The Mental Barriers Preventing You From Building With AI

I asked 14 AI builders how to overcome them.

New here? I'm James — I help founders and creators build AI systems that make them impossible to replace. Most people think skill is the barrier. It rarely is. The real block is identity — the belief that "someone like me" doesn't build things.

This piece draws from conversations with 14 AI builders on what actually shifted for them, and offers five doors into the same room.

If you want the broader philosophy on becoming irreplaceable, start with Unpromptability: 5 Steps to Become Irreplaceable in the Age of AI.

You’re sitting on an idea.

Maybe it’s a system that could save you ten hours a week. Maybe it’s a tool that would make your business harder to replicate. Maybe it’s something you’ve been sketching in your head for months — something that you think would be really, really cool to have. A workflow, an automation, a product that you know AI could help you build.

You haven’t built it.

You know AI is capable. You’ve seen what builders are shipping. You’ve watched non-technical founders launch apps, spin up automations, create systems that would have required a development team two years ago.

And still, you’re frozen.

The gap looks like skill

“The biggest misconception is that you need to learn to code first. You do not.” — Sam Illingworth

I surveyed twelve builders — founders, creators, consultants — who crossed the line from “interested in AI” to “actively building with it.” I asked them what held them back before they started.

Their answers were unanimous. The biggest perceived gap is skill. Technical knowledge. Some course they hadn’t taken, some language they hadn’t learned.

I felt this too, at the start. I assumed I needed to understand Python, understand how AI works, before I could build anything real. I assumed the people shipping AI products had some foundation I was missing. So I waited. I consumed more content. I watched others build.

That assumption kept me frozen for longer than I’d like to admit.

A deeper layer: identity

“Skills aren’t the real barrier. Identity is. People don’t think someone like them builds products.” — Jenny Ouyang

The skill gap is real. I won’t pretend it doesn’t exist.

But it’s not the thing keeping people stuck.

For me, Jenny’s insight cut deeper than anything else in my survey. She came from a scientist background, became a parent, moved into software development. At every stage, she was put in a category — and before becoming a developer, nothing about that category was about “building.” Even with technical skills, she didn’t see herself as someone who builds products and ships them to real users.

I recognize that pattern in myself.

I run an AI automation business. I build defensible systems that clients pay thousands of dollars for. Last night, I taught a COZORA class on my AI philosophy to a community of builders. And somewhere in my head, I still think: People like me can’t build real things. Not like what everyone else is building.

The longer you wait, the wider the gap feels. But the gap was never skill. It was permission. Permission you were waiting for someone else to give you.

What cracked that belief open for me was realizing I’d already been building. Architecting solutions. Delivering outcomes. Solving problems with duct tape and determination. I just hadn’t called it building because I didn’t see myself as a builder.

The evidence was already there. I just needed to name it.

Screenshot of survey premise.

The mindset shifts builders made

I asked every person in my survey: What changed?

What shifted in their thinking that made building feel like something they could actually do?

The answers weren’t about technology. They were about mental models.

Here are five patterns that emerged — five different doors into the same room. Pick the one that fits your specific block.

1. Start with your annoying problem, not a grand vision

“Don’t start with a grand vision. Start with that one annoying task. Then open Claude Code or any AI tool and describe exactly how you do that task today, step by step. Build the simplest version that saves you 30 minutes. Ship it. Or put it where you can use it on a daily basis. Then improve it.”

Dheeraj Sharma

Most people stall because they’re trying to build something impressive. An app. A startup. A system that will change their industry.

The builders who ship started smaller.

A task that annoyed them this week. A workflow that wasted their time. A low-level frustration that they’d be glad to get rid of.

Dheeraj built his entire content operating system — five research agents, a full pipeline, multiple products — from a single frustration: “I’m tired of spending three hours researching every article.”

Elena Calvillo said something similar: “Stop reading theory and solve a tiny, annoying problem you have right now.” All of Jenny’s products followed the same pattern: “I set out to fix my own problems, then realized other people had the same ones.”

The first step isn’t building a product. It’s noticing what drains you. One task. One workflow. One friction point that eats your time every week.

That’s your first build.

Tonight: Write down one repetitive task you did this week that annoyed you. Not the biggest problem in your business. The most irritating small one. That’s your starting point.

2. Talk to AI like a thinking partner, not a search engine

“It’s the closest thing to having a strategic co-founder available at 2am. Open it. Have a real conversation. Tell it what you’re building. Ask it to poke holes in your idea. That conversation alone will teach you more about AI than a week of tutorials.”

Sharyph

Most people use AI for isolated tasks. Write this email. Summarize this article. Fix this sentence.

