the moment they got stuck.

- Rank the groups by how often they appear and how much the person seemed to care.

Output a table: problem, how many mentions, which sources, one representative quote.

4. SPECIFICS

- Quote exactly. Never clean up or paraphrase what someone wrote.

- Two people describing the same struggle in different words belong in the same group. Two people using the same word for different struggles do not.

- If a group has only one mention, say so and keep it. Small signals matter at this stage.

- If the material is too thin to find patterns, tell me that directly and tell me what to collect next.

5. TOOLS

My pasted material, labeled by source. Nothing else. Do not use general knowledge about my industry to fill in gaps.

6. EXAMPLES

[worked example to be added]

Prompt 2: The judgment test

What it does: takes the directions that survived and scores them by how much of your expertise the product can absorb. A product that only needs public information stays cheap and copyable. A product that needs your judgment is the one worth owning.

1. KEY RULE

Two products can solve the same problem and only one of them is hard to copy. The difference is how much of my judgment the product has to carry to work. Score for that, nothing else.

2. ROLE

You are a competitor with money and a chat window. You want to rebuild my product after buying it. You are scoring how hard I am about to make that for you.

3. TASK

Ask me for my candidate directions (from Timo’s prompt and from the evidence sweep) and a short description of my background.

For each candidate, answer:

- What would this product have to decide for the buyer in order to be useful?

- How many of those decisions could you make correctly using only public information?

- How many need something only I would know from doing this work for years?

- If you bought this product, how long would rebuilding it take you?

Then rank the candidates from most copyable to least, and say which one you would least like to compete with.

4. SPECIFICS

- Judge the product, not the topic. A common problem can still need uncommon judgment.

- Name the specific decisions. “It needs expertise” is not an answer.

- If a candidate needs no judgment from me at all, say so plainly.

  That one belongs in the freebie.

- Do not rank by market size, demand, or profit. Only by how hard it is to rebuild.

5. TOOLS

My candidate list, my background, and the evidence sweep output if I have it.

6. EXAMPLES

[worked example to be added]

And now, for Timo’s optimizations.

Optimization #1: Match the Format

One of the biggest conversion boosters is keeping both products in the same format.

If someone downloads a GPT, they’re already comfortable working inside ChatGPT. Asking them to suddenly use a separate AI app or another platform creates unnecessary friction.

Examples:

• GPT → GPT

• Text course → Text course

• Notion template → Notion template

Whatever the format is, keep it consistent and just make the paid version more valuable at solving the next problem.

The part of your product a stranger cannot rebuild

Timo’s rule is sound. Same format, less friction, more conversations.

But if you really want to differentiate in true Unpromptable fashion, you need to extract your judgement, your own experience, and be sure it helps a specific person, at a specific moment, with a specific problem.

Your product should give the answer you’d give if the person were sitting across from you, which usually depends on things the textbook never mentions. That answer lives in your head and it has never been written down.

Run this before you build anything.

1. KEY RULE

I know more than I can explain on demand. Ask one question at a time and keep pulling until I give you a specific instance. General answers are useless here.

2. ROLE

You are interviewing me to extract the decisions I make without thinking about them, the way a good apprentice would after watching me work for a year.

3. TASK

Ask me what product I’m building and who it’s for. Then run three rounds, one question at a time.

Round one, edge cases. Ask about five situations where the standard answer fails. What did I do instead, and why.

Round two, rules. Ask what I always do that others skip, and what I refuse to do even when a client asks.

Round three, corrections. Ask what a beginner gets wrong that I would fix in the first minute.

Then write the output: a set of decision rules, phrased as instructions the product itself can follow.

4. SPECIFICS

- One question per message. Wait for my answer.

- When I answer generally, ask for the last time it actually   happened.

- Ask why after every rule I give you. The reason is often the more   useful part.

- Do not suggest answers or offer options. Ask and wait.

- In the final output, keep my phrasing. Do not smooth it into   professional language.

5. TOOLS

My answers only. Do not add best practices from your general knowledge.

6. EXAMPLES

[worked example to be added]

Optimization #2: Match the branding

Your two products are related, so the branding needs to say that out loud.

The name should share a structure, and the visuals should feel like they belong to the same family. If the visual identity is completely different, the connection feels off even if the reader can’t say why.

If you’re building related products, make their branding related.

Consistent product branding with AI, without creating average slop

Timo’s rule is the standard for a reason. Products that look related sell each other without you saying a word.

Holding that consistency by hand gets harder with every product you add. Six months from now you’ll be building product four at eleven at night, and you’ll pull up product one to remember how you named things and how you opened it. That works until it doesn’t.

The boost is writing it down once. Not a prompt you run, a file you keep. Every product you build after reads from that file, so product four carries the same decisions as product one without you going back to check.

Use this:

Asset 4: The house style file

What it is: a plain text document you attach to every build. Write it once, update it when something changes.

What goes in it:

• Naming convention, with three real examples from products you’ve already made

• How every product opens, and how it closes

• What every product always includes, and what it never includes

• Words and phrases you don’t use

• Your visual defaults: colors, fonts, layout

• Two or three samples of your actual writing, pasted in full

That last one carries more weight than the rest combined. Rules tell the model what to avoid. Samples tell it how you sound.

1. KEY RULE

Find the pattern in what I already made. Do not tell me what good style looks like or suggest improvements.

2. ROLE

You are documenting an existing style so it can be repeated, the way someone would write down a recipe by watching a cook who never measures anything.

3. TASK

Ask me to paste three to five of my best products or pieces of writing.

Then produce a house style file covering:

- Naming convention, with the pattern spelled out and my examples   shown

- How I open and how I close

- What shows up in everything I make

- Words and constructions I clearly avoid

- Sentence rhythm and paragraph length

- Reading level

Write it as instructions, not description. “Open with the reader’s situation in one sentence,” not “the writing has an engaging tone.”

4. SPECIFICS

- Every rule must point to where you saw it in my samples.

- If my samples disagree with each other, say so and ask me which   one is right.

- Skip anything you can only find once. One instance is not a   pattern.

- No compliments and no critique. This is a record, not a review.

5. TOOLS

My pasted samples. Nothing else.

Optimization #3: Keep the Price Impulse-Friendly

This purchase should feel effortless.

There isn’t a universal perfect price, but remember: this isn’t where you’re trying to maximize profit. You’re trying to convert a reader into a customer.

For that reason, I usually price my first upsell at $7.

The real goal isn’t making money from this product.

It’s creating your first paying customer.

As the saying goes:

“You don’t make money with the upsell. You make money with the upsell of the upsell.”

Once someone has bought from you once, buying future products becomes much easier.

What you’re actually charging for

Before you set the price and hit publish, run one last check on the product itself.

Two products can look identical from the outside and be worlds apart underneath. One carries public information anyone could have assembled. The other carries decisions you only make after doing the work for twenty years. You can’t tell which one you built by reading it back, because everything in there looks obvious to you.

So hand it to a machine and tell it to rebuild the thing. Whatever it can’t reproduce is what you actually made. That’s the part your buyer pays for, at seven dollars or wherever Timo’s logic takes you.

Prompt 5: The replication test

What it does: tries to rebuild your finished product from scratch, then reports what it couldn’t reach.

Run this on your product draft.

1. KEY RULE

Try to rebuild my product using only public information and your general knowledge. Report honestly on where you fell short. Do not be generous with me.

2. ROLE

You are a competitor who bought my product this morning and wants a version of it live by Friday. You have no access to my experience, my clients, or my history.

3. TASK

Ask me to paste my finished product in full.

Then rebuild it from scratch, working only from public sources and your own knowledge.

Then report:

- What you reproduced easily, close to identical

- What you got roughly right but thinner or vaguer

- What you could not produce at all, and what you would have needed   to know to produce it

- How many hours your version took

Finish with a list of the parts that survived the attempt.

4. SPECIFICS

- Actually attempt the rebuild. Do not describe how you would.

- Judge output, not effort. If your version works for the buyer, it   counts as reproduced.

- Name what was missing in specifics. “Personal experience” is not   an answer.

- If you reproduced almost all of it, say so plainly.

5. TOOLS

My pasted product and public knowledge. Nothing I have told you about my background in this conversation.

Whatever survives is what you’re charging for. If nothing survives, run the judgment interview again and put the answers back into the product before you launch it.

This framework doesn’t stop at $7

The problem-solution loop you just used to go from freebie to $7 product works exactly the same way for your next upsell too.

Your $7 product solves the problem your freebie created. But solving that problem creates a new one, and that new problem is exactly what your next upsell (at a even higher price point) should solve.

You can use the same framework and same 3 optimizers from today to build out your funnel more & more.

Freebie → $7 product → $50 product → TO infinity and beyond! :D

Assets only you can make

The gap between a product that sells once and a product nobody can copy is entirely in what you fed it.

Timo’s method gets you a real product, and getting there beats most people. Run his prompt, pick the direction, ship at seven dollars. But if you really want to ensure that a competitor can’t copy it next Friday, you need to load it with something they don’t have.

Your edge cases, your refusals, the corrections you make when a situation goes sideways in a way the textbook didn’t cover. Those can’t come from a prompt. Only from the years you spent getting them wrong first.

The house style file is where that compounds. Write it once, attach it to every build after, and your fourth product carries the same judgment as your first without you rewriting a word.

If the launch mechanics are what’s stalling you, Timo’s Gumroad Game Plan walks through the setup end to end.

Your judgment is the input. AI shapes it into something people can buy. Give it a prompt instead and it hands back a product that already exists, sold by someone who got there first.

A few years ago, AI took my income.

I was a freelance writer. Clients who used to pay me for articles stopped, nearly all of them, inside a few months. I spent weeks wondering if I had built a career on a skill with no future.

So I learned the tool that replaced me. I rebuilt my whole process around it, got faster, and started earning more than before.

And in the middle of all that, I nearly handed over the one part that made any of it worth paying for.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace.

This piece is the sorting method I use: which work to hand over completely, which to let AI draft while you keep the call, and which to never touch. Plus why your judgment fades even when you protect it, and the one-minute habit that stops it.

For the foundation underneath this, read The Only Moat Left In The AI Age. For how other operators are running this in practice, read how 20 AI-powered entrepreneurs actually work.

The ceiling moved while you were busy

A few years ago this question answered itself. “What to automate with AI? Well, everything that it can automate!”

And that usually meant “the boring work.” But, you automated them because the boring work was all the machine could reach.

Now, it can reach far more into you. It reaches your judgment. It drafts your proposals. It scores your leads. It writes in something close to your voice and suggests the things you would have suggested.

The ceiling is gone. And while most people are treating that as good news, a few are cautious. I'm one of them. While I use AI extensively, I automate conservatively. And I do that by sorting every task in my business onto a ladder with three rungs.

The three rung automation model I follow

Automation starts from the bottom.

The limit is up to you. Your red line is the highest rung you let it reach. Here are mine.

Bottom rung: hand it over and forget it

The bottom rung is everything a client never sees and never credits you for.

Medicine has the cleanest example. AI scribes listen to the appointment and write up the note. A study across six health systems in JAMA Network Open found real drops in documentation time, mental load, and burnout.

AI took over the transcript work. It never took the diagnosis.

Mine look like this:

• Client calls get recorded and turned into notes and a task list I never type.

• Contract drafts, timelines, and standard client emails come out of templates I built once.

• One finished article becomes about fifteen pieces of platform content, and I never touch the reformatting.

Nobody hired you because your invoices arrived in a beautiful font. If a client could not tell whether you or a machine did it, hand it over and stop touching it.

The mistake down here is doing too little. If you are still doing this work by hand because you never got around to it, you are paying yourself minimum wage out of your own margin.

What bottom-rung automation looks like in an expert business

Most consultants underestimate how much of this is already buildable:

• A system that reads your calendar, preps you before each client meeting, and nudges the client to actually show up

• Meeting recordings turned into notes, next steps, and a draft follow-up email waiting in your drafts folder

• An intake form that creates the client folder, the CRM record, and the kickoff checklist the moment someone signs

• A searchable archive of every call you have ever had, so you can find the thing a client said in March

None of these touch a decision. That is why they are safe to build for your own business.

Middle rung: let the machine draft, keep the decision

The middle rung holds most of the money … and most of the danger. The steps stay the same. The right answer does not.

Mine look like this.

Every client we work with has a project holding their public business information, their marketing, and their goals. I can ask questions of that pile.

This client keeps hitting the same wall on LinkedIn, so give me three approaches nobody has tried. I pick the one that fits the person. Sometimes none of them fit, and that tells me something too. For research, I might feed in hundreds of comments and transcripts and ask what patterns are there. The machine finds them. I decide what they mean.

Other versions an expert could build:

• A proposal drafter trained on your last ten won proposals, that stops before scope and price

• A prep brief before every session: last session’s notes, open threads, a suggested opening question

• Lead scoring that sorts and flags, and never replies to anyone without you

• A content engine that pulls ideas from the questions clients actually ask you

• An assessment system that reads, scores, and communicates with your people about a specific problem you’ve already solved 100x before

Let the machine lift. You keep the choice, the judgement.

Why “keep a human in the loop” is not enough

The trap on this rung is comfort. Harvard Business School and Boston Consulting Group ran the biggest study we have on this, with 758 consultants.

On tasks the model was good at, the ones using AI did more work, faster, and better. On the one task the model was bad at, the ones using AI did about 19% worse than the people working alone. Fabrizio Dell’Acqua, who led that part of the research, calls it “falling asleep at the wheel.”

The answer looked confident, so people stopped checking. Even when there was a human in the loop. The draft is usually good enough. Good enough is the sound the trap makes when it closes.

Another study: a 2026 Harvard Business School working paper read the chat logs of more than 70 consultants who worked with LLMs for business problems. As they worked on validating the model’s initial findings with fact, their models did not back down. It apologized, corrected itself, then restated the same wrong answer with more detail and cleaner reasoning. The researchers describe professionals getting talked into it by the thing they were supervising. (Interestingly, one of the writers, Prof. Karim Lakhani, has a Substack, also about AI! Find it here.)

Granted, Gen AI has moved fast, and this paper was published only last year (2025). But I think the larger point still stands: even if you were paying attention, even if there was a human in the loop, you can still make the wrong, AI-influenced decision without proper judgement.

Build your judgment into the system

Slowing down does not fix this. Neither does watching harder. What works is putting your judgment inside the build, so the system cannot finish the job without you.

The systems I build does that three ways.

1. It stops at fixed points. The system runs at full speed up to the stop and then sits there. No nudge, no timeout, no default that fires if I ignore it.

2. My standards also live in a file the system reads on every run, instead of a prompt I retype when I remember. On a bad day, the file still holds the bar.

3. And it hands me options with the reasoning attached, so I am choosing between things instead of approving one thing. Approving takes a second. Choosing makes me look.

None of this makes the system slower. Everything below the stop still runs while I sleep. The only thing I added was a place where the work has to pass through me.

That is also what makes it hard to copy. Someone can buy the same tools and clone the same flow. They cannot clone where you put the stops, because you placed them using the judgment they do not have. Your decisions become part of the infrastructure.

The tell that you have slipped is boredom. When the review stops feeling like a decision and starts feeling like a keystroke, that task already moved down a rung without asking you.

Top rung: what should not be automated

Everything on this rung comes down to one line. Never automate the point where your work meets a person.

Everything before that point is machinery. The point of contact is the product.

For a coach, this point might be the session, and the silence right after someone finally says the true thing. For a consultant, it is the moment you tell a client what you actually think, and then stay in the room while they deal with whatever good or bad news you got. For an advisor, it is the call where someone decides whether to trust the recommendation.

Automate everything around those. Never automate it.

For most expert businesses the list looks like this:

• The session itself, and the read on what a client is avoiding

• Delivering bad news, and the work of fixing it with them

• The call on whether a prospect is a fit, and the decision to turn one away

• Every reply, comment, and DM you send to your own audience

• Asking for a testimonial, or pricing one specific job

None of that is efficient, but that’s kind of the whole point. Human contact is inefficient, but it’s the most valuable.

These are also the moments that get more of you every time you automate a rung below well. The hours you take back do not disappear. They go here.

Mine covers every point where a client meets me.

The first call. The progress updates. The call when something breaks and we are working out the fix. I write my own emails and I try to make them warm, because that email is the relationship, and nothing is underneath it if I hand it over. Same rule for my brand. I do not automate relationship building, so I built Daily 10, a relationship habit builder that puts ten people in front of me every day. It runs the queue. It has never written a comment for me.

A system can decide who you talk to, when, and about what. It cannot be the one talking.

Automation on this rung has a cost

Automate this rung by accident, and people do, and you change what you sell. You stop selling a person who knows and start selling an output anyone with the same tools can make.

And the price of a commodified output only moves one way.

Deloitte Australia showed the fast version. Last year, the firm delivered a 237-page review to a government department for around AU$440,000, and had to refund the government back.

A researcher at the University of Sydney checked the sources. The report cited academic papers that did not exist and quoted a federal judge saying something no judge had said. Deloitte refunded the final payment. The department said the analysis and recommendations still stood, which is the whole lesson. The thinking survived (inasmuch as it survived after having suffered a shameful, public, blow). The sourcing did not, and the sourcing was the part carrying the firm’s name.

But, you can guard the top rung perfectly and still lose it.

In 1983 a psychologist named Lisanne Bainbridge published a paper called Ironies of Automation.

Automate most of a process and the operator keeps only the parts that could not be automated. Their job turns into watching. They stop practicing the skills the job used to build. Then something goes wrong, a person has to take over, and that situation is hard by definition, so it needs more skill than the old job did.

They have less.

Your judgment works the same way. It is a practice, not a possession.

