Can AI write your PRD?

I recently joined the MVP Club, a professional development community focused on the adoption of AI. Last week we kicked off a challenge to build and launch an app in a month. I was pumped. I had an idea for a tool that would help simplify the home-buying decision for first time homebuyers — possibly one of the biggest and most stressful decisions we'll make in our lives.

Naturally, I turned to AI.

I fed the high level concept into ChatGPT and asked it to draft a product requirements document (PRD) and brand guidelines. Within seconds, it delivered something surprisingly comprehensive: goals, personas, feature set, the works. I skimmed it, made a few tweaks, and thought, wow, this just saved me hours, if not days. I dropped the PRD into Lovable, a vibecoding tool that turns product ideas into websites.

A few minutes later, I had a prototype. It looked good. Really good.

I felt a rush of pride. No way I could've done this on my own so quickly. It felt like I had a superpower. For a brief moment, I believed the hype — maybe we really are heading toward a future where entire products are conceived, scoped, and shipped with minimal — or no — human involvement.

But then I looked at what had been built.

It wasn't right.

The site was polished, but too complex. Somewhere between the prompts and the output, the soul of my idea had gotten lost.

And that's when it hit me: this wasn't a failure of the tools. It was a failure of mine. I had asked the tools to take the lead before I had done the foundational work myself.

Back to basics

The reason I fell in love with product management is because it starts with the customer. Who are they? What are they struggling with? What does their journey look like without your product?

I had skipped that. I'd asked AI to do the thinking for me before I had done the work of understanding the problem.

So I started over.

This time, I began with the customer: a stressed-out homebuyer trying to weigh tradeoffs they've never had to think through before — location vs. commute, school ratings vs. access to green spaces, balancing their desires with that of their partner; all while having to navigate an insane amount of online listings. I mapped their journey, their emotional highs and lows.

Then I asked myself: what kind of questions would help them gain clarity? What kind of support would actually reduce their stress? And what information should this tool highlight — or intentionally leave out?

Only then did I return to the product, the interface, and yes, the AI — but now as a partner in execution, not as the architect. It's taking me a bit longer to build now, but my vision is clear. My goal is to share the MVP in early June — stay tuned!

What AI can and can't do

Don't get me wrong — the outputs from ChatGPT and Lovable were incredibly impressive. They gave me inspiration, speed, and even ideas I plan to incorporate down the line (just maybe not in the MVP). But what those tools couldn't do — what they still can't do — is care.

They can't ask, is this worth building? They don't know the emotional weight of a home purchase. They can't sense when something "feels off," when the interface doesn't match the intent.

And they certainly can't spot when the solution is misaligned with the problem.

That's still our job.

The evolving role of the PM

