If AI is so good at solving problems, why do you still need employees?

Reading that back, that sounds brutal. But as time goes on and the models improve, it's a reasonable question. The tools are genuinely impressive. But the question misidentifies the bottleneck. Product development isn't slow because writing code is hard. It's slow because understanding what to build and why is hard. AI works to requirements. It doesn't work to intent.

Requirements vs Intent

Imagine asking your AI tooling to help simplify an onboarding flow.

This looks like a sensible goal. The AI can see exactly what you were optimising for, so it recommends removing a step that on the surface looks like it is adding unnecessary complexity.

The problem? That step isn't complexity. It's the core business proposition. Remove it and you're not building a simpler product. Best case you've lost some context of what you were trying to build. Worst case you're building a different product entirely.

You could argue that you should have given the AI better context. And you'd be right. But that's precisely the point. The context has to come from somewhere. It comes from months of customer conversations, failed experiments, and hard-won decisions that sit in the heads of the people who built the business. You can't shortcut to it. You have to earn it.

Requirements are what you ask for. Intent is why you're asking for it, and what you're ultimately trying to achieve. The difference between the two is where products win or lose. The AI could see the requirement was to simplify onboarding but had no way of knowing the intent behind the step it recommended removing.

The tools don't compress the learning that builds that intent. They don't replace the iteration. They don't skip the moment where you realise you built the wrong thing and have to go back. Customer feedback, market signals, the hard conversations with users - these are still what separates a product with a point of view from one that just functions.

What AI changes is how quickly you can execute once that intent is clear. That's genuinely valuable. But the intent still has to be defined first, and the only way to build it is to go through the process.

What This Means For Your Business

AI is very good at the parts of product development that are routine and repeatable. It's the unique parts — your business model, your customers, the decisions that make your product yours — where it falls short. That's not a temporary limitation waiting to be solved by the next model release. It's the nature of what these tools are.

As building code gets faster and cheaper, the constraint shifts. It's no longer about how quickly you can produce something. It's about how clearly you can decide what to build.

The businesses that succeed in this AI era won't be the ones who generate the most output. They'll be the ones who remain clear about what makes them valuable — and use AI to build something that reflects that, rather than something that just looks like everything else.

That's a higher bar than it used to be, not a lower one. Speed is easy now. Knowing what to build is the hard part.