So sorting your ladder correctly is not enough. Going back to doing the grunt work will not fix it either. Making your own decisions visible to yourself will.

Build the log, take back your judgement

So every time you override one of your systems, write three lines. What it recommended. What you did instead. What you saw that it missed.

Ninety seconds. No tool, no template. One running document.

That log does three separate jobs:

• It gives you the rep back. Naming the reason is the practice that watching takes away, and it costs a minute instead of an afternoon.

• It shows you which overrides repeat. Anything you have written four times is a rule you can now state, which means it just moved down a rung on purpose instead of by accident. Feed it to the system as a check. The decision still stops with you.

• It piles up into a record nobody else has. Not what you know. What you decided, when, and against what advice.

A year of that log is what you teach from. It is what you hand a new hire instead of a slide deck. For a business made of expertise, it is about as close to inventory as you get.

Find your line this week

Make this the question you ask before you build anything:

Which rung is this task on?

Try to write down the rule you follow when you do it. If you can finish the rule, all the way to what finished looks like, it is bottom rung. Hand it over and stop touching it. If you can write the steps but not the right answer, it is middle rung. Let the machine draft, and keep the decision. If you cannot write it down at all, or if it is a point where your work meets a person, it is top rung. Leave it alone.

Then draw the line on purpose. Every time, on purpose. Set it too low and you drown in work you should have handed off years ago. Set it too high and you give away the thing that made you worth choosing.

The question today is not “What should I automate with AI?”

It’s “What should I not?”

PS. I built a short diagnostic that shows where your line probably sits and which rung you are most at risk of crossing. Takes a few minutes, and it is free.

Find your line: take the AI Readiness Diagnostic

Someone reads your post. They tap a button. A number shows up next to your work: 55% AI-assisted. You made that post.

A machine helped you tidy the draft, sure, but the ideas were yours and the years behind them were yours. Now a stranger holds a number that seems to say you faked it.

Last Monday I wrote about how to keep yourself inside your writing when you bring AI into it. This is the other half of the story. Your work is not only the words on a page.

It’s everything you do around them, and that is where a reader decides whether you’re real.

New here? I’m James. My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can copy. If you want the idea underneath everything I write, start with How I Use AI To Become Irreplaceable.

What the scanner actually reads

If you’re a founder or business owner who publishes to grow your brand, a new button on Substack probably turned your stomach this month. But not to worry.

Pangram looks at one post and hands back a number. That’s it. It studies the shape of your sentences and guesses whether a machine had a hand in them.

Picture two people.

Your post scores 55% AI-assisted. You used a model to clean up the draft, and the scan caught it. Look at what you do around the work, though. You leave real comments on other people’s posts. You share the ones you admire. You send a note to set up a project with someone you respect. You answer the readers who reach out. You show up, over and over, as the same person.

There’s a second person: they write every word by hand. Their post scores a clean 0%.

Not a trace of a machine.

At the bottom of every piece, though, sits a link to a sales page. They never reply to a comment. They only show up to push an offer, then vanish. They are human on the page and missing everywhere else.

One of those two reads as real to an audience. It isn’t the one who passed the scan.

That’s the whole trouble with treating a score as a verdict. It doesn’t know why you made the post. It doesn’t know the ten years of work sitting behind the idea. It doesn’t know that you answered forty comments last week, or spent an hour helping someone who will never pay you.

It reads the page, and gives a number. But the number answers a question almost nobody cares about. Readers don’t want proof that your hands typed every letter. They want proof that a real person, with real judgment and real stakes, stands behind the work.

Five things a scanner can’t read

Your authenticity is a full-body thing. It shows up in signals that live outside any single post, and those signals are what a reader is really weighing, even when they don’t notice. Five of them matter most, and not one shows up in a scan.

1. Transparency

Transparency is you being open about why you make what you make. Your bigger mission, the reason a post exists, where you draw your lines.

Substack handed you a tool for this. Next to the scanner, writers now get a small box to explain how they made a piece.

You can say plainly that you used a model to draft, that the thinking is yours, that you’d never let a machine invent a client story. A reader who understands how you work stops guessing and starts trusting.

A score can’t build that trust for you. Only you can.

(If you want a fuller standard for keeping AI honest in your process, I laid out five rules for that.)

2. Consistency

Consistency means showing up as the same person over time and across every place you appear. Who you are in a post matches who you are in the comments, in a note, in how you handle someone who disagrees.

This is where a scanner is most blind.

It reads one post, frozen in a single moment.

Your readers read you across months. They watch whether your actions match your words. The more they see the same steady person, the more you become someone they feel they know.

A scanner sees a snapshot. Everyone else — your audience — sees the whole film.

3. Audience alignment

Audience alignment means making things for the real people you serve, not for a trend, not for a metric, and now, not for a scanner either.

You know who is on the other side of the screen and what keeps them up at night.

This is where years of real work become an edge. You’ve sat across from these people. You know the question they’re too shy to ask out loud. When you make something straight for that person’s real problem, they feel seen in a way no scrubbed, scan-proof post can fake.

4. Lived experience

Facts inform. Your experience is what connects. The client you lost and what it taught you. The launch that went sideways. The judgment you earned by getting things wrong for years before you got them right.

This is the center of the whole thing. A model can copy a shape. It can copy a tone.

It can’t copy your Tuesday. It has never sat in your chair and agonized over a decision that mattered.

When you build your real experience into the work, you hand a reader something a machine has no way to reach. Encoding that experience into a system you own is the real project.

It’s also the exact thing a competitor with an LLM subscription can never clone, because they didn’t live it.

5. Engagement

Engagement is the simplest signal and the one most people skip.

You reply to comments. You join the conversation. You treat readers as people who want to connect, not as views on a dashboard.

Go back to the handwriter with the perfect score. Their miss isn’t the writing. It’s that they talk at their audience and never with them. They broadcast, then disappear.

The person at 55% is in the comments and the notes, showing up in the room and building the one thing a scanner will never catch: an actual relationship.

Nobody can prompt their way into the trust you build one reply at a time.

Why this hits experts hardest

If you’ve spent ten or fifteen years getting good at something, and are now just talking about it with the help of AI, this scanner probably landed on you as a threat.

You tried AI to help you publish, it worked, and now a number is out there that seems to shrink all that skill into a percentage.

The mistake sits right there in the panic.

The score is the wrong thing to worry about.

Chasing a low number means editing your sentences to fool a machine, which is a strange use of a whole career. Your actual moat was never the raw text. It was the judgment behind the work … and the presence around it, the two things you built over years and the two things a scan can’t see. Most experts panicking this month are guarding the wrong door.

My last post scored 55% AI-assisted. I didn’t flinch. I know I’m human, and being human was never something a page, a score, a detector — no matter how good — could prove.

I comment on work I respect. I share what I admire. I reach out to people I want to build with. I answer the readers who answer me.

The scanner can have my 55%. Everything that makes me worth reading happens where it can’t look.

Don’t write to beat the scan that’ll now be there. Write to serve. Write to outgrow.

PS. If you want to see how much of your current setup a stranger with a prompt could actually copy, and where your real moat is hiding, take the free AI Readiness Diagnostic. It’s a short set of questions that shows you which parts of your work are promptable and which parts are yours alone.

For two years, I wrote slop for a living.

Dozens of articles a month on best laptops for students, best paints for your interior, best paints for your exterior. How to unclog a drain. Ten foods that help you sleep. All that and more.

I got a brief with a topic and a list of keywords, and I packed those keywords into the text until Google was happy. No AI anywhere near it. Just me, a list of words, and a deadline.

When AI got good enough to do that job, it did. My kind of work was one of the first things to go. And rightly so.

The keyword-stuffed article I used to create is table stakes now, automated away. It was pattern with nothing underneath it. A machine could do it because there was nothing in it that needed me.

I wasn’t in those articles. My judgment, my voice, wasn’t.

I think about those two years a lot this week, because Substack just did something big.

New here? I’m James, and I help founders and creators build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can copy.

My forever free Substack publication is where I, and the experts I invite, explore that idea of using AI to become irreplaceable.

If you want the foundation this piece sits on, start with How to Train AI on Your Brand Voice.

Chris Best against Claudefishing

On July 21, Substack’s CEO published a post called “Against Claudefishing” and rolled out a detection tool built with a company named Pangram. Readers can now scan a post, a note, or a comment and get an estimate of how much of it a human wrote by hand.

The feed lost its mind. People read this new Substack AI detector as going to war on AI, the start of a witch hunt, a machine that would drag honest writers into the town square for using the same tools everyone uses.

It was, of course, more nuanced than that.

For one, Best was careful. He said the problem is not AI. He even explicitly said not every AI-assisted piece is slop.

He said Substack is not against people using AI to help their work, and that writers should get to pick their own tools.

What he named as the crime was very specific. He called it Claudefishing, the act of faking a human on the other end of the words, so a reader spends real attention on something with no real thought behind it.

To combat this, the Substack team introduced two new features: integrating with Pangram AI detector, and a disclosure statement. Theoretically, these will give readers the transparency that they need to combat “Claudefishing.”

And for the most part, I agree with this. I have watched that con fill timelines for two years. “Claudefishing” is the same emptiness I used to manufacture by hand as a freelance content writer, only now it scales with AI. It’s what my whole brand is categorically against: promptability, replicability. Also known (in my head) as the Gray Sea Of Mediocrity, also known more simply as AI slop.

But while the principle is sound, I’m not sure that the steps are enough.

Sound in principle, noisy in practice

I agree with the principle, but I also understand those who don’t. I believe everyone’s unease is about what people will do with the tool more than the tool itself.

A detector in the hands of a crowd that already suspects everyone becomes a weapon for accusation. It’s always the way this goes with AI detectors. This has happened to me: countless employers and partners have implemented “Must be below X% in [insert AI detection tool],” ruining the (completely human, btw) quality of the original draft. It’s happened in school, as well, with students penalized from AI detector false positives.

Best was careful about this, but the narrative will get flattened. “Scanned as AI, therefore fake.”

That is the default this drifts toward unless the narrative about AI, and AI writing, shifts. If not, this would set the conversation back a long way.

Pangram is good, but how good?

The scan can see the tool being used. It cannot see you.

Pangram checks one dimension of your writing.

It reads your words and estimates whether a machine shaped them. Plenty of tools do that. Pangram is one of the better ones, at least by its own 2024 technical report, which claims near-perfect detection across a wide range of writing.

Moreover, Pangram has been independently assessed against other commercial detectors. Researchers Brian Jabarian and Alex Imas via the National Bureau of Economic Research (NBER) published a September 2025 study specifically benchmarking Pangram vs Originality AI, GPTZero, and an open-source RoBERTa baseline. It was the only detector that kept policy-grade error rates across all four AI models tested.

Though, it’s still not perfect, as ’s own testing, and ’s experiment shows (btw, he’s offering a free skill in that article).

It’s just … not perfect.

So yes, Pangram might be something … but so what?

These tools can tell you if AI is used. But it cannot tell whether your judgment is in the sentence. It cannot tell whether the piece is aligned with what you stand for. It cannot tell whether a reader on the other end is spending their attention on a real point of view or on filler dressed up to look like one.

I am proof of this gap, from before AI was even here.

I wrote soulless, hollow work for years with my own two hands, and no detector on earth would have flagged it, because a human typed every word. The absence of a machine did not make it real. It was slop with a pulse.

The gap runs the other way too.

A writer can sit with a model open, argue with it, throw out most of what it says, and end up with a piece so full of their own conviction that it could not have come from anyone else. The tool was used, yes, but only to amplify the humanity.

A score tells you a tool was involved. But that should never be the only benchmark.

So … Can you write authentically with AI?

Yes. Authentic AI writing happens when the work stays aligned with who you have said you are and what you are for. A detector reads the tool. Your judgment and your mission are what make the writing yours.

If the tool isn’t the measure, something has to be. The standard I use held up long before tools like Pangram existed.

Your writing is authentic when it stays aligned with who you have said you are and what you are for.

Your promise, in other words. The reader came to you for a specific way of seeing the world. Authentic work keeps that promise. Hollow work breaks it, whether a human or a machine did the typing.

Most of what any of us make is repeatable pattern, the parts that follow a shape we have used a hundred times. That work can be handed off. The exception, the judgment call, the moment where you decide what is actually worth saying, stays with you.

Encode the pattern. Protect the exception.

A machine can carry the repeatable weight so you have more of yourself left for the part only you can do.

The standard is easy to say and harder to live.

Three habits keep me honest about it.

1. Think critically about where AI belongs in your process

Critical thinking here comes down to knowing the trade you are making, and defending it out loud. When you can’t say why the model touched a given part of your work, that part is running on autopilot, and autopilot is where slop comes from.

Ask yourself:

• Can I explain exactly where in my process I use AI?

• Do I understand what I give up when I use it there?

• Do I have concrete reasons for making that trade?

2. Know what you’re for

Alignment needs something to align to.

Until you have named your promise, you cannot tell whether a draft keeps it or breaks it.

Get specific about who you serve and what you stand for, then hold every AI-assisted choice against it.

Ask yourself:

• Can I say in one line what I stand for and who I serve?

• Does this piece keep that promise or wander off it?

• Would the reader who came for me recognize me in this?

3. Guard your voice

Your voice is how you show up on the page, separate from your goals, your motivation, or your mood on a given day.

A model will pull you toward its own smooth default the moment you stop paying attention. So, know the rules of your own voice well enough to catch them slipping.

Ask yourself:

• Can I name the rules that make my writing sound like me?

• Does this read like me out loud, or like anyone?

• Did I let the model overwrite a choice I would have made?

The disclosure box is a gift

Substack also gave writers an optional “How I make this” statement, a place to explain your process. Most people are treating it as a confession booth. That is backwards.

That box is the most unpromptable thing on your page.

It is where you get to say what you are for and how your hand shows up in the work. A model cannot fill that box in for you, because it has no mission and no scars to draw on.

Fill it in yourself.

Tell people what you refuse to automate and why. You will find it does more to earn trust than any clean detector score ever could.

Authentic AI Writing: How I keep myself in the work when using AI

People assume that using AI to write means teaching a model to copy your voice well enough to slip past a scanner. That is the shallow version, and it is the one Best is right to be against. Voice worn as a costume.

But there’s a better, more mindful way to do things. This is what I’ve tried to do.

I have encoded myself. My values, my worldview, my years of writing experience, the things I have lived through, my specific pet peeves, the principles I refuse to break. Those are the criteria my drafts get measured against before anything ships.

This is a long way past a bot that knows my posting schedule.

So when I bring AI into a piece:

• It works like a mirror held up to my own promise

• It checks whether the writing still matches what I stand for while I juggle a dozen things at once.

• It helps me straighten a promise that has drifted.

• It pressure-tests an argument from a reader who disagrees with me.

• It widens my research past what I could gather alone, and it pushes a half-formed idea until it becomes a real one.

I’ve discussed AI writing in my Substack many times. I wrote my process for it here. I’ve also invited guests who showed their own processes:

• on Using AI for Writing Without Losing Your Voice

• on Training AI on Your Brand Voice, and

• on How to NOT Use AI for Writing

Used this way, AI leaves more of you in the work. It clears away the mechanical parts, so your attention goes to the parts only you can carry. What reaches the page comes out more concentrated.

Stay in your work

If this week rattled you, be plain with yourself about what the scan actually is.

It is not the judge of your worth as a writer. Run it on your own drafts if the number steadies you, then set it down and ask what matters:

Is my judgment in this?

Is it aligned with who I say I am?

Would a reader feel a person here?

When the answers are yes, publish and don’t flinch. Write authentically with AI, no matter what the numbers turn out to be.

The scanners will get better. The anti-scanners might get better too. The feeds will keep filling with hollow text that scales faster than any of us can write by hand.

None of that decides whether your work is real.

The way through is to be so present in what you make that the question of which tool touched it stops mattering, because a reader can feel you standing there.

I spent two years absent from my own writing, back when it was just a way to earn a living. I am never going back, and no button on a platform is going to send me there. If you are in your work, you are safe. Stay in it.

So, you know that I just recently moved out of my hometown. What I didn’t tell you is that in the place I moved into, the toilets didn’t work. Two bathrooms, both with non-functioning toilets.

I’ll spare you the gruesome details, but I’ll leave you with this: I consider myself a pretty handy guy (as many people probably do). I know how to figure stuff out, know my way around tools, all that. But I took one long look at that system and went, “I need a plumber” (also, I thought, “I need to argue with my landlord,” but that’s a different story).

There are problems we know how to fix. There are problems we’ll eventually figure out. And there are problems that we can figure out, but really don’t have to.

Unfortunately, this is how we’re talking about AI.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can replicate. Here’s an example of my work, turning a therapist’s 20+ years of experience into an AI-powered diagnostic.

The fantasy of the all-powerful tool

AI has taught everyone that they can do everything themselves, or rather, that they can do everything by asking AI themselves.

This has some truth to it. AI multiplies what you can do by a factor of 100x. Writing for non-writers, coding for non-coders, analytics for the non-analytical.

It also has limits, and this is what AI influencers fail to acknowledge.

Your capability is only limited by what you know to prompt in the first place. If you don’t know that something is possible, it won’t be possible. If you don’t know that something HAS to be done, it won’t be done.

Now apply that to an AI project inside your own practice.

You’re a coach, a consultant, or any number of credentialed expert, thirty years of practice behind you. You decide to build an AI assistant that talks to leads the way you would, because you know AI can do it and you’ve watched other people pull it off.

Three months in, one to two hours a day, and prompting turned out to be the smallest part of it. The thing answers questions about your own framework and gets it wrong in ways that make your chest tighten. You’ve realized that deploying it introduces practical and legal risks that you haven’t even heard about before then.

