Don't forget your role as a PM when working with AI tools

When I started building my AI travel product with Claude Code, I had an almost magical expectation: I tell AI what I want, and AI builds it. For a while, that actually worked. Then I noticed the product was starting to drift in random directions.

I Let My AI Engineer Become the Product Owner

Claude Code is an extremely capable technical partner. It can inspect a codebase, diagnose problems, propose solutions and ship changes at a speed I couldn't match myself.

And that capability made me increasingly passive. Instead of actively deciding what the product should become, I was sometimes just reviewing what Claude had built and reacting to it.

That's when I realised I had temporarily forgotten my own job.

AI Can Build It. It Doesn't Know What "Good" Looks Like.

Claude knows how to build a working product. It doesn't automatically know what I envision, who the target user is, what trade-offs I'm willing to make, or what would make someone genuinely love the product.

Those decisions still need a product owner.

I have to actively question its recommendations, clarify ambiguous requirements, challenge its assumptions and inject my own product judgement. And surprisingly, there have been moments where Claude has effectively acknowledged: I hadn't considered that.

The Best Workflow Isn't "AI Does Everything"

I've started thinking about Claude less as an autonomous product builder and more as an extremely capable engineer on my team.

I provide the direction, context, priorities and product judgement. It helps me investigate, build, debug and iterate at a speed that would otherwise require a much larger team.

The relationship works best when I'm driving, not when I'm watching.

AI Makes Product Judgment More Important

This changed my view of whether AI will replace PMs.

I don't think AI replaces a good PM simply by becoming better at execution. In fact, the better AI becomes at execution, the more valuable clear product direction becomes.

A strong PM brings vision, user empathy, product sense and the ability to decide what is worth building. AI can dramatically amplify the ability to turn that vision into reality.

Conclusion: AI didn't remove the need for me to think like a PM. It taught me that when execution becomes incredibly easy, I need to be even more deliberate about owning the product direction.