Here’s the short answer: you actually can.
And here’s the longer answer: the number of people who can drive AI well enough to do it properly is a handful. And the gap between those two statements is where most businesses are currently losing money.
What you lose when you replace a team with a chat window
Think about what a successful digital product actually involves.
Pull it apart into its component pieces: the user research, the product design, the technical architecture, the frontend, the backend, the security considerations, the testing, the deployment, the ongoing maintenance. In a normal project, you’d have different people handling different parts of that. Each of them brings specific expertise. Each of them asks questions and pushes back from within their discipline. Each of them catches things the others miss.
Now imagine replacing all of those people with a chat window.
The dream AI is selling right now is that you can lie in bed, get struck by a lightning bolt of an idea, open a laptop, and have something built by morning. And in a narrow technical sense, that’s true. You can generate code, create interfaces, connect systems, and ship something faster than was possible at any previous point in history.
But here’s what didn’t happen while you were doing that: nobody pushed back on the idea. Nobody asked whether the user actually needs this. Nobody questioned the technical approach. Nobody ran the numbers on whether it’s worth building. The room full of humans who would have stress-tested your thinking. They were replaced by a system that is, at its core, trying to help you do what you asked.
AI feels like a better shortcut. It often produces the same outcome.
The right way to use AI to build
There is a right way to use AI in this process. I see very few people doing it.
It looks like this: use AI to massively skill yourself up before you build anything. Use it to pull your idea apart critically, to understand every component, every dependency, every assumption baked into what you think you want to build. Use it to ask hard questions of yourself, not to execute before you’ve finished asking them.
That approach does bring genuine speed and efficiency. It compresses the learning curve. It lets one person hold significantly more of the picture than they could before. But it requires you to bring the judgment. The AI supplies the knowledge and the execution. You supply the thinking.
Most people invert this. They bring the idea and let AI supply everything else. That’s not a new kind of digital project. It’s the same old digital project, running faster.
The Forward Deployed Engineer, and what it doesn’t solve
The market has noticed that something new is needed here. A role has emerged, the Forward Deployed Engineer, that’s being hired aggressively right now by the biggest AI companies in the world. The concept came originally from Palantir, who for years had more Forward Deployed Engineers than software engineers. The idea is someone who can sit inside a real business and actually deploy AI in a real operational context, bridging the gap between the technology and the business problem.
It’s a genuine and valuable role. I find the concept compelling.
But here’s what it doesn’t solve: the thinking problem. A Forward Deployed Engineer can deploy AI faster and more effectively than almost anyone. They can’t make the underlying problem worth solving. They can’t guarantee the right question is being answered. They are, by definition, an efficiency play.
Efficient mistakes are still mistakes.
AI makes you feel like you’ve done the thinking
There’s something else worth naming; and I say this as someone who uses AI every day and has built real products with it.
AI makes you feel like you’ve done the thinking when you haven’t.
It’s a remarkably validating tool. You describe an idea, and it builds on it, elaborates on it, helps you develop it. It doesn’t tell you the idea is wrong. It doesn’t tell you you’ve missed something fundamental. It tells you what a good version of what you described might look like. And that feels, viscerally, like rigour.
It isn’t. It’s enthusiasm dressed as analysis.
The speed AI gives you is real. The thinking AI appears to replace is an illusion. And the faster you move, the more expensive that illusion becomes when it catches up with you.
So can you build it yourself with AI?
So can you build your digital projects yourself with AI?
Yes. If you’re one of the handful of people who can drive it properly (who bring the judgment, the critical thinking, the willingness to slow down before you speed up, and the experience to know what questions to ask before you start building), then AI has made you significantly more powerful than you were five years ago.
For everyone else: AI has made it faster and cheaper to build the wrong thing.
The question to ask before you start isn’t “can I build this with AI?” It’s the same question it’s always been: “Should I build this at all, and do I actually understand what I’m building?”
That question doesn’t have an AI shortcut.
Garth Shoebridge has been building digital products for 25 years, and using AI as part of his workflow for the last few. If you’re trying to work out whether to build something yourself or bring in the right expertise, start with a conversation.