It’s not ready at all. And worse, those hours came out of the practice that pays you.

You’re becoming a real builder, for sure, but it’s costing you time, energy, and money. When all you wanted was the product.

When you should build your own AI

Don’t get me wrong, I’m not saying you shouldn’t ever tinker with AI. You should definitely build something cool with it, some time.

But with how powerful AI has become, and with how much influencers hype it, many founders believe it’s the answer to their business problems. In many cases, it’s not. It can be, but you need to know when, and where.

Here are three rough criteria to decide this by:

1. When experiments light you up. If you have ever lost a Saturday to a workflow and looked up at 9pm happy about it, you’re the person these courses were built for. That feeling alone could be the payment for a long stretch of frustrating work. I took a version of that road myself, and it gave me a career.

2. When you have time and energy beyond your practice. The real cost was never the course fee. It’s the hours after client work, when the judgment your clients pay for sits idle while you debug.

3. When revenue doesn’t depend on it. Learning to build pays back in seasons. If this quarter’s number is the thing keeping you up at night, building an AI workflow yourself will likely just add to your problems. It’s the wrong road this quarter. It might be the right one next year.

AI educators are not lying to anyone. They’re just handing out maps to a journey suited for a specific kind of people, ones with disposable time, energy, and income. As founders, business owners, and creatives, you might have all these three — and that’s perfect.

But in many more cases, experts don’t really want to be doing more than what’s already making them money.

Where self-built AI hits the wall

Something happened to small business AI adoption between 2024 and 2025 that nobody selling courses talks about. It went backward. A survey of 1,500 small business owners found usage fell from 42% to 28% in one year. Millions tried the tools, and enough of them quietly walked away to drag the national number down.

The walking away is expensive. When a project dies at a small business, the sunk cost lands between $25,000 and $100,000 in time and abandoned work.

Bigger budgets tell the same story. MIT’s NANDA project found that 95% of generative AI pilots inside large companies produced zero measurable return, meaning nobody could point to a number that moved after the money was spent. The same research showed systems built by outside specialists reaching deployment at about twice the rate of in-house builds. And S&P Global found enterprises abandoning most of their AI initiatives jumped from 17% to 42% in a year.

Teams with engineers on payroll and dedicated budgets stall when they realize the complexity of what they’re trying to accomplish, and how AI prompting is just a tiny step in this process.

You’re working solo, after client hours, with none of all that.

When should you hire outside help?

Experiment all you want. Prompting, testing tools, seeing what a chatbot can do with your intake questions. That work is cheap and it teaches you things.

Past a certain point, though, you’re not experimenting anymore. You’re running a project. And a project needs someone who has finished one before.

That someone is a builder, not another course. A course sells you capability. A builder sells you the thing itself. Both are real products and both have their place. The question is which one you’re short on. If you’ve already bought capability twice and your system still doesn’t exist, you were shopping in the wrong aisle.

So where’s the line? Three criteria draw it.

1. The outcome is public

Some AI work fails privately. Your summarizing workflow breaks and you lose an afternoon.

Some fails in front of the people who pay you. For example, I built an assessment for a therapist that her audience takes before their first session. It’s the first thing they experience of her practice, and it runs on twenty years of clinical judgment. That build had to be right, because a bad version wouldn’t just be useless. It would have told a hundred strangers something untrue about who she is.

Private failure costs you time. Public failure costs you the thing you spent twenty years earning. Here are some questions to ask about the build:

• Will a client, a lead, or a reader touch this directly?

• If it produced something embarrassing tomorrow, who would see it?

• Does it speak in your voice, or answer questions people are trusting you to answer?

2. The cost of delay is real

Pain is a bad signal on its own. Pain at 11pm is what makes you buy material you will abandon once your practice consumes your time.

The better question is what another six months of this costs you. Not how much it hurts today, but the hours, the dropped follow-ups, the leads that went cold while you meant to get to them.

Run the calculations here. It’s usually worse than you think and it makes the decision for you.

• How many hours went into this last month?

• Multiply by twelve. Would you pay that to make it stop?

• Have you already paid for a tool that was supposed to fix it?

3. The domain knowledge is far from your practice

You could learn retrieval systems, embeddings, evaluation, and the rest of it. You can learn how AI works for marketing.

It would take months, and at the end you’d hold a skill that has nothing to do with the work your clients pay for.

When the knowledge is far enough away, you can’t tell AI what you need either. Your capability is limited by what you know to prompt in the first place, so a gap in domain knowledge is a gap in your prompts. AI will answer the question you asked. It won’t tell you that you asked the wrong one, and it won’t mention the thing you didn’t know to ask about.

That’s how three months disappear. The tools worked fine. Nobody in the room knew what was missing.

• Would learning this build a skill you’d use again after this project?

• When the output looks wrong, would you know why?

• Could you tell a good version of this from a mediocre one?

“But that will make me dependent”

Hire a builder and you depend on the builder. I hear this one a lot, and it deserves a straight answer.

The plumber leaves. The pipes stay in your walls, the toilets stay fixed, carrying your water.

A system built from your expertise works the same way. Your frameworks power it and your judgment shapes every answer it gives, which means it belongs to you the way your pipes belong to you and no plumber ever will.

Compare that to the dependency nobody talks about: an expert running client work through generic tools that hold none of their thinking, where every output could have come from any stranger with the same subscription.

There is a word for that kind of output: promptable.

It carries nothing of you, and clients feel that even when they can’t name it. A system built from your archive and your judgment makes work no stranger can reproduce.

Unpromptable.

That word sits on the masthead of this newsletter because it’s the point of hiring a builder: you walk away holding a moat, and the moat is made of you.

Done right, hiring is how you end up owning more.

Ride the wave, charter a boat

I never learned plumbing. I made a call, argued with my landlord (although argued is a strong word, he had no leg to stand on, knew it, and paid for everything but the hassle), and by the end of it the problem was gone.

I got two working bathrooms and my attention back. The plumbing knowledge stayed with the plumber, where it belongs.

Building your own AI system or product is exciting, like a roller coaster ride, filled with its own ups and downs. But having a system built for you is smoother, quieter. The follow-ups happen without you. The intake runs itself. The knowledge that lived only in your head now works hours they don’t.

That’s the outcome worth wanting. Your own expertise, running without you in the room.

Just because you can doesn’t mean you should. You can build it. Give it enough Saturdays and you’d likely finish. The question is whether the version of you who finished it is better off than the version who spent those Saturdays on the practice that pays.

A system built from your expertise belongs to you either way. You don’t have to be the builder to own it.

You don’t have to be swimming, or drowning, to ride the AI wave. You can charter a boat.

If this way of thinking is useful to you, subscribe. Two essays a week on building systems from expertise, real numbers and sideways builds included.

PS. The AI Readiness Diagnostic takes 5 minutes and tells you whether an AI project makes sense for where you are, including when the answer is no.

I lost my writing clients to ChatGPT.

I didn’t lose them to a better writer or an agency, but to a bunch of code. The client worked out that the thing I sold him, an okay article on a subject he chose, could be produced in thirty seconds by a machine costing less than my invoice. He was right. What I sold was a format, and formats fall first.

That year taught me the only rule I trust online now. Whatever a machine can produce in bulk gets cheap. Whatever it cannot produce at all gets valuable.

Which brings me to you, and the advice you keep being handed.

New here? I’m James. I help founders and experts turn what they know into AI systems no one else can replicate. Here on Substack, I write about Unpromptability, exploring what it takes to become irreplaceable in the age of AI.

If you want the philosophy underneath all of it, start with How I Use AI To Become Irreplaceable.

The job description you never applied for

You have twenty years of practice behind you. Clients who found you by referral. A body of judgment you can’t fully explain because you live it every day.

And you know, in the uncomfortable way people know things they’d rather not, that referrals are thinning and the people who need you are looking online first.

So you go looking for how to be seen online, and the internet hands you a job description. Post daily. Show your face. Ride whatever is trending this week. Share the morning routine, the humble setback, the lesson you learned at 5AM.

Build a personal brand, which somehow always means performing a personality at scale.

I want to be precise about who I mean, because “influencer” is a wide net and I’m not throwing it over everyone.

I mean the business influencer. The personality-first operator whose whole apparatus exists to funnel attention into something at the bottom. The course. The supplement stack. The cohort. The coaching program about running coaching programs. The daily text post engineered to feel like a confession, the curated life, the ragebait dressed as a strong opinion. Every piece of it is intimacy, manufactured on schedule, pointed at a checkout page.

That’s the job you’re being told to take. And your gut says no, and you’ve been treating that as cowardice.

It isn’t. It’s judgment. Sit with it.

The intimacy is the product, and the intimacy is being faked

Here is the part almost nobody selling you the influencer playbook will say.

Attachment is what an influencer actually sells. Not information, which is free. Not expertise, which they often lack. The audience keeps coming back because it feels like they know the person, and that feeling is the entire inventory.

Which means the whole model rests on the audience believing the intimacy is real.

Fake intimacy damages trust

Researchers at USC’s Marshall School published a study in the Journal of Consumer Research this year, combining behavioral analysis on TikTok with eight preregistered experiments.

They found that when content carries an AI disclosure, engagement drops. And the drop had a specific cause. It wasn’t concern about quality or a general dislike of AI. The parasocial connection weakened. The audience read the disclosure as low effort, and the bond that was the product quietly dissolved.

Manufactured closeness only works while nobody looks at the machinery.

Most influencers use AI to manufacture this closeness

And the machinery is now everywhere. Originality.ai ran long-form LinkedIn posts from ninety-nine influential voices through AI detection and found that more than half of what those accounts published in 2025 was likely machine-written.

The people selling you authenticity had already outsourced theirs.

The platforms noticed. LinkedIn published Keeping conversations real this year and rolled out the changes the industry now calls the Authenticity Update, going after engagement bait, automation pods, and the inauthentic behavior that built half those accounts.

The tactics that made the playbook work are now the tactics that get it buried.

AI content is giving diminishing returns

Meanwhile the ground is giving way underneath everyone still playing.

Richard van der Blom’s Algorithm Insights research puts LinkedIn views down around 50% year over year, engagement down 25%, and follower growth down 59%, with average organic reach falling to 8 to 12% of followers, from 15 to 20% the year before. Over on TikTok, Socialinsider’s benchmarks show views down 23% year over year and follower growth down 33%, with the smallest accounts hit hardest.

I’ll be honest about the limits of this. Influencer marketing spend is still growing. The industry is NOT dying.

What is happening is worse, especially for the person doing the work without the AI crutch.

The price of admission keeps rising, the return on each unit of effort keeps falling, and the exact behaviors the playbook demands are now both mass-produced by machines and actively suppressed by the platforms.

You are being handed a job description that is being automated out from under the people who wrote it.

Decline it.

How experts build authority without performing

There is another way to gain online visibility, and it starts with a different question. Instead of asking how to perform, ask what a stranger with a prompt could never make.

I call that being unpromptable. It’s the only thing left to do.

A machine can produce a daily post that sounds like you. It can produce a vulnerable confession, a contrarian take, a carousel about the five lessons you learned from failure. What it cannot produce is twenty years of your practice, the passion burning behind every year of work, the grudge against your industry that you’ve been carrying since time immemorial.

Three parts of your working life sit permanently outside its reach. None of them require you to film your breakfast.

1. Use your practice as material

The influencer runs on lifestyle. The day becomes the show, and eventually the wall between your life and your feed comes down.

Experts hear that and conclude they have nothing to publish. That’s the wrong conclusion, though drawn from a healthy instinct. You don’t have a lifestyle worth broadcasting. You have something better: twenty years of pattern recognition that other people would pay to learn, or borrow.

Mine came out of that bad year. I want experts to turn what they know into systems nobody can take from them, so that nobody else gets replaced the way I was. A machine can imitate my sentences all day. It cannot hold that grudge, keep that mission burning.

That conviction is the only moat left in the AI age, and it’s the one thing the flood can’t touch.

Yours is sitting in your caseload. The mistake you watch clients make in the first session, every time, for two decades. The industry advice you know is wrong and can prove is wrong. The thing you tell people at the end of a consultation that lands like a small detonation, because you’ve refined that sentence over hundreds of repetitions.

Start capturing it. Open a file and keep it open during your working week.

• Every time a client says something that stops you, write it down in their words.

• Every time you catch yourself explaining the same thing for the hundredth time, write the explanation down.

• Every time your advice contradicts what the industry says, write down both and note why you’re right.

Do that for four weeks and you will have more material than you can publish in a year. It writes itself, because you already said it out loud. You just never wrote it down.

None of that comes from your private life. It comes from your working life, which is where the good material was hiding the whole time.

And it filters for you. Lifestyle content collects people who are bored. Practice content collects people who have the problem you solve, and those are the only people who were ever going to hire you.

2. Run your own equation

Influencers need scale because of how they get paid. Sponsorship rates are priced on reach, so a bigger number is the business model. Chasing followers is rational when you’re selling attention wholesale.

Now run your own numbers. Write down what you charge, and how many clients you need this year to have the year you want. Most experts have never written that number down, and it shocks them when they do.

A therapist with a full caseload needs maybe twenty-five clients. A consultant with a good year needs eight. A surgeon, a fractional CFO, a specialist advisor, most of you are working with a number small enough to fit in one room.

Ten thousand followers is not a goal against a number like that. It’s noise. You could have four hundred readers, and if thirty of them have the problem and trust you completely, your calendar is full and your business is done.

So stop performing for a crowd and start hosting a room. Keep a short list of the people who have told you something real about their situation. Twenty names is plenty. Reply to every one of them properly, in full sentences, using what they told you last time.

More: write your next piece for the person on that list who has the problem most clearly, and let everyone else watch you do it.

A room also survives things a following can’t. Reach gets throttled on a random Tuesday for reasons nobody explains. Platforms rise, platforms rot. Borrowed attention vanishes the day you stop paying for it.

The people who trust you will follow you anywhere, including off the platform entirely. In that way, relationships are your strongest defense against being replaced.

3. Set an anti-hustle pace

The treadmill has one speed. Post daily or decay. Feed it through illness, grief, family, and every other thing life does to a person who has an actual practice to run. Burnout there is almost a certainty.

I’m moving across the country right now. Half my life is in boxes and every day hands me a small emergency. By the influencer playbook, this is a goldmine. Film the packing. Sell the stress before the feeling cools.

I didn’t. I have five rules for weeks where your life falls apart, which can be boiled to the phrase “just try to do something”. The most important thing of the week ships and everything else drops without guilt. That rule is why the newsletter hasn’t completely decayed while my life is basically falling apart.

You need a pace you can hold in your worst month, not your best one.

Two thoughtful pieces a month, held for two years, builds more than daily posting for a season followed by a year of silence. Trust is built on a pattern, and a pattern only counts if it survives the bad month.

There’s a second reason, and it matters more than the first. Nobody hires a frantic advisor. Steadiness reads as competence, because it usually is. The expert posting through a panic spiral is teaching the audience something, just nothing that helps them.

Pick the pace you could keep during the worst month you can imagine. Then keep it.

When you stop performing

The dread lifts first. Publishing stops feeling like a performance and starts feeling like talking about your work, which you already do all day for free.

Then the work changes. A stranger with a prompt can generate infinite morning routines and endless contrarian takes. They cannot generate the framework you built out of a failure that still stings, or the sentence a client said to you last Tuesday that you’re still turning over. That gap is how you build authority without becoming an influencer, and it widens every week the flood rises around everyone else.

Then the business changes. Influencer growth is a spike you re-earn every morning. This kind is a slope. Slower, steadier, compounding, and attached to you rather than to a platform’s mood.

You already have what this requires. A practice worth twenty years, judgment nobody can download, and a small number of people who need exactly what you know. If you want to pressure-test where you stand, the Unpromptability audit will tell you in an afternoon.

The visibility was never the hard part.

The costume was, the performance was. Take it off.

P.S. I write for experts who want to be visible without becoming a personality. If that’s the road you want, subscribe and walk it with me.

Subscribe now

My whole life is in boxes right now.

I moved across the country. Since the last week I’ve been here, my house has been a maze of cardboard, and the newsletter went quiet while I packed up one life and set down another.

If you have started something big lately, you might know this feeling. I have a floor covered in half-unpacked boxes, you might have a messy inbox and a week full of hour-long meetings. But all the same -- the quiet worry that everyone else kept moving while you fell behind is there.

That’s why I’m writing this; to remind you that you can get through this big thing gracefully, and to remind myself that I can, too.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace, by turning your expertise into something no one else can copy. I write 2x week deep dives on Substack, and am looking for writers to collab with. If you’re interested, let me know!

Also, if you have a build you keep meaning to start, read AI Made Building Easy. So Why Can’t You Start?

A big change floods you and pulls you off course

A move is a machine for making small tasks. Set up the utilities. Learn the commute. Assemble the furniture. Sort the mountain of packaging that appears from nowhere. None of it can wait, and all of it eats the hours you meant to spend on the work that matters.

A big build does the same to a founder. Stand up a new business, or roll out your first real AI system, and the daily minutia arrives in the same flood. Stakeholder messages pile up, and a design call you cannot skip always lands at the worst time. The work you can point to on a map gets buried under the work you cannot.

If you are standing up a new business, or rolling out a change as big as your first AI system, you already know this floor. Same if you are a creator launching or rebuilding while everything else is unsettled. The Big New Thing demands everything while your old routines fall apart underneath you.

That flood does two kinds of damage.

It pulls you off your direction so slowly you do not notice until you are weeks off course. And it makes you feel so far behind everyone else that you are tempted to stop showing up at all.