That’s not building. That’s errands.

The shift is relational. When you treat AI as a vending machine, you get vending machine outputs — isolated, transactional, forgettable. When you treat it as a thinking partner, you start giving it context. You start having a conversation. You start iterating toward something real.

Kim Doyal recommends starting with voice: open a dictation app and talk to AI like you’d talk to a colleague. “This is what I want to do. This is who I’m serving. This is the problem it solves. What do I need to know? How do we get started?”

That conversation alone will teach you more about AI than a week of tutorials.

Tonight: Open your AI tool. Don’t ask it to do a task. Tell it what you’re trying to solve, who it’s for, and what’s getting in the way. Have a conversation. See where it leads.

3. Ship something embarrassingly small

“The founders who get stuck are the ones who plan for six months before building anything. The ones who ship are the ones who start with something embarrassingly simple and let it grow.”

Sam Illingworth

The gap between “I built a prototype” and “someone is using this” is where most people stall.

Perfectionism is a stalling tactic dressed up as quality control. The version you ship this week will teach you more than the version you plan for six months.

I recall using the same principle when I shipped a Google Doc as a product. Embarrassingly small, but it helped me build that mental muscle and put me on a path. And I’m not alone.

Sam built his first game as a text quiz with ten questions. It now has 192 entries and thousands of plays. Jenny put it plainly: “Aim for shipped, not perfect. Cross that gap as fast as possible, even if the product is rough.”

The version you’re embarrassed by is the version that teaches you what actually matters.

Tonight: Whatever you’re building — or thinking about building — ask yourself: what’s the version I could ship this week that would be useful to one person? Build that. Don’t look at the mountain ahead, start with an embarrassingly small step below you.

4. Pick one tool and go deep

“Just really try to understand and master a certain tool. See which one is a fit with how you work, and then just take a deep dive into that one tool. Don’t think you’re missing out by not trying every single tool out there.”

Joel Salinas

Non-technical founders often stall in research mode. Downloading apps. Watching tutorials. Comparing features. Two weeks later, nothing built.

The trap is optionality. Every new tool feels like progress. Every comparison feels productive. But comparison isn’t building. Depth is building.

Pick one tool. Push its limits. Learn its quirks. The builders who ship aren’t the ones who tried everything — they’re the ones who went deep on one thing and figured out how to make it work for their specific problem.

Tonight: Choose the AI tool you’re most comfortable with — Claude, ChatGPT, whatever. Stop researching alternatives. Start building with the one you have.

5. Find other builders

“Don’t build alone. Isolation kills more products than bad ideas. Find two or three people who are also building. Share what’s working, what broke, what you’re stuck on. The builders who ship are the ones with other builders around them.”

Jenny Ouyang

Building in isolation makes every obstacle feel like a personal failure. Building alongside others normalizes the friction.

This doesn’t require a formal community. It requires two or three people who are in the same phase. People who will tell you when you’re overcomplicating things. People who will celebrate when you ship something small.

Alex Randall Kittredge started by watching and following other Substackers who were trying new tools out. “I followed their lead and experimented on my own.” That’s all it takes — someone slightly ahead of you, someone building alongside you, someone to share what’s working and what broke.

Tonight: Find one person who’s also trying to build something with AI. Send them a message. Ask what they’re working on. Building becomes sustainable when it’s shared.

You might already be a builder

The thing holding you back wasn’t skill. It was the belief that building belongs to someone else.

But the evidence might already exist — in places you haven’t thought to look.

That infoproduct you created. The case study you assembled for a client. The onboarding system you duct-taped together because no one else was going to do it. The spreadsheet that grew into a workflow that now runs a chunk of your business.

None of that required code. All of it required the same mental process: see a problem, architect a solution, deliver an outcome.

That’s building. You’ve been doing it all along.

Transferring that process to AI is a shorter hop than you think. You’re not learning a new skill from scratch. You’re applying a skill you already have to a new medium.

And if you’re not there yet — if you genuinely haven’t built anything, if the evidence doesn’t exist — then start tonight.

Pick one of the five doors. Go all in on that single thing until you hit a wall. Talk to AI like a thinking partner until you run out of things to say. Ship something embarrassingly small, and keep improving it until someone actually uses it. Solve your annoying problem until it stops annoying you.

When you hit the wall, try another door. Go deeper on the problem. Find another builder to unstick you.

The only wrong move is staying frozen.

---

P.S. If you want to take this further, I write about AI systems that make businesses irreplaceable — philosophy and practice, every week. If that resonates, you’re already in the right place.