What five weeks of upheaval taught me about not stalling

I have been in this transition since around June, when a week-long event pulled me offline and the move swallowed everything after. Long enough to watch which habits kept me upright and which ones let me tip over.

The first thing a big change threatens is your sense of where you are going, so the first lesson I learned was to protect your direction. The rest of what it threatens is whether you keep moving without burning out or quitting on the thing entirely. Thus, the next four protects your momentum.

1. Don’t lose sight

A big transition is when it pays to have a north star.

Most days, the work in front of you will try to become the whole world.

Anyone who has done a cross-country move can tell you how total it feels. The setup list never ends. The environment is new, the people are new, and you are building routines from nothing. It crowds out the reason you moved in the first place.

A build does the same. You get pulled into debugging, stakeholder updates, and the hundred small design choices that cannot be ignored, the same way I cannot ignore the furniture and the garbage bags. The work is real enough, but it can also pull you off-course.

A mission is how you see through it. That’s why it’s the first Pillar of my Unpromptability philosophy -- a mission that you truly care about.

When you know the one change you are here to make, you can look up from the daily handiwork and check whether you are still pointed at it. You catch the drift in week two instead of month three, and course-correction early will only cost you a small turn.

2. You’re more than your latest post

I have felt behind since June. First the event, then the move, and all the while Substack kept moving without me. Everyone in it going forward with life as normal while I sorted boxes.

It is hard not to feel behind when the timeline never pauses for you. The plain version is basic, and true anyway. You are more than your latest post, and more than your latest ship.

It is fine to drop the ball on your online presence when the rest of life gets loud. A founder who goes quiet mid-build has not stalled. The work just moved somewhere the timeline cannot see it. Your health and your relationships outrank the feed every time, and so does anything you are actually learning.

Your energy will ebb and flow, and so will your attention. You will not run every project, AI or otherwise, at full power all the time. That is normal.

You still have your north star. When you have been heads-down for a while, look up. At the very least you will know which way to go.

3. Nobody’s watching that closely

The move wrecked my publishing and my community engagement, and it made me anxious. Most of that anxiety traces back to a vague, faceless “them.”

I worry “they” will be put off by my absence. Readers will notice I went quiet. Potential clients will forget me. Peers will think less of me.

Most of it is not real. People are not paying as much attention to us as we imagine. The market is not studying your half-finished project the way your own head is at 2 a.m.

And the few who are paying real attention tend to have your back. They have had their own setbacks and busy seasons, so they give grace for the things you cannot control. Give yourself the same. Nobody out there is deducting points because you missed a week.

4. Half-assing is underrated

Do things sad. Do things scared. Do them lonely, and do them badly. Do them imperfectly.

I keep repeating this to myself, because my place is in shambles. Boxes open and unopened, laundry starting to climb the walls. On days like that it is easy to wait for a wave of motivation to lift you back into action. But that’s a trap. Big transitions and big projects keep pulling you along, and they leave no gap for motivation to catch up.

So move imperfectly. This is where a build rewards you fast. Ship the rough version. Push the ugly first automation that still saves you an hour a day. Commit the code that works before it is clean.

A half-done system running in the world beats a perfect one still living in your head.

Something is almost always better than nothing.

5. Always pick yourself back up

A friend once told me why he regretted his gym membership. “I knew I shouldn’t have signed up,” he said. “I know myself. Once I miss a day, I never go back.”

I have a rhythm I hold myself to. Two articles a week, every week. When I miss one, the pull to just not start again is enormous. Right now, in the middle of all this, it is stronger than ever.

It is not only the newsletter. There is a build I set down when the move started, an experiment I told myself I would reopen and have not touched since. The unfinished thing gets easier to abandon the longer it sits.

Get buried in the day to day, run at less than full power, drop the ball on the project and the publishing both. All of it is survivable. But remember: the successful did one thing differently: they picked themselves back up.

At some point you just have to tuck your chin and do it. I have to. Writing this is me reaching.

What to carry out of your own mess

You might be in your own version of this. A launch mid-flight. A system half-built and waiting. You might not have the moving boxes that I do, the half-built furniture, and the second-floor that, as I discovered, had such bad ventilation that it feels like an oven at noon. Your problems will look different, but swimming through it might feel the same.

Write one sentence for your direction, the change you are here to make, and keep it somewhere you will see it on the hard days.

Let yourself go quiet when life gets loud, without reading it as failure. Move imperfectly, and let the rough version ship. Keep one rule for restarting, so a missed week never turns into a missed season.

Do that, and you keep your direction and your momentum through the part of the year that takes both from most people. That is most of what makes someone hard to replace. You kept pointing the same way, and you kept building while the mess ran its course.

My life is still in boxes. This one is as much for me as it is for you.

Tell me the Big New Thing you are pushing through right now. I want to know what my people are building while everything else is up in the air.

P.S. If you have been quiet lately and felt bad about it, take this as your permission slip. Look up, find your direction, and move imperfectly.

Michael Thompson sat on a goldmine of content and couldn’t dig into it.

Over a decade he had built something most people — most writers, even — never will.

A hundred recorded transcripts, with sixty more nearly done and more than that in the works. Primary writing. Secondary writing. And underneath all of it, hundreds of books’ worth of quotes and references he had pulled out by hand, one passage at a time, across years of research and study.

He didn’t come to me because he was out of things to say. He came to me because he couldn’t reach the things he had already said.

If you have made a lot and can no longer find most of it, this is your story too.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace, by turning their expertise into something no one else can copy. This piece is built on a real system I designed for a client with a decade of work behind him.

For the idea underneath all of it, start with The Only Moat Left In The AI Age.

The problem nobody warns you about

Most people who want to write fear the empty page. They think the hard part is having something worth saying.

The people who have been at it for years have a different issue. They have said plenty, but it’s buried.

Michael writes on illumination and practical mysticism, a field so narrow that on most days he is the only person in the room who knows it well. There is no colleague down the hall to check his thinking against, little community outside of the ones he’s in already. He has spent years recording ideas, refining them, linking one to the next until the whole body of work became a single connected thing. And almost all of it came before LLMs could read a word.

So when he wanted to write something new, he ran into a wall of his own making. Where did I cover this before? What did I say that connects to it? Which transcript, which essay, which page? The answers all existed somewhere in ten years of files. Finding them ate hours he didn’t have.

A new idea arrives, and Michael half-remembers saying something close to it three years back.

So he opens folders. He scans transcripts. He reads his own old paragraphs looking for the one line he needs, and forty minutes later he has found a different good idea and lost the thread he started with.

The work was all there, a goldmine of information. The path to it was gone.

That is the strange cost of depth. The more you make, the harder it gets to reach. A small library you can hold in your head. A large one turns on you, and starts hiding things from the person who wrote them.

The AI idea and translation system we built

Underneath everything is the foundation: his entire body of work, broken into pieces and embedded so the machine reads it by meaning instead of by keyword. Plain file search needs the exact words. He never remembered the exact words. Every transcript, his primary and his secondary writing, the full stack of references he kept. Ask it about an idea and it finds each place he wrote about that idea, even when he reached for different words every time.

On top of that sits his voice. The system learned how Michael writes from Michael’s own writing, so when it helps him reshape a passage, the words come back in his cadence, built from his own patterns. It captured his voice, all Michael.

Then there’s the library. Every quote and reference he ever pulled, held with its source and its page, formatted the way he likes it, searchable in a breath. The collection he built by hand over years finally sits in one place that answers when he asks.

And around all of it, the front end: the room he works in. Two panels. His past conversations on one side. The AI in the middle.

The first time he ran it the way he meant to, he asked for an article referencing both internal and external sources. It came back with his exact brief, with sources across years of work, that he had half-forgotten making. The forty-minute dig became a five-minute answer. He kept his thread this time.

The part I’m proudest of cost me money to say.

Early on I told Michael he might not need a custom build at all. A good setup inside a twenty-dollar AI subscription would cover most of what he does on a normal day. The full system only earns its price if he plans to turn it into a product for other people in his field. If this stays personal, I told him, keep your money.

He chose to build anyway, with the product in mind. I wanted him spending on the right thing, for the right reason. A builder who only ever says yes is a salesman. I would rather tell you the truth and build you less.

Every use pointed at work he’d already done

Once he started living in the thing, a pattern showed up that I couldn’t unsee. Every job the AI did for him reached backward, into work he had already finished. Not one of them gave him a new idea. Look at the five.

1. Indexing what he’s already written

He asks where he has touched a topic before, and the workspace lays it out. Every transcript, every essay, every offhand line that connects to the thread he’s chasing.

Work he wrote years ago, surfaced in seconds instead of the hours it used to cost him. The AI knows nothing about western mysticism. It knows exactly where Michael’s thinking on it lives, and it hands him back his own words at the moment he needs them.

2. Curating his outside references

For years Michael read widely and pulled quotes by hand, hundreds of books’ worth, each passage chosen because it struck him as worth keeping. That curation was slow, careful, human work.

The system now holds every reference with its source and page, searchable in a breath. He selected those passages with his own judgment. The AI simply remembers where they all are, so the library he built by hand finally behaves like a library.

3. Bouncing ideas off a wall

Working alone in a narrow field is a quiet kind of work. No one nearby knows the material well enough to push back on it.

So Michael uses the AI as a wall. He throws a half-formed idea at it and watches what bounces back, adds volume to a thin thought, tests whether an argument holds its shape. The ideas are his. The wall gives them something to hit, which is more than a blank document ever did for him.

4. Summarizing his longform

Michael writes long, and he writes long well. Ask him to cut a piece down to a paragraph and he’ll lose an afternoon to it, and the result still won’t sing.

The AI does it in a moment, and the moment beats his hours. He told me this plainly, which most experts never will. Writing long and compressing it short are two different skills, and he only owns the first one. The AI summarizes the work.

5. Translating his work for newcomers

He writes for readers already deep in the subject. A newcomer opens one of his pieces and drowns by the third paragraph.

So he hands the AI a dense essay and asks it to widen the door, to add the context he normally leaves out because his usual reader doesn’t need it. His voice stays intact. His ideas stay intact.

The AI only builds the on-ramp, so someone standing outside the field can find a way in. The thinking was already on the page.

Why this isn’t AI writing for you

Not one of these five uses put a fresh thought in Michael’s head. Every one worked on material he had already made: his writing, his quotes, his arguments, his voice. The substance was his before the machine ever touched it.

That is the whole truth about AI and people with real depth, sitting in a single pattern.

When you have nothing yet, AI generation feels like a miracle. Type a prompt, get words. For someone with ten years behind them, generation is the weakest thing AI has to offer, because the substance already exists.

The work is reaching it, ordering it, opening it up to other people. Your body of work is the moat. The AI is the index and the mirror, never the author.

What this means for your archive

If you’ve been making things for years, you are probably sitting on the same goldmine of content Michael was. A body of work so large and so connected that you can no longer reach into it fast enough to use it.

Your instinct will be to make more. Resist it. The most unpromptable thing you own is already written. No one can copy a decade of your thinking, the way your ideas wire together, the references only you thought to keep. That archive is your asset. The one move that matters now is making it reachable.

AI will not hand you a mission, or a voice, or a single original idea. You did that work already, over years no prompt can shortcut. What it can do is build you the room where all of it sits within arm’s reach.

Michael spent over ten years building something no prompt can replicate. The system I built him adds nothing to the pile. It lets him stand inside it.

That is the version of AI worth wanting. The one that makes your own depth usable again, instead of quietly telling you depth stopped mattering.

P.S. If you’ve got a real body of work and you can’t reach it, that’s a system worth building, and it’s the kind of thing my team builds for founders and creators every week. Build your system now!

P.P.S. Michael is a knowledge worker, vibe coder, practical mystic, and entrepreneur. He writes the newsletters Illumination and Knowledge Worker, check him out!

Subscribe now

Everybody hates AI generated comments.

A few months ago, a Note about how bad AI comments are went viral on Substack. I don’t recall who posted it anymore, but I remember hundreds of people piled into the replies to agree.

That many people, agreeing, all saying the same thing: we can tell, and we hate it.

Meanwhile, the platforms are starting to throttle and ban automated comments outright. The verdict is in. Nobody wants a place where the relationships are machine-made.

And yet people keep reaching for that ever-so-magical automation button.

There’s a better way. A harder way for sure, but a way that’s more intentional, grounded, and best of all, effective. I have five steps to offer below, so read on.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace, by turning your expertise into systems no one else can replicate. My mission is to help you become irreplaceable WHILE using AI, not in spite of it.

This article discusses something I built that serves that mission: the Daily10 web extension. For more information on what it is, why I built it, and how it impacted users in my community beta test, read these:

• How To Stop Letting The Algorithm Decide Who You Engage With

• What I Learned About Intentional Networking From Beta-Testing A Habit Tracker With 13 Strangers

• Your Relationships Are Your Strongest Defense Against Being Replaced By AI

Automated engagement: The trap that creators love to fall into

Automated engagement. You connect your account, point it at a feed, and it develops relationships for you while you sleep. Or you copy the article, plug it with a prompt, and then comment what the AI generated.

For many founders and creators whose business relies on building relationships, this sounds like a dream come true. Or at the very least, a dream appealing enough that they grab for it again and again, turning auto-likers and auto-commenters into a million dollar industry.

But.

Some of those comments get your account flagged or banned outright. The rest do something worse: destroy trust.

People reach for it for a reason I understand. They know relationships grow a business. They also have no system for keeping up with relationships by hand, so they hand the showing-up to a machine. The machine shows up. It never connects. And the person on the other end feels the difference in about a second.

A suspended account is the smaller cost. The bigger one: the people you were trying to reach now trust you less than if you’d stayed quiet. A bad comment is worse than no comment.

You taught them that you’ll let a bot speak in your voice.

I’ve said for a while that this was coming. You can’t automate belonging. You can’t prompt real trust. Someone can copy your posts, but they can’t copy your relationships, the comments, the banter, the shared history.

I believed that when I wrote it. I just hadn’t built the thing that turns it into a daily practice instead of a nice idea.

Now I have.

Why the friction is the point

The tool is a web extension called Daily 10, and I’ll tell you upfront what it doesn’t do: It will not write your comments. It will not engage for you. It is, on purpose, not easier than the automations you may have already tried.

Because in relationships, the friction is the point.

Think about a gym. Nobody wants a machine that lifts the weights for them. The whole reason you walk in is to do the reps yourself, because the reps are what make you stronger. A robot that did your sets would leave you exactly as weak as the day you started, with a great-looking exercise log.

Automated engagement is that robot. It does the sets. You get nothing.

So I drew a hard line down the middle of the work. The tool handles the friction that never should have been there: remembering who to show up for, finding them without digging, tracking what you’ve done. You keep the friction that is the entire point: reading the post, writing the comment that proves you read it, building the relationship brick by brick.

The tool owns the logistics of memory. You own the relationship.

How to build real relationships online

Connection is a muscle. You train it in five rungs, climbed in order, because skipping one breaks the rungs above it.

Step 1: Choose the right ten

Everything starts with picking the right people, and most people pick wrong.

They choose the loudest accounts in their feed, the big names, whoever the algorithm put in front of them that morning. Then they spend their energy chasing attention from people who will never matter to their actual work.

Your ten should be the people who move your mission. Start there.

Name your mission, name what you’re trying to do over the next few months, then work backward to the specific humans tied to it. This will become your community. Some will be peers who do what you do from a different angle. Some will be the exact people you want to serve. A few might be a step or two ahead of you.

Ten is small on purpose (though it could be 5, even 3). You can’t show up for fifty people like a human. You can show up for ten.

Daily 10 starts you right here, with a simple, guided assistant that walks you through it. It asks about your mission, your current goals, and who actually matters to them, and you come out with a real list instead of a vague plan to network more.

When I built my own list, I filled up with mission-oriented founders, peer builders, and creatives navigating the AI transition. The people doing serious, mission-driven work in my corner of the internet, not the biggest accounts.

Step 2: Save them where you can reach them

The smallest obstacle kills your intention to engage.

If showing up means hunting through the feed to find someone, you won’t keep doing it. Friction at the start of a habit is how the habit dies.

So make your ten reachable in one move. No searching, no scrolling, no digging through a bookmark folder you’ll never open again.

That’s the core of Daily 10. You save each person’s profile and photo once, and from then on you click straight to them. The people you chose stay one click away, instead of buried under everything the algorithm would rather show you.

Step 3: Show up, daily or weekly

Showing up is the part no tool can do for you, and it’s simpler than people make it. Go to the person. Read what they actually posted. Leave a comment that could only have been written by someone who read it. React, share, add a thought.

That’s it.

Do it daily for the few closest to your work. Weekly for the rest. The cadence matters more than the volume. Ten thoughtful touches a week will build more than a hundred drive-by likes.

Daily 10 helps you log each interaction as you go, so showing up stops being invisible. You can see the work you’re doing (or not doing) instead of wondering whether you’ve talked to someone in the last month, and you can get as detailed as you like so you know where the relationship is at.

The comments and the banter are the whole asset. Most of the relationships I value most started in someone’s reply section, mine or theirs, one real exchange at a time.

Step 4: Stay consistent

Consistency is where good intentions go to die.

The first week is easy. Week six, when nothing has obviously paid off yet, is when people quit. Relationships rarely break in a dramatic moment. They cool slowly, while you’re busy, until one day you notice you haven’t spoken in months.

A system has to do two things here. Make the invisible progress visible, so you can feel it working. And warn you before a relationship goes cold, while there’s still something to save.

Daily 10 scores your interactions over time and flags the relationships that are cooling, so you catch them early.

It also motivates you to go on a connection streak — letting you track how many days in a row you’ve completed your roster. This was intentional, a little bit of gamification to improve your consistency. Plus, there are built-in features to keep it from being too punishing, which sounds small until you get one super busy day, or forget to interact on the weekends.

Going back to the gym analogy, you don’t build an athletic body from one heavy session. Not even from one week, or one month. You get stronger from showing up again and again, over months.

Step 5: Move it to the next level

Relationships don’t — shouldn’t — stay flat. They either deepen or they fade, and the deepening happens through bigger commitments.

Think about how this works everywhere else. Say in your romantic relationships. You don’t go on lunch dates forever. At some point you take a trip together, meet the family, build a life. Each step is a larger commitment than the last, and that progression is a healthy thing.

Online relationships work the same way, and most people never let them graduate. They like and comment for a year and never once reach for the next step. So the relationship plateaus, then drifts.

At some point, you’ve earned the right to propose more. A collaboration with a peer. A real conversation with someone you’d like to work with. A question for a mentor.

This is a relationship moving up a level, the same way every good relationship does.

Daily 10 calls these Relationship Gates. Once a relationship reaches a certain level of real, consistent contact, the tool surfaces the next step, so you propose it when it’s earned, instead of too early or never.

I’ve watched this happen inside my own ten. Some of the builders I now write with and appear on screen with started as names I simply showed up for. The relationship grew through ordinary contact until a joint piece made sense. None of it was a pitch. It was a relationship that reached the next floor.

And most of my ten never reach a gate at all. They’re just people I’m glad to know. The relationship was always the point. The gate is what happens when it’s ready.

Building real relationships online

Machine-made connection is hollow, and people can feel it in a second. So show up like a person, and use something that makes sure you actually do.

Building real relationships online by hand is slow, and slow things lose to easy things, right up until the easy thing gets you banned or ignored.

Keep doing the slow thing anyway. Ten real relationships, built by hand over a year, become the one moat nobody can copy: people who know you, trust you, and pick up when you call.

That’s the muscle the internet is trying to amputate. Keep it.

Daily 10 runs this whole practice, the choosing, the saving, the logging, the warnings, and the gates. I built it to keep my own ten alive, and this week I’m giving it to subscribers for free.

If the five steps made sense and you want the thing that makes you do them, subscribe now. You’ll get it in your welcome email. I want you to have it before your next handful of relationships goes cold.

Subscribe now

A guy messaged me on LinkedIn last year pitching a platform for me to try.

He’d seen my work on AI systems and wanted to tell me what he was building — a platform to connect AI talent with the businesses hunting for them. The first version was a clean signup page. Simple, the kind of thing you can make in an afternoon now.

A week later, he posted a major update. He’d overlooked a few things in v1, he said, but he was on it.

A month after that, the tone had changed. Apparently, things were breaking left and right. The fixes cost more every week. Then the post: an apology to the ten or so builders he’d pitched the platform to, and a quiet pivot into writing about how AI is mostly hype.

It turns out … It was his first time vibe coding. A twenty-dollar subscription and one broken version stopped him cold.

Here’s how you can avoid breaking your first AI build.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace, by turning your expertise into systems no one else can replicate.

This piece grew out of a survey of ten solo AI builders I ran last week, and a string of conversations with founders who started a build and quit.

If you want the foundation underneath all of it, start with AI Made Building Easy. So Why Can’t You Start?

Every build hits the technical wall

The first version breaks. Costs creep up. The part you assumed AI would handle turns out to need you, your judgment, your hours, your patience. This happens to everyone who has ever made anything. The wall is the same height whether your idea is brilliant or borrowed.

What decides who climbs it is whether you have a reason to.

The LinkedIn guy didn’t.

I’m not posting this to pile hate on him, but rather to illustrate a lesson based on what I’ve observed: don’t borrow someone else’s conviction.

It seemed like he grabbed the idea because the market looked hot. AI hiring was everywhere, and the experts posting about it sounded so sure of themselves. He borrowed their certainty and assumed AI would supply the rest. So when the wall showed up, there was nothing underneath him to help him over it.

This happens all the time. People work on build ideas they’re fired up about, and half the time the fire belongs to someone else. They saw a passionate expert ship something, felt the pull, and went looking for their own version of it.

AI will build whatever you point it at.

It has no opinion about whether you should care.

How to find an AI build you’ll actually survive

Before you make anything, run your idea through three questions. They take five minutes… and save you a month.

Test 1: Are you user number one?

Would you use this thing yourself, every day, to solve a problem you actually have?

If the honest answer is no, you’re building on borrowed belief.

Daily 10, the relationship tracker I built, had exactly one guaranteed user the day it existed.

Me.

I made it because I kept losing the people who mattered to my work. Not leads in a CRM. People. The ones whose work I wanted to follow, whose posts I meant to engage with, whose relationships I kept dropping because the feed buried them under everyone else’s noise.

Daily was the solution — a Chrome extension that functioned as a glorified bookmark with tracking, motivating, and analytics function. But still, all in all, a glorified bookmark.

I’ve done this with other “products” as well, some even started with a Google Doc. Embarrassingly simple. But it worked for one person before I ever wondered whether it could work for anyone else.

Selling it later was a bonus stacked on top of a thing that already earned its place in my day.

A build with one real user is alive. A build with zero real users is a pitch.

Test 2: Your mission, or borrowed conviction?

Scratch your own itch is old advice. Every builder has heard it. It’s also not enough on its own, because plenty of people scratch their own itch and quit anyway.

Add one condition: this itch has to sit downstream of your mission.

Watch the two stories line up.

The LinkedIn guy could have been his own first user of that hiring platform. Maybe he did need to find AI talent. He still would have quit, because the problem floated on its own, hooked to nothing he was on a mission to do.

My mission is some version of helping creators and founders stay irreplaceable while using AI. Every relationship I let slip was the work failing in miniature. Daily 10 was that mission showing up as friction. It was me serving that mission with a tool I built for myself, even though I didn’t know how to code.

Friction in the thing you care about most is the kind you’ll cross a wall to fix. That’s why I kept paying the twenty dollars, metaphorically and otherwise. Quitting the build meant quitting the mission, and I wasn’t going to do that over a bug.

So don’t just ask whether you’re willing to use it. Ask whether using it will serve a greater purpose, a mission.

Test 3: Does it feed the rest of your work?

The best builds aren’t lucky one-offs. They’re what everything you’ve already built produces when it’s working.

Daily 10 is a good example, because I can trace exactly where it came from. My mission pointed me at the general problem worth solving: helping people stay irreplaceable. Writing about that mission, over and over, sharpened the blurry problem into a specific one. I could see its shape. The people who showed up around that writing became the ones I tested with. I read their comments on my Notes to decide what to actually build, recruited a handful of them for the beta, and used their data to make the thing better.

Only then did the build exist. And once it did, it paid all of that back. It gave me more to write about. It deepened the relationships it was tracking. It became a tool I own outright instead of a subscription I rent forever. For one of the beta testers, the impact was an opportunity.

That’s the loop. What you’ve built feeds the new thing, and the new thing feeds everything back. A build that only does its one job stays a hobby. A build sitting at the center of that loop is an asset, and an asset compounds.

I track this across five pillars when I go deep — mission, marketing, community, assets, and impact (the full breakdown is here).

The test underneath the framework is simpler: could you only have built this because of work you’d already done, and does the build now make that work stronger?

Final thoughts

The founder who messaged me wasn’t wrong that building is hard. He was wrong about what was supposed to carry him through it.

He thought the tool would. He thought AI had made building so easy that belief became optional. Then the wall showed up, and there was nothing on the far side of it worth reaching for.

Your first AI build will hit the same wall. It comes whether you picked the right thing or the wrong one, and that part isn’t up to you. What’s up to you is whether you’re still standing there when it arrives. Pick the problem that’s your mission complaining, the thing only you would actually use, and you’ll keep paying the twenty dollars long after it stops being fun. Because you’ll have a reason to.

The reason underneath the build is the thing that survives the build. Everything else is just code.

Daily 10 is the build I’ve been describing this whole time. It started as a vision to keep the people who matter to my mission from slipping through the cracks, and it’s now a Chrome extension that helps me do it every day.

I’m releasing it free to subscribers this week.

If the three tests got you thinking about your own first build, this is the one I made by passing them, and you can have it. Subscribe now so it lands in your inbox the day it goes out.

Subscribe now

P.S. Want help finding the build that’s worth your month? The AI Readiness Diagnostic shows you where your foundation is solid and where it’ll crack under a real build. It takes a few minutes.

I kept losing people.

Specific people, ones I wanted to build real relationships with. They’d show up in my feed one morning and I’d interact. Then the algorithm would bury them under trending posts and strangers I never asked to follow. By the time I remembered to go back, the moment was gone.

This kept happening. Week after week. I’d find someone worth connecting with, engage once, and lose them to the scroll.

So I built something. A Chrome extension called Daily 10 that saves ten profiles, prompts me to visit them daily, and logs my interactions. Simple. Manual by design. It solved a problem I’d been sitting with for months.

And it changed how I think about what’s worth building with AI.

I decided to write this article to see if other builders had the same moment. And, if so, if there’s something there that founders and creators new to building with AI can work with.

After reading this, I hope you’ll know what exactly to build with AI, not just how.

New here? I’m James — I help founders and creators build AI systems that make them impossible to replace.

My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can replicate. This piece draws from a survey of 10 solo AI builders sharing what they built, where AI helped or failed, and what they left out.

For the philosophy underneath all of it, start with How I Use AI To Become Irreplaceable.

Your build starts with a problem you care about

Last August, I wrote an article called Stop Looking for a Niche. Start Looking for a Problem You Can’t Ignore. The argument: your mission starts with friction you’ve lived through, and the work worth doing is the work that solves it.

That article was the philosophy.

This one is what the philosophy looks like in practice.

Daily 10 was the first time I felt the thesis as a builder. I didn’t start with a market analysis or a competitive landscape. I started with irritation. I was losing track of the people who mattered to my mission because the platform I was using was engineered to scatter my attention. I looked for a solution, but nothing out there was simple enough, light enough, or cheap enough for what I needed.

So I built one.

AI handled the technical side — the code, the scaffolding, the interface. I handled the direction, the design, and one decision that shaped the whole product: when beta testers asked me to integrate AI for auto-logging interactions, I said no. Harvesting feed data would mean potentially violating platform TOS and turning a simple tool into a bloated one.

Plus, the friction of manual logging helped the mission. It kept things more intentional, more human.

I wanted to know if other builders experienced the same thing. So I ran a survey. Three open-ended questions:

• What problem did you solve?

• Where did AI help or fall short?

• What did you leave out?

Ten builders responded. They work in different industries, at different stages, building wildly different products. The same patterns kept showing up.

5 insights from solo builders creating mission-aligned builds

A mission-aligned build is an AI-powered tool or system designed to solve a problem the founder personally observed and cared enough to fix. The build carries the founder’s expertise, judgment, and context into the product itself — which is what separates it from generic AI output that anyone could generate.

These five insights came from pulling on that thread across all ten responses.

1. Start with a problem you couldn’t ignore

Dheeraj is a software engineer with 20 years of IT experience who hadn’t written a line of code in seven years. He’d been running a content business — newsletter, YouTube channel, courses — and the bottleneck was throughput. Research alone ate 20+ hours a month. He had 17 years of content creation experience, so he knew what good output looked like. He just couldn’t produce it fast enough.

He didn’t set out to build a product. He built a research agent that cut those 20 hours down to a single session. Then a content pipeline that compressed 2+ hours of article production to 15 minutes. Then a tool for Substack creators that he built in a 3-day weekend sprint for $135 total.

The product came from the problem. The problem came from his own workflow.

Every single builder in this survey followed the same arc. Nick Quick was losing a war against his own Substack feed and built a CRM because nobody else had. Anfernee kept drowning in scattered thinking across tools and built a Notion AI agent to hold his weekly reflection process. Elena Calvillo couldn’t manage collaboration requests flooding in after her AI Advent Challenge gained traction, so she built DraftKit. Dallas Payne had prompts scattered everywhere with no way to organize them, so she built her own Prompt Vault in Lovable.

Ten out of ten. Every build started with personal friction.

If you’ve read AI Made Building Easy. So Why Can’t You Start?, you know the mental shift that needs to happen before you open a tool. This is the step before that shift. You don’t need a business plan. You need a problem that’s been costing you time, energy, or sanity — one you know well enough to describe in plain language.

Your move: Write down the friction you experience every week in your own business. What do you keep solving manually? What problem would you fix even if nobody paid you for it? That’s your build.

2. AI pulled every builder toward generic — pull away

The pattern was universal. Every builder in the survey hit the same wall.

Nick fought it on every spec rewrite. There is no Substack CRM — nobody had built one. So every LLM he worked with had no frame of reference, and a model with no frame of reference doesn’t say “I don’t know.” It guesses. It kept trying to turn Stackwise into a generic contacts database or an email tool because those things already exist.

Chris Tottman calls AI’s default output “90% vanilla.” Anfernee had to override AI’s tone because early outputs were too clinical for solopreneurs still figuring things out. Manas Moon had to constantly simplify because AI overcomplicated explanations his beginner audience didn’t need. Dheeraj caught AI generating 600 words of setup before getting to the point in a 2,500-word script draft. Joel Salinas scrapped his AI-drafted outreach emails entirely because they sounded like someone else wrote them.

The mechanism is consistent, and we all know this. AI produces the most statistically likely output — the average of everything it’s trained on. That average is usable. It’s also forgettable.

The builder’s job is to recognize when the output matches their specific context versus when it matches what already exists elsewhere.

Nick put it well: the “why” was the one input AI could not generate for him. He rewrote his spec so many times he started to suspect AI’s real job was to make him defend his own idea until there wasn’t a soft spot left.

Your move: Next time AI gives you output that feels “right,” pause. Check whether it feels right because it matches your context, or because it matches what already exists. If you can’t tell the difference, that’s the gap your judgment needs to close.

3. Show your edge in the structure, not the features

The builder’s expertise didn’t show up in feature lists. It showed up in architecture decisions — the way the tool was organized, the workflows it enforced, the logic underneath the interface.

Anfernee built the Solopreneur Idea Validator with three separate paths: one for Digital Product/SaaS founders, one for Service/Consulting, one for Education/Content. AI would have built a single generic validation flow. Anfernee split it because he’d watched enough solopreneurs fail to know that a SaaS idea and a coaching offer require fundamentally different questions. That knowledge came from experience, and it shaped the skeleton of the entire tool.

Elena built DraftKit around a discovery she made after her first version shipped. She thought the core problem was scheduling — so she built a calendar-first tool. Writers told her otherwise. “If I can’t write with someone in here, I’ll just use Google Docs.” The real gap was collaboration, and she had to rebuild around it. That redesign — calendar demoted to supporting feature, writing workspace promoted to core — came from listening to her users and having enough product instinct to act on what they said.

Nick built Stackwise around a constraint nobody else had to work with: Substack has no public API. He had to construct the entire CRM from the outside, working around the platform rather than through it. That architecture decision came from understanding how Substack works at a level that only someone deeply embedded in the platform would.

Dheeraj structured his content pipeline around 17 years of knowing what “good” looks like. AI can draft a script. It can’t tell you that the opening has a hedge in it that undercuts every confident claim that follows. Dheeraj caught that — at minute 7 of a recorded video, after the audio was already done. Expensive lesson. But it proved where the real edge lives in an AI-powered system: in the human who knows what to keep and what to cut.

Your move: Before you add features, map the structure. What workflow splits, categories, or decision trees come from knowledge you earned through experience? AI can build the features. The structure is yours.

4. Protect the end user with strong constraints

The most telling decisions in the dataset were about what builders removed — and why the removal made the product better for the person using it.

Anfernee cut the validation score from his Idea Validator. The first instinct was to give founders a grade at the end. It felt satisfying to build. He killed it because scores create false confidence. A founder can answer every question well and still be chasing a problem nobody will pay to solve. The tool needed to surface thinking. Manufacturing certainty would have undermined the entire purpose.

Elena learned this the hard way. DraftKit’s SMART Draft feature used AI to generate a starting point for a collaboration. Early on, it auto-applied the draft to the workspace — and writers pushed back hard. “Did you write this or did the AI write this?” became real friction. She redesigned the flow so AI only stages a suggestion. The human always pulls the trigger. That was her product sense, and AI had no mechanism to figure it out on its own.

Nick cut every integration that pointed outside Substack. The first version in his head tracked subscribers flowing in from YouTube, Instagram, SEO — full attribution, very dashboard-with-a-corner-office. Then he asked the only question that mattered: does growing on Substack require any of that? The answer was no. So he cut it all. What remained was a tool that does one thing and refuses to blur it.

Dheeraj killed an entire product category. He’d been building and selling n8n workflow templates — individual automations, specific use cases. Each one made sense in isolation. Collectively, they were scattered. Worse: n8n was already his biggest free asset. 42 lessons, 31 hours of YouTube content, completely free. Trying to monetize a category he’d already given away was, in his words, “offer sprawl with extra steps.” He consolidated what remained into a single $27 bundle and moved on.

If you’ve been thinking about what to build, your constraint decisions will matter as much as your build decisions. The AI prompts and workflows you’re already using probably contain the same kind of judgment — features you instinctively left out because you knew they’d hurt more than help.

Your move: For every feature in your build, ask one question: does this serve the user’s outcome, or does it just make the product look more complete? If you can’t answer clearly, cut it.

5. Learn when to keep friction on purpose

This was the most counterintuitive pattern. Three builders chose to keep manual steps in their products — and the friction was the point.

Dallas built her Prompt Vault entirely for herself. It lets her tag prompts, sort them into categories, link back to original sources, record when a prompt is used and any tweaks she made. She can combine prompts in a play environment and merge them into something new. She loves it. And she deliberately did not connect it to any language model, because the manual engagement with prompts — the act of reading, evaluating, combining — was where the thinking happened. Automating that step would have removed the value.

I made the same call with Daily 10. Some beta testers wanted AI to auto-log interactions by harvesting data from the feed. I said no. The extension needed to stay platform-agnostic, light, and simple. More than that — the act of manually visiting a profile and manually logging the interaction was an added mechanism that made the relationship-building intentional. Remove the friction and you remove the intent.

Joel landed in the same place through trial and error. He asked AI to draft outreach emails to potential sponsors. The emails came back generic. They didn’t sound like him, and he didn’t want that to be his first impression. So he flipped the workflow: AI researched each contact and gave him background. He wrote the emails himself. The act of writing — by hand, or dictated into his microphone — was the signal that made the outreach real.

Everyone tells you to automate everything. These three builders found the places where automation would have destroyed the thing that made the build work.

Your move: Before you automate a step in your build, ask whether the friction is where your judgment or relationship equity lives. If it is, protect it.

Mission-aligned building is the way forward for solo builders

Every response in this survey traced the same arc.

• A problem the builder couldn’t walk past.

• A build shaped by their expertise.

• A set of constraint decisions that protected what mattered.

And at every stage, AI was useful — but it pulled toward the middle. The builder’s job was to pull it back toward something specific.

That’s what makes a build mission-aligned. The founder’s judgment is in the product itself. Their experience shaped the structure. Their convictions determined what got cut.

AI handled the bricks, but the blueprint came from someone who cared enough about the problem to get the architecture right.

I experienced this with Daily 10. The build started because I cared about a specific problem — losing track of people who mattered to my mission. AI handled the code. I handled the direction, the design, and the decision to keep it simple when the easy thing would have been to make it complex. That pattern held across all ten builders in this survey. And I think it holds for you, too.

You have friction in your business right now. Something you keep solving manually, something that costs you time or focus every week. You have expertise — years of knowing what good looks like in your domain, what questions to ask, what to cut. And you have access to tools that can turn that expertise into something that compounds.

The build worth building starts with a problem you couldn’t ignore. Start there.

P.S. If you’re a passionate, mission-driven founder or creator figuring out where to start with a clear vision of what to build, but not the time and expertise to make it a reality, we should talk. We can give you a clear picture of where your biggest opportunities are.

I spent 2 hours picking a font for my personal website.

Claude had offered me Georgia for headings and a clean sans for the body in the first sixty seconds. Both reasonable, I liked them, and have even used them repeatedly across multiple landing pages. But this time, I rejected that, and wanted to build something different.

I paid for that decision with 2 hours plus immeasurable frustration.

After that, I had narrowed my choices down from 5 pairs, to 3, to 1, choices built on top of a stack of mockups, inspiration sites, and burnt tokens. Despite the speed that AI promised, I still ended up wasting precious hours hung up on this seemingly minor decision.

And yet, I wouldn’t trade this friction for anything, because, in a time where everyone can make everything, where I’m willing to add friction is what created something COOL.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do by turning your expertise into systems no one else can replicate. If you’re frozen on what to actually build with AI, start with AI Made Building Easy. So Why Can’t You Start? — it’s the foundation underneath this whole piece.

Everyone can make (almost) anything … and it’s a problem

The capability gap closed two years ago. The subscription gap closed last year. The knowledge gap is closing right now.

Everyone is on the same five tools. Everyone reads the same threads about which prompts actually work. Everyone watches the same creators teach the same workflows. For example, 87% of creators now use AI in their work, with most using three or more tools across their stack. The playing field seems to have never been more level.

Some creators see this and freeze.

You might see how powerful AI is. You see how much everyone seems to be leaning into it, and feel behind.

Worse, you feel like you won’t ever catch up. Everything’s moving too fast. Everyone seems to have no trouble keeping up.

Except you.

The freeze is more common than people admit. 60% of small business owners say they don’t know how to apply AI to their actual business — they have the tools, the subscriptions, the awareness. They just can’t move. I see it in my client calls. Expert with three months of tutorials behind them, sitting in front of a blinking cursor with no first AI project they actually want to build.

This is a widely observed problem. I’ve written about it before in The Mental Barriers Preventing You From Building With AI and in AI Made Building Easy. So Why Can’t You Start?. Both pieces come back to the same starting move: build with AI to solve a problem. Yours, or your audience’s.

That’s a good first step, but it’s not the only way.

Because once you’ve answered should I build with AI (yes) and for whom (you, or them), there’s a harder question underneath. What should you build with AI?

The default answer everyone gives you is some version of “automate your annoying admin.” And yes, that’s great. But it’s also boring. Search “AI project ideas” or “AI projects for beginners” and you’ll find a thousand interchangeable suggestions that all do the same five things slightly differently. The internet is full of “fine.”

So I want to give you another answer:

Build something cool.

Not your audience’s idea of cool. Not the market’s. Yours. Just like how my website build might not resonate with others, but I find it awesome.

The rest of this piece is how to actually do that without it sounding like a poster on a coworking-space wall.

What to build with AI: Three steps to make something irreplaceable

Two ideas to set up first.

Coolness, the way I mean it, isn’t a vibe. It’s a recognition signal. You look at a thing and your pulse quickens, your mind races with a million ideas, and you think yes, that, exactly that. That’s the quality I’m talking about.

It’s also specific to you. Most of what the world calls cool feels flat to you. Most of what you find cool, other people don’t even notice.

Identity alignment is the muscle that lets you spot this coolness, this passion or obsession, in your own work before it ships. It’s the ability to look at something AI just handed you — a layout, a hook, a feature, a font — and feel, fast, whether it’s aligned with who you are and what you’re for.

For me, there are three steps to work with this trait. Here’s how you build something cool with the same stack as everyone else’s.

1. Articulate your mission

Cool work that lights you up doesn’t come from nowhere. It comes from a mission you can say in one sentence.

Not a tagline. Not a mission statement in the corporate sense. The actual thing you care about so much that it’s slightly embarrassing to say out loud.

Mine is some version of: help creators and founders stay irreplaceable while using AI. That’s why everything I make — even a personal website — has to feel like it could only come from me.

Not a template, or a default Claude suggestion. Not the average of what worked for someone else.

This is also why the standard advice on what to build with AI feels hollow. It assumes the goal is output, but it’s not. The goal is output that’s yours. Those are different jobs: the first one needs tools, the second one needs a mission.

If you can’t say your mission in one sentence, you can’t recognize cool work when you see it. You’ll keep accepting the tool’s defaults because you have nothing to measure them against.

How to apply this idea:

Answer honestly, do you already have a mission? If not, open a notes app. Write “I care about ___ because ___.” Fill in both blanks honestly. If the first answer is “money” or “growth,” push further. Money for what? Growth toward what?

Keep going until the sentence makes you slightly uncomfortable to read back.

That’s the one.

2. Find the smallest possible way to fulfill it

This is where most people make a big mistake. They articulate a big mission, then try to build the entire universe of that mission in one project.

Don’t.

Pick the smallest possible expression of your mission. The smallest cool thing that could exist. For me, on this round, it was a personal website with fonts that felt like me. Not a course. Not a platform. A website. With fonts.

This is the same move that the AI builders I surveyed made a case for in AI Made Building Easy. So Why Can’t You Start? One of the best ways to work on AI is to start small, the smallest possible way. Small protects your capacity. Small lets you finish. Small forces real decisions instead of theoretical ones.

The best first AI project isn’t ambitious, but specific. A landing page. A weekly email. A two-minute video. An AI side project nobody asked for. A small mission-aligned thing beats a big mission-aligned plan every time.

How can you operationalize this?

Take your mission sentence. Imagine the smallest possible thing that would express it. For example, in my quest for being irreplaceable, I needed to build relationships with the right people. I couldn’t do that relying on my feed.

I created a tool to combat algorithmic drift. With the Daily 10 extension, you can not only save your ICP, you can track your interactions with them, score every checkpoint, and determine the next steps.

It’s nothing fancy, but then again I started with a Google Doc.

3. Run an identity alignment check

Once you start building, AI will hand you a hundred options. Headlines. Layouts. Names. Copy. Color palettes. Features you didn’t ask for.

Don’t accept any of them on the first pass.

The identity alignment check is one question: does this feel like me?

Not “is it good.” Not “is it on-brand.” Not “would this convert.” The goal right now is to build something cool, so this check must be simple. Just: does this feel like me. If yes, ship. If no, change it. If you’re not sure, you don’t know yourself well enough yet — go back to step one.

The font story from the intro is this check, applied.

Georgia was good, it was okay. The default sans was good. But it didn’t feel like me. It failed the identity check. So I kept looking until I found a pairing that didn’t — Space Grotesk paired with Fraunces, an inverted serif/sans convention with little oddities in the letterforms that you don’t notice on the first pass but sticks with you.

Two hours of friction, but it resulted in a font system that felt like me. The same check applies whether you’re a writer, a builder, a coach, or any of the other shapes AI for creators takes these days.

So, the next time AI hands you an output you almost accept, pause. Ask the question, “Does this feel like me?”. If you can’t say yes, that’s me with conviction, the work isn’t done.

What happens if you don’t do this

When you don’t build from your mission, there are two possible outcomes.

The first: you never build.

You stay frozen because you have no internal compass pointing anywhere specific, so all directions look equally valid and equally pointless. This is the freeze from the start of this piece. It’s the most common answer to how to start with AI — and it’s invisible because it happens in private.

The second is sneakier. You do build, but without a mission backbone.

You research for a problem you don’t care about, and it starts off great. But you don’t really care about it, so you take the AI’s first suggestion. You add every feature it proposes. You let the project bloat. The thing you make can be functional, but forgettable, and it’s more like the AI’s. It’s not yours.

Worse, you lose interest in it halfway through, because nothing in it is actually yours. You’re not invested. The motivation that was supposed to carry you to the finish line was never there to begin with. It was borrowed from whatever inspired you to start, and inspiration evaporates fast when the work doesn’t feel like yours.

The outcomes look different from the outside, but they all feel the same. They’re pointless, discouraging, and you’re back to square one.

And you can’t work on this with more discipline.

The fix is passion — alignment with what you want or what you care about. It’s making something cool that you think is worthwhile.

This isn’t just a website or app thing

You might be reading this thinking it’s about apps you can build with AI. It’s not.

AI is a tool. The internet was a tool. Social media is a tool. A camera is a tool. A notebook is a tool. We’ve been handed tools for thirty years.

Everyone got the internet. Everyone got social media. Everyone got smartphones with cameras better than what magazine photographers shot with twenty years ago. AI for small business owners is the latest version of this same handout — same tools, access, same starting line. And yet — not everyone makes the same slop.

The people whose work you remember from the last decade weren’t the ones with better tools. They were the ones with clearer obsessions. The thing they cared about leaked into everything they made, until you could recognize their work from across the room. When tools commoditize, your obsession becomes the moat.

That’s always been the game. AI didn’t change it. AI just made it more obvious, because the floor of “fine” is so high now that fine is no longer interesting.

What’s interesting is what you, specifically, find COOL. What you can’t stop thinking about. The detail you spent two hours on when nobody asked you to.

So whether you’re picking fonts, writing emails, designing offers, naming a product, or figuring out what to build with AI on a Tuesday night — the question doesn’t change. Does this feel like me? Does this excite me?

If yes, keep going. If no, the work isn’t done.

This is how you build something irreplaceable using the same tools as everyone else.

P.S. Are you building something that lights you up right now? If not, what’s stopping you? Let me know in the comments.

There has never been a better time to build a brand online.

There has never been a harder time to be believed.

Every founder, operator, and consultant is being told the same thing. Become a creator. Build in public. Marketing is no longer optional. It is the work.

But every niche looks saturated before you have written your first sentence. The floor of the internet has dropped out. Bad-but-passable writing is now free and infinite. Readers have learned to scan a paragraph, decide it was made by a machine, and scroll on.

Edelman’s Trust Barometer has been tracking the shift for years. Trust in institutions keeps falling. Trust in brands keeps rising. People buy from brands — and people — they recognize. Research from groups like Stackla and Nosto tells the same story from a different angle: peer voices outperform brand voices by a wide margin. As more AI-generated content floods every channel, the trust currency gets more expensive to mint.

To do business now, you have to build trust. To build trust, you have to market consistently in a voice that is recognizably yours, and for many, that means writing.

But many, many people hate writing.

I have read some version of this sentence in comments, calls in three languages, and a dozen client calls and interview.

I cannot write. But I must market.

New here? I’m James. I help founders and creators build AI systems that make them impossible to replace. My mission is to help you use AI to become irreplaceable at what you do, by turning your expertise into systems no one else can replicate.

For the foundation this article builds on, start with How to Train AI on Your Brand Voice.

AI writing systems solve the problem … but only a specific kind

The first time AI helped me write something that actually sounded like me, I was not impressed by the AI. I was impressed by myself.

The article would have taken me eight hours of agony. It took two hours of excited conversation instead. I do not remember the title. I do not remember where I posted it. If I went back today, I would probably spot tells I would not let slip now.

But I remember the feeling.

At the time, I read it back and recognized it as mine. Not the AI’s version of a person. Mine.

And it was like a whole world opened up.

That moment is the reason this newsletter exists. The reason a Medium article ships every week. The reason three to five LinkedIn posts go out in the same window, while three builds sit in progress and multiple full-time clients keep moving. (Yes. This is too many things. I am aware. I have a problem.)

I am not the best writer in the world. I do not have to be. I have an AI writing system that lets me ship in a voice readers can recognize, at a volume that earns trust over time.

Here is the reframe most of the AI writing conversation gets backwards. The point of a good system is not to make writing easier for you. That is the side effect. The point is to make trust easier for your audience. A consistent voice, shipped weekly, infused with a real worldview, earns trust in a way one heroic essay every six months never will.

I can already hear the objection forming. Is AI writing not the exact thing eroding trust in the first place? Hold that thought.

The writing purists are right about one thing

The strongest version of the case against AI writing goes like this: Writing is thinking.

If you outsource the writing, you outsource the thinking, and what comes out belongs to no one. Readers feel the difference. Every founder broadcasting a prompt result is contributing to the slop pile that is destroying the internet’s trust supply. So learn to write, hire a human, or stay quiet.

Grant them the strongest part of that argument first. They are right about how AI is failing audiences. Most AI writing online belongs to no one. It is what happens when you treat a model like a vending machine. Generic prompt in, generic paragraph out, shipped because shipping feels like marketing. Readers learn one more time not to trust a writer who looks suspiciously fluent.

That’s the slop tax we are all paying.

But the purist’s argument assumes only one method of AI writing exists: The lazy one.

They are generalizing from slop. It is like judging cooking by gas-station hot dogs and concluding humans should stop using stoves.

The right question is not, did AI touch this? It is, does this genuinely help me? Do I feel for it? Can I feel the writer’s worldview encoded in the output? Are the opinions there? The lines this person would not cross?

If yes, the writing builds trust. If no, the writing erodes it. The answer does not depend on which fingers typed the words or how much tokens got burned by generating it.

A bad human writer with no point of view erodes trust just as fast as a badly prompted AI essay — the main point: the tool (AI writing) isn’t the problem, but how it’s used.

And no, I am not defending the slot-machine crowd either

The version of AI writing that earned the purists their suspicion is real. I am not running cover for it.

“Write me a LinkedIn post about X.” Paste. Ship. Repeat. That feed is a cemetery. The posts are interchangeable. Swap the headshot, swap the name, the post stays the same. Readers know. They scroll past faster every quarter.

The standard I am holding is simple. A piece of writing should be unmistakably the person who shipped it.

AI is allowed in the kitchen. AI does not get to be the chef.

Trust is built when readers recognize the same specific mind showing up over and over. A slot-machine system cannot produce a specific mind, but a mission-infused system can.

What a mindful AI writing system actually is

So, can you use AI for writing newsletters, social posts, blog articles that build trust even if you’re not a “professional writer”? My answer is yes.

But you need to have a system that supports and amplifies you.

A mindful AI writing system is not a tone-of-voice prompt. It’s not just an AI that “writes like you.” For me, this is an encoded vessel of you with four layers. Strip any one out, and the whole thing collapses back into the kind of output that loses you the reader you were trying to win.

Encoded profile, not just “voice”

For a well thought-out AI writing system, having an accurate voice is a given, so I’m not going to tackle it here. If you want, I have three different methods, 2 from guest authors here, and here, and one from me.

Most “AI voice” work in the wild is depressingly shallow, because they end there. That misses the point of an AI writing system entirely. What good is an AI that sounds like you if it doesn’t know what it is you’re trying to say?

At most, they’re adjectives. Vibes. “Warm, conversational, with a touch of wit.” Models pattern-match the surface, and readers recognize the mimicry instantly.

A real “voice” encodes something deeper. Your worldview. Your philosophy. The origin stories that lights you up every time you tell them. The opinions you will not soften for a stranger. The things you refuse to say even when a model thinks they fit. Done right, your encoded profile becomes an asset in its own right.

Can your writing system disagree with a bad idea in your name, the way you would? Does it know what you consider to be a bad idea in the first place? If yes, it knows what you believe, not just how you sound.

Consistency is writing style is important, yes, but a good AI writing system does more than that. It creates consistency of conviction over time, not consistency of adjectives.

Source material from your life

The second layer decides where the system gets its words and ideas from.

Most AI writing fails at this step. By default, a language model writes by predicting what would sound reasonable given everything it has read on the internet. That produces writing that sounds reasonable to anyone, which means it does not sound specifically like you. It sounds like the average of every person who has ever written about your topic online.

A mindful system pulls from a different source: your own material. Your past articles. Voice notes you have recorded. Transcripts of client conversations. Direct interviews the system runs with you to gather new material on a specific topic, which is exactly how the article you are reading right now was built. The system’s job is to find what you have already said or thought, and put it into a new piece of writing.

If the system does not have the material it needs, it asks you for it. It does not make something up. The moment a system invents a fact, a quote, or a story you never told, the trust you were building with the reader is gone.

You as the final ratifier

The third layer is the role you play at the end of the process. The system writes a draft. You read it. You decide whether it goes out.

Nothing should ship under your name unless you have read the full draft and confirmed that the ideas, opinions, and phrasing all sound like yours. A simple test works here: if you would not be comfortable reading the draft out loud to people who know you well, do not publish it.

This is the step that separates a thoughtful AI writing process from a careless one. The careless version skips this step. The writer prompts the model, copies the result, and posts it without reading carefully. The thoughtful version treats every draft as a proposal that needs your sign-off before it becomes public.

Your name is what readers are learning to trust. That trust is what makes the writing valuable. So the responsibility for what gets published has to stay with you.

Mission as override

The fourth layer is what keeps the writing aligned with what you actually stand for over time.

Every AI writing system needs clear rules about what it is allowed to write about and what it is not. These rules come from your mission: the specific work you are doing, the audience you are serving, and the ideas you want to be known for. Without these rules, the system will gradually produce writing that is more general and less specific to you, because general writing is what language models produce by default.

This layer does three jobs. First, when the draft starts to sound generic, the rules pull it back toward your specific topics and positions. Second, when the system cannot find real source material from your life, it tells you so instead of filling in the gap with invented content. Third, when a draft does not fit any of the topics you have decided to write about, the system does not publish it.

This is the layer that allows you to write at high volume without losing the things that make your writing yours. Every piece you publish should make you more recognizable to your readers, not less.

How this works for client AI writing systems

Everything above is framework. The harder question is whether it survives contact with a real client. Here is how the four mechanisms looked when I built AI writing systems for two client engagements.

One is a cardiologist with decades of patient experience, now building a wellness platform. He has a method, an offer, and strong opinions about how the work should be done. He knows that writing on social media would be one of the strongest ways to establish his brand and get to know his framework, but he doesn’t have consistent writing routine.

The other is a naturopathic consultant. He runs a small practice and is growing an audience in the natural health space. He knows his work. His business is set up. He just cannot get the writing out of his head and onto the page on any kind of schedule.

Both have something to say. Neither writes easily. Both needed an AI system that would not flatten their voice into the generic content already filling their feeds.

1. Operationalizing the encoded profile. For each of them, I built four profile documents. A personal profile captures their worldview, origin story, and the positions they will not soften. A writing profile captures how they actually speak, pulled from transcripts of them explaining their work to a real person (me). A business profile captures what they sell, who they serve, and what every piece of content has to support. An AI system guidelines profile captures the instructions, constraints, and output formats the model is allowed to use.

2. Operationalizing source material from their life. Every system pulls only from their own material. Past writing. Voice notes. Transcripts. Client material. Previous work. For the cardiologist, most of what he knew had never been written down, so I sat with him for hours and turned the recordings into the knowledge base. For the naturopath, some of the material was already there in volume, so the job was structuring it so the system would retrieve from his voice rather than defaulting to the generic wellness corpus on the open web. Same idea in both builds. Very different work to get there.

3. Operationalizing them as the final ratifier. This is the layer most AI writing systems skip when they sell to non-writers, and it is the layer that fails them inside a year. A system that lets a non-writer ship without reading carefully quietly turns them into a slot-machine operator. So both of them have a hard ratifier step. Every draft sits in front of them before it goes out under their name. That is the floor. On top of that, both are in a coaching engagement built to grow their writing muscle next to the system, not behind it. The goal is not to let AI become a permanent crutch. It is to ship now while they grow into writers who can approve their own drafts without flinching.

4. Operationalizing mission as override. The AI guidelines profile and the business profile do the override work together. The guidelines tell the model what it can produce, in what formats, and under what constraints. The business profile tells the model what each of them is here to sell. A draft that fails either check gets pulled before it ships. For the cardiologist, that keeps the writing from drifting into generic wellness talk that has nothing to do with his clinical perspective. For the naturopath, it keeps every piece tied to his practice and his actual offer, not the wider influencer aesthetic that would absorb him fast if the guardrails were not there.

Both started as non-writers. Both have told me, almost word for word, that writing used to take them days. They would sit down, stall on the gap between what they knew and what they could get written, and either ship something they did not love or not ship at all.

With the system running, the first aligned draft lands in minutes. The publish-ready piece comes together inside a few hours. Quite a big leap.

Permission, with a standard

If you are the founder I started with, the one who keeps muttering I cannot write but I must market, here is what I want you to take with you.

You are allowed to build a presence with AI. The purists are working from a strawman. The version of AI writing they are arguing against is not the only version available to you.

But.

The permission comes with a standard. You do not get to slot-machine your way into a brand. The bar for any AI writing system worth building is the four mechanisms. Encoded profile. Source material from your life. You as the final ratifier. Mission as the override. All four. Before you commission or build one, there are five questions worth answering honestly.

Trust has never been more valuable. Trust has never been harder to earn.

The way through, for founders who cannot write but must market, is not refusing AI. It is also not surrendering to it. It is mission-infused systems, operated by someone who actually have something to say. That is how trust gets rebuilt at scale.

The Unpromptable voice is not the one that refuses to touch AI. It is the one whose mission, worldview, and opinions are so encoded that no generic prompt could reproduce it.

P.S. I write up the architectural decisions behind each of these builds at my personal website. Walk through them there, and reach out if you have any questions.

TL;DR: Most AI-built websites look the same because most people skip the identity work and accept AI's defaults. This article walks through the five-step process I use to build websites that actually reflect my brand — developing taste, working with pre-existing assets, building custom ones, sketching before prompting, and treating AI like a collaborator instead of a vending machine. None of it requires design experience. All of it requires caring enough to push past "good enough." If you want help for anything similar, feel free to reach out.

I just finished building my third landing page when I tabbed between all three of them and realized something:

They looked the same.

Not identical — different colors, different copy, different purposes. But the bones were the same. The spacing, the section flow, the way the hero text sat above a centered call-to-action button. If you squinted, they could have been siblings.

I had spent some time on these, and was quite proud of them the first time I bought them to life. They worked — responsive, fast, clean. And for a guy with zero design background, the fact that AI let me build functioning websites at all was a small miracle. A month earlier, I would have been thrilled just to get something live.

But now, I realized these three pages could have belonged to anyone.

New here? I’m James — I help founders and creators build AI systems that make them impossible to replace. Most people chase tools. I teach you to build infrastructure around your thinking, your voice, and your specific bottlenecks. If you want the foundation this builds on, read AI Made Building Easy. So Why Can’t You Start?

The ignorance tax

I’ve seen this pattern before — with writing.

When AI writing tools first hit the mainstream, non-writers started producing articles at volume. They were happy with them. The paragraphs were grammatically correct, the ideas were organized, the tone was professional. That felt like enough.

It wasn’t. Those articles were verbose, generic, and interchangeable. But the people writing them couldn’t see it, because they’d never developed the eye for good writing. They didn’t know what they didn’t know.

The same thing is happening with AI website design.

AI gives you something that works. Responsive layout, clean typography, a hamburger menu that actually opens on mobile. You feel accomplished — and you should. A year ago, building a site without coding experience would have taken weeks and a freelancer. Now you can have one live by dinner.

But “works” and “branded” are different planets. AI defaults are designed to be generic. They use safe spacing, predictable section orders, the same card layouts you’ve seen on a thousand SaaS landing pages. When your only reference point is other AI-built sites, “good enough” feels great. I wrote about this dynamic with visual branding on Substack — recognition only forms through deliberate constraint, and AI’s instinct is to do the opposite.

This is the ignorance tax you pay when building with AI. You pay it in three ways: your site looks like every other AI-built site (which weakens your brand), visitors bounce because nothing catches their eye (which kills conversions), and small backend issues pile up in places you can’t see (which costs you later).

There’s nothing wrong with paying the ignorance tax. Every builder pays it when they start. I paid it on three separate landing pages. The question is whether you want to keep paying it — and if not, what to do about it.

Why identity is the real bottleneck

There are excellent technical guides on building websites with AI from people like , , , , and a particularly memorable build story from . They have excellent manuals for the engineering side — clean code, robust architecture, token-efficient prompting, deployment best practices. Those guides are valuable, and I used several of them when building my most recent projects.

My focus is different. I’m focused on identity — making your website look and feel like you. Because the technical execution is downstream of knowing what you want. Most people skip that part entirely because AI lets them. You can go from zero to deployed site without ever answering the question: what does my brand actually look like?

This matters because a clean, robust, token-efficient website that looks generic is still generic. The market is drowning in them. With the sheer volume of AI-built sites flooding the web right now, your website’s brand identity is what separates a site someone visits from a site someone remembers.

The same principle holds for your brand voice in writing — adjective-based prompting (”make it modern, make it clean”) gives you AI’s version of modern, which is everyone’s version of modern.

You need constraints, and taste. And these five tips will help you with those.

Agency-level AI website design in five tips

A caveat: I’m still in the middle of learning all of this myself. I’m sharing the process that would have saved me a ton of time and tokens when I first started, and the things that made the biggest difference once I stopped settling for “good enough.”

I’ve only pushed one website to production, and am in the middle of building a second one. Doubtless they won’t be perfect as I had to learn along the way, and I plan to go back and improve them. But I believe they’re a far cry from what I used to build, and I’m going to share why.

The purpose of these tips is to expand your concept of what AI is capable of, so you can direct it with more specificity instead of giving it basic instructions and accepting the defaults.

Here they are.

1. Develop your taste before you touch a single tool

When someone remodels a kitchen, they don’t start by ripping out cabinets. They browse photos. They save references. They develop a feel for what they want before a single hammer swings.

Website design works the same way, and most people skip this step because AI makes it so easy to generate something immediately. You type a prompt, get a site, and start tweaking. But tweaking without direction is how you end up with three identical landing pages.

Before you prompt anything, spend time collecting visual references.

I spent hours collecting pins for the “vibe” that I want to portray. At that time, Pinterest was my go-to place for it and it has excellent stuff, but I have since discovered that there are many other sites for this exact purpose:

• Behance has professional design portfolios that will recalibrate your eye.

• Dribbble and Awwwards showcase award-winning web design.

• Don’t overlook your competitors’ sites — study what pulls you in and what pushes you away.

Save 10-15 references. Notice what draws you in — the whitespace, the typography, the way a hero section uses photography or videos or interactive elements, the color palette. You’re training your eye the same way you’d train your ear for good writing; by consuming a lot of good work before producing your own.

This step alone changed everything for me. Before I did this, my prompts were vague: “build a modern landing page for an AI consulting business.” After, my prompts got specific: “I want a dark minimal layout with a large serif headline, asymmetric grid, and a single accent color — reference this mood board.”

The output went from “another AI website” to something I recognized as mine.

2. Work with pre-existing assets

Many non-technical builders don’t realize this, but you don’t have to build everything from scratch.

Many developers and designers have created open-source resources specifically for AI-assisted builds, and using them gives you a better starting point than a blank prompt. I wasted hours and tokens getting parts of my sites to function just right, before I realized that others have solved that specific problem, and are giving away the solution for free.

For example, there are specialized skills you can plug into your AI workflow to solve a wide number of issues. You can get them from sites like Skills.sh:

I’ll link my favorites below:

• Audit. A skill for “Systematic quality audit across accessibility, performance, theming, and responsive design with prioritized findings.” Run this to surface technical and design issues you weren’t even aware existed.

• Design taste. This skill is a “humanizer” skill for UI development. This skill will help your first pass avoid common AI conventions.

• UI/UX Pro Max. “Comprehensive design intelligence for web and mobile UI/UX across 10 technology stacks.”

These aren’t magic bullets, but they raise the floor of what your first draft looks like by a wide margin.

On the component side, libraries like 21st.dev offer pre-built, high-quality UI components that you can drop into your projects. Instead of letting AI generate a generic testimonial section from scratch, you can hand it a polished component and say “use this as the base, customize it for my brand.”

The principle: start from something good, then make it yours. This is how professional designers work. They don’t draw every button from nothing — they use systems and modify them.

3. Learn to build your own assets

Sometimes pre-built resources won’t cut it. Your brand might need something specific — a custom animation, a unique graphic element, a video background that says something only you can say.

The tools available now would have been unthinkable two years ago. You can generate custom video assets through platforms like HuggingFace. You can create 3D elements and embed them directly. You can design brand-specific illustrations using AI image tools and refine them until they match your exact visual identity.

I won’t turn this into a tool tutorial — that’s a different article entirely. The point is this: if you can imagine it, there’s probably a way to build it with AI right now. Most people don’t explore these options because they don’t know they exist. Knowing they exist changes the questions you ask.

4. Sketch before you prompt

This one might sound old-fashioned, but drawing things out before I open a code editor has become a non-negotiable part of my process.

Sketching forces you to think about layout, visual hierarchy, and content flow before AI fills in the blanks with its own defaults. When you skip it, you hand over creative decisions to the model — and the model will always choose the safest, most statistically average option. That’s what language models do. They trend toward the center.

My sketches are rough. Boxes and arrows and scribbled labels. They don’t need to be pretty. They need to capture the structure I want so that when I sit down to prompt, I’m directing the AI toward a specific vision instead of asking it to generate one for me.

It’s the same principle as outlining an article before writing it. The outline doesn’t need to be polished. It needs to exist, so the draft has somewhere to go.

5. Adopt the “Everything is figure-out-able” mindset

This is the single biggest lever, and it has nothing to do with any specific tool.

Most people use AI like a vending machine. They put a prompt in, they get a website out, and they work with whatever comes back. But AI is a collaborator, and the quality of the collaboration depends on the quality of your questions. I wrote about this in the context of building AI systems — the same principle applies here.

• Start with “Can you...” and “How can we...”

• Can you research what kind of content should go on a coaching services page?

• How can we create a custom hover animation for these portfolio cards?

• Can you audit this page for accessibility issues and fix them?

• How can we make the mobile version of this hero section feel less cramped?

The AI that builds your website can also launch research agents to figure out what content belongs on each page. It can generate copy, create specific visual assets, test responsive layouts across screen sizes, and debug issues you didn’t even know were there. Most people tap into maybe 20% of this because they treat the first output as the final output.

Your website prompt is the beginning of a conversation. Push back on the defaults. Ask for alternatives. Say “this looks generic — what would make it feel more distinctive?” Ask it to reference your mood board. Ask it to explain why it made certain layout choices, and whether those choices serve your brand or just the model’s statistical average.

The more you ask, the better the output gets. And the more you learn about what goes into good design — which feeds back into step one, developing your taste.

When to build, and the alternative

Everything above assumes you have the time and willingness to iterate.

For many people, that process is worth it — you learn what your brand looks like from the inside out, and that knowledge pays dividends for years. Build your own when you’re still clarifying your brand, when the site is a learning project more than a revenue-critical asset, and when you’re energized by the creative repetition.

But consider hiring when:

• Your website is a direct revenue driver and you need it operational fast

• You’ve already tried building and the results consistently fall short of your vision

• You need backend complexity — CRM integrations, automated booking flows, custom data pipelines — that goes beyond visual design

• Your time creates more value in your core business than it does talking to a code editor

And even if you do hire, an AI-native developer or agency will build your site faster, cheaper, and at a higher quality ceiling than traditional freelancers or agencies ever could.

The old quote of $5,000-$15,000 for a custom site is a different market now. AI-native builders work with the same tools described in this article, but with the design eye and technical depth to push them further. The economics have shifted in your favor.

Both paths work.

The framework above will make you a better client even if you do hire, because you’ll know what you want and you’ll articulate it with specificity. Whether you build or hire, the identity work has to come from you.

AI can execute. Only you can decide who you are, and therefore, only you can decide what your brand looks like.

The real shift

Six months ago, I was thrilled that AI could build me a website at all. That thrill was real, and it was earned. Building a custom functioning site without a technical background is an unlock that didn’t exist two years ago.

But the game moves fast. What felt like a superpower six months ago now feels like table stakes. Everyone can build a website with AI. The edge comes from building one that looks like you — that carries your visual language, your brand weight, your specific identity. Not because AI can’t do the work, but because most people don’t give it enough to work with.

Develop your taste. Study what moves you. Sketch before you prompt. Push back on defaults. Keep asking better questions.

Your website is a storefront for your mission. Make it look like yours.

P.S. If you’re building AI-powered systems for your business and want help figuring out whether your web presence matches your brand, I’d love to chat. You can always reach out.

TL;DR: I built Daily 10 because the most valuable thing AI cannot do for you is care about people consistently. The first cohort of 13 beta testers proved the thesis: real conversations, working partnerships, people showing up for each other in ways the feed had quietly stopped allowing. Their feedback shaped Daily 10 v2 features, which I talk about here. The core thesis of this article: Relationships make you irreplaceable, and irreplaceability is a practice you maintain.

Almost everything on your screen right now is AI.

The article you’re reading was touched by an AI model. This image above was made by one. Outside this article, the captions on every video, the suggestions in every inbox, the order of every feed — AI, most of it.

The exception is your online friends.

People here who you actually know. The five people you can always rely on to publish articles you read on a bad day. The three people who would say yes to a collaboration before they’d read the brief. The one editor who knows your work well enough to commission a piece without a pitch.

Those relationships are not generated. They are accumulated. And the way they are accumulated is boring: You show up for the same people. Repeatedly. Across enough surface area that they recognize you when it counts.

This is why community is a big part of Unpromptability.

And that’s why I built something for it.

New here? I’m James. I help founders and creators become irreplaceable in an AI-saturated internet. I teach you to build infrastructure around the parts of your work AI cannot copy: your voice, your taste, and your relationships.

This piece walks through Daily 10, a Chrome extension I built and beta-tested with 13 creators, and what their feedback shaped in v2.

For the philosophy underneath all of it, start with The Only Moat Left In The AI Age.

The problem of algorithmic drift

The internet rewarded showing up for your people for about a decade.

Then the algorithm took the wheel. You stopped seeing your people. They stopped seeing you. What got measured as “engagement” became a slot machine attached to a feed optimized to keep you watching, not keep you connected.

I built Daily 10 to take that wheel back.

I call this algorithmic drift. The slow replacement of the people you meant to follow with the people the platform wants you to see. You opened the app to keep up with twenty creators you respect. The feed quietly swapped eighteen of them out for engagement bait, trending posts, and a recommended account that already has a million subscribers.

Your feed weakens your relationships, and weak relationships make you promptable.

I write Unpromptable because the strongest defense against being replaced by AI is not a better tool.

It is a network of humans who actually know you, would rather work with you than search for an alternative, and would actually choose you over a search result. That network is the one thing on the internet a model cannot generate. It can only be built by you, slowly, over time, by remembering specific people and showing up for them.

Daily 10 is the smallest piece of infrastructure I could build to stop letting the algorithm decide who you engage with. What you’re reading now is what 13 testers and ten weeks have done to that principle in practice.

A mission-driven MVP

Daily 10 is a Chrome side panel that holds ten people (or 3, or 5) in front of you every day. You add their profile URL from Substack, LinkedIn, X, anywhere, and you log your goals and their priority level.

The extension surfaces the top people you should engage with, prioritizing the ones you’ve neglected. You click a name. Their profile opens in a new tab. You go engage. Comment, react, send a DM, jump on a call, whatever fits the relationship. You come back. You log what you did. You move on.

That’s the entire loop. There’s no automation. No DOM scraping. No AI suggestions. No generated comments. The extension does not engage on your behalf, because the entire point is that it should not.

What it does is make showing up easier. A streak counter for consecutive days at 10 of 10. A 30-day score for each profile that quietly tells you which relationships are warming up and which are cooling off. A small visual flag when you haven’t touched someone in seven days.

The design choice underneath is the part I want you to take with you. Most software in the AI era is being built to do the thing for you.

Daily 10 is built to make sure you do the thing yourself. It’s intentional friction, exactly what human relationships are.

The beta challenge and what it produced

I ran a 10-day challenge with 13 testers in April. Mostly Substack writers, each picking ten people they wanted to be more present for. Five of them have filed full survey responses so far, and every one of those responses contained a moment that would not have happened without the tool.

logged a writer she’s wanted to get closer to on Substack. She’d just discovered him through a mutual friend. For seven days she paid attention to his notes and posts. Some days she left a thoughtful comment. Some days nothing he posted moved her, and she didn’t force a reaction. On Day 8 he noticed she kept showing up, sent a message, and they ended up on a Zoom call where he walked her through how to use AI in her workflow. The conversation she now describes as warm and friendly did not exist before the extension reminded her to look.

wanted to get to know the work of our mutual friend, Pinkie from . By the end of the ten days, Pinkie had featured her in a post and they were DM-ing weekly about Claire’s Substack rebrand strategy. A digital nudge became a working partnership.

, who runs a Substack of his own, was actively looking for collaborators. Daily 10 became the reminder loop that surfaced the right people often enough that when he reached out, they already knew his name and his work. He told me later that the side panel turned a vague intention into a daily mechanic.

Aarav had a call with someone designing a statistics and machine learning course for product managers. He’d assumed senior creators were inaccessible. The conversation showed him they aren’t, when the outreach is genuine and consistent.

Then there was who, while she connected with people she normally wouldn’t have, felt that Daily 10 felt like another to-do on top of being already very active on Substack. She was right. For her, it was friction without payoff. That is also a result. A tool that produces a real outcome for one creator and produces nothing for another is not a failed tool. It is a tool with an audience. Tracy’s note told me who Daily 10 is for, and who it isn’t.

Building unpromptable infrastructure means listening to the people using it

Ten days of real usage. A WhatsApp group I checked every morning and night. Multi-day threads where testers told me what was breaking, what was confusing, and what was quietly working. v2 is what comes out the other side of that. A set of small, deliberate changes that make Daily 10 serve the person using it.

1. Streak Shield. Monica lost her streak twice from timezone confusion. Claire lost hers when her schedule got busy. Both told me the streak reset felt punishing in a way that didn’t match the spirit of the habit. v2 ships a Streak Shield, a manual protection that lets the streak survive a missed day.

The friction is still real, it just provides allowances for the daily realities we have to face.

2. Workweek Mode. Des said it cleanly: “I don’t work weekends, so that resets the streak number I think.” v2 added a Workweek Mode that scopes the daily loop to weekdays only.

The streak respects how you actually work.

A small briefcase icon flips on when it’s active, gold-filled, so you can see at a glance which mode you’re in.

3. Three new interaction modes: Depth (3), Balanced (5), and Breadth (10). The most consistent complaint from v1 was that a 10-person checklist punishes you on the days you have no time. 10 people is a best practice. It is not a law. Some founders and creators have the bandwidth for ten people a day. Many can only show up for the three or five who matter most that week.

v2 lets you choose. The mode flexes with your week. The principle stays.

4. More context per profile. Several testers said the roster felt flat once you got past the names. Useful in week one, less useful in week three. v2 adds context controls on each profile: who the person is to you, what your goal with them is, your priority level, and what the next step in the relationship looks like.

The roster stops being a list of names. It becomes a small, legible map of why each person is on it.

5. Relationship gate. The 30-day score tells you a relationship is warming up. It doesn’t tell you what to do about it. v2 adds a relationship gate that surfaces a prompt when a profile crosses a threshold: you’ve created a relationship with this person — take the next step.

Actually comment, instead of liking, or even shoot a DM. This stops Daily 10 from quietly becoming a comment-and-react loop.

6. Weekly and 30-day analytics. Several testers wanted to see consistency over time, not just today’s score. v2 surfaces weekly and 30-day analytics so you can watch the habit build.

7. Accessibility and a warmer palette. The v1 dark UI had real contrast issues some testers worked around silently for ten days before mentioning them. v2 widened touch targets, fixed text contrast against the panel background, and shifted the palette warmer. A Chrome side panel you’ll have open at multiple points per day cannot feel cold.

Each of these is small. Together they are the difference between a beta tool and one I’d happily let a stranger install.

Unpromptability is a practice

I keep returning to this idea because the longer I work in AI, the more obvious it becomes. The strongest moat against being replaced is not a tool. It is not a workflow. It is the network of humans who know you, trust you, and choose you over the algorithm’s next suggestion.

That network is not built by posting more. It is not built by going viral. It is built by interaction, comments, and nervous DMs. It is built by talking to them on a Monday and then doing it again on Wednesday. Slowly. With intention. Across enough months that it becomes a relationship and not a transaction.

Daily 10 is one asset in a larger practice. The smallest piece of infrastructure I could build that makes that practice easier. There will be others. Different surface, same principle. The unpromptable asset is whatever lets you be present with the specific people, the specific work, or the specific expertise only you can bring.

But this one I’m proud of, because every feature in v2 is the voice of a real person who used the thing and told me what was missing. The wisdom is theirs. The code is mine. The outcome belongs to whoever picks it up next. Ten relationships warmer than they were a month ago.

Join the second cohort

I’m running the Daily 10 Challenge again. Same shape as the first. Ten days, ten people, one daily check-in. Same goal. Show up consistently for the people who matter to your work, and notice what changes.

This cohort gets v2. Streak Shield, Workweek Mode, the cleaner logging flow, the onboarding screens, the warmer interface. You also get the WhatsApp accountability group, a daily prompt from me, and a final survey that shapes v3.

If you want in, reply to this email or comment with the word “cohort” and I’ll send you the link. Spots are limited so I can keep the feedback loop tight.

Pick ten people this week. Show up for them. Watch what happens.

The coaching space is drowning in free lead magnets.

Everyone has an opt-in. Everyone has a freebie. Everyone has an ebook, a checklist, a PDF gathering dust in someone’s downloads folder. And everyone is competing for the same exhausted attention with the same playbook: give away value, build a list, then sell.

Ellen Scherr took a different approach.

She launched a $139 AI-powered assessment for women navigating midlife transitions. In-depth, personal-coaching-level information given to you, but no discovery call required. Pay, answer 60+ questions, receive a personalized clarity report in your inbox.

Since launching on January 28, 2026, 56 people have taken the assessment — roughly a dozen of those complimentary, offered during the validation phase to pressure-test the product. At least four have converted to high-ticket coaching engagements in the first three months. That’s over $5,500 in assessment revenue before a single coaching call was booked — and the system keeps running while Ellen sleeps.

New here? I’m James — I help founders, coaches, and creators build AI systems that make them impossible to replace. Most people chase tools. I teach you to build infrastructure around your thinking, your voice, and your specific bottlenecks. This piece is a full case study of a client build. For the foundation this builds on, read 5 AI Systems Nobody’s Talking About (That Actually Build Moats).

The attention problem

The coaching and wellness space is saturated. The market rewards volume — more content, more opt-ins, more followers. Volume attracts browsers. Ellen needed buyers.

Her work sits in a gap most offerings don’t serve. Midlife transition for women isn’t a crisis to be fixed with a template. It’s a slow unraveling of identity — career shifts, relationship recalibrations, existential questions that don’t have searchable answers. The standard “here’s a 5-step framework” approach fails because the work requires depth. It requires someone who can see patterns across years of accumulated experience and reflect them back with precision.

Ellen spent 18 years as a licensed therapist before expanding into coaching. That combination — clinical depth paired with forward-looking coaching methodology — is rare. One-on-one work delivers that depth. Ellen loves doing it. But it doesn’t scale, and it requires someone to already trust her enough to book a call.

She needed an entry point — something that demonstrated her methodology, delivered real value, and filtered for the people most likely to want deeper work.

Free downloads attract tire-kickers. Low-ticket courses compete on price. Group programs dilute the intimacy that makes Ellen’s work effective.

Ellen needed a product that did three things: captured attention in a crowded market, delivered standalone value worth paying for, and naturally surfaced candidates for high-ticket coaching.

The standard playbook didn’t fit. We needed to build systems that encode her specific advantage — not systems that make her marginally faster.

What Ellen has now

Ellen now has a $139 AI-powered diagnostic called the Midlife Clarity Assessment that mirrors the early work she does with private clients.

Users answer 60+ questions designed to surface patterns most people can’t see on their own. The output is a personalized clarity report — pattern recognition, reflection prompts, and a 90-day action plan — delivered directly to their inbox.

The experience is available on-demand. No waiting for a calendar slot. No discovery call required. Users get access to Ellen’s frameworks whenever they’re ready, at a fraction of 1:1 coaching cost. The report becomes a working document, a map for their next 90 days, grounded in 18 years of therapeutic and coaching insight.

Results in three months since launch:

• 56 people have taken the assessment (roughly a dozen complimentary during the validation phase)

• At least 4 have converted to high-ticket coaching engagements

• $5,500+ in assessment revenue

• Ellen receives a weekly report identifying warm leads for direct outreach

• We’re actively iterating on positioning and distribution to scale volume while maintaining conversion quality

The assessment generates revenue on its own while functioning as a qualification engine. Ellen isn’t chasing leads. Buyers are self-selecting into her ecosystem, demonstrating intent with their wallets before she ever speaks to them.

How we built it

Each component of the system serves a specific function in the pipeline.

The assessment layer

Users complete a detailed Fillout form with 60+ questions. These aren’t data collection fields but are introspective prompts. Each question surfaces specific patterns in how someone relates to identity, relationships, purpose, and future. The act of answering is already valuable. By the time someone finishes, they’ve done real reflection work.

The processing layer

Answers flow into an n8n automation that handles routing, processing, and delivery. The workflow manages the entire pipeline — from form submission to report generation to email delivery — without Ellen touching anything. She wakes up to completed assessments and revenue in her account.

The wisdom layer

This is where the system earns its value.

The AI doesn’t generate advice from its general training. It retrieves from Ellen’s second brain — a curated knowledge base of her core concepts, frameworks, essays, and approaches built through RAG (retrieval-augmented generation). When the system offers a reflection or suggestion, it pulls from Ellen’s actual body of work. The output is grounded in her philosophy, her language, her accumulated insight from two decades of practice.

This is the layer that makes the assessment feel like Ellen.

The delivery layer

Users receive their clarity report directly in their inbox.

The report includes pattern recognition based on their responses, tailored reflection prompts, and a 90-day action plan. The language matches Ellen’s voice. The structure mirrors her coaching methodology. Users get a tangible artifact they can return to — guidance that travels with them beyond the assessment itself.

Privacy-focused architecture

User responses are processed and used to generate the report. They aren’t stored for training, shared externally, or used for marketing purposes.

Ellen has access to response data for personalization — so she can tailor follow-up if someone books a coaching call — but the data stays private. This matters for the kind of intimate reflection the assessment asks for.

Users share real fears, real struggles, real uncertainty. That deserves protection.

Weekly lead assessment report

Ellen receives a recurring report that evaluates recent assessment takers. The system flags likely candidates for her high-ticket coaching engagement based on their responses and engagement patterns. This turns the assessment into an ongoing lead qualification engine.

Ellen knows who to reach out to, and she has context on their situation before the conversation starts.

Ongoing marketing support

The build didn’t stop at the tool. We’re actively collaborating on marketing — refining positioning, testing distribution channels, adjusting messaging based on early buyer feedback.

The system is designed to improve as we learn what resonates. The 56 assessment takers so far represent the baseline, gathered in three months. The goal is to scale volume while maintaining conversion quality.

Would this work for your business?

What we built for Ellen works because a few specific conditions were true — and they’re worth naming if you’re wondering whether something like this could work for your business.

First, she had deep accumulated expertise. Eighteen years of practice, thousands of client conversations, a body of written work that captured her frameworks. The AI had something meaningful to retrieve from. If your expertise lives mostly in your head and hasn’t been documented in any form — essays, SOPs, recorded sessions, frameworks — the system has nothing to pull from. That’s the starting line.

Second, her audience faces complex, personal problems. Midlife transition doesn’t have a one-size-fits-all answer. That complexity is what makes a personalized assessment valuable enough to charge for. If your audience’s problem can be solved with a checklist — even a technical checklist — an assessment won’t carry the same weight.

Third, there’s a natural progression from the assessment to deeper work. The report gives users real clarity — and for some of them, that clarity reveals they want a guide. The assessment surfaces those people and gives Ellen context before the first conversation. The $139 product funds itself while filtering for $3,000+ engagements.

If those three conditions are true for you — deep expertise, complex audience problems, and a clear path from diagnostic to deeper engagement — this model has legs.

AI as force multiplier

Most creators worry AI will replace them. That fear is valid — if your value is information delivery, AI does it faster and cheaper.

Ellen’s advantage lives in what AI can’t generate on its own: pattern recognition developed over 18 years of therapy and coaching. The ability to see what someone can’t see about themselves and reflect it back with clarity. That accumulated insight — earned through thousands of hours of real conversation with real women in real transitions. That’s the unpromptable core.

The AI only amplifies that core. It retrieves from her knowledge base, applies her frameworks, delivers in her voice. But it doesn’t replace her. It amplifies her. The wisdom is still Ellen’s. The AI is the delivery mechanism.

Anyone in the world has access to the tech stack. The Fillout form, the n8n workflow, the RAG retrieval, these are tools anyone can access. But you cannot copy Ellen’s accumulated insight. And you cannot copy the work we did in thinking through the assessment concept, value delivered, and delivery.

Those are the moat.

This is what AI-powered infrastructure looks like in practice. Identify what’s genuinely irreplaceable about your expertise. Then build systems that extend it without diluting it.

The best use of AI isn’t to replace your thinking. It’s to make your thinking travel further than you ever could alone.

If you’re navigating midlife

I built this system with Ellen. I’ve read her essays, studied her frameworks, watched how she thinks about the women she serves. So I want to speak directly to some of you reading this who might resonate with her work.

If you’re in the middle of the kind of transition Ellen specializes in — identity shifts, career pivots, the quiet questions that don’t have Google-able answers — the Midlife Clarity Assessment is a good place to start.

Ellen brings something rare: 18 years as a licensed therapist combined with a coaching methodology built for forward momentum. She doesn’t offer quick fixes or tidy checklists. She offers clarity — the kind that comes from someone who’s walked the path and spent two decades helping others do the same.

Take the assessment

The assessment takes about 30-60 minutes. The report lands in your inbox within 24 hours. And you’ll walk away with a 90-day map built from frameworks that have helped hundreds of women find their footing.

I built it. I’ve seen what’s under the hood. And I can tell you the depth is real.

If you’re stuck, this is a good place to start. Don’t hesitate.

P.S. If you’re a founder, coach, or creator sitting on years of expertise you haven’t figured out how to scale — I’ve been in that seat. What we built for Ellen is one version of what’s possible when you pair real knowledge with the right AI infrastructure.

If you’re curious about what that could look like for your business, I’d love to hear about it. You can learn more about how I work and get in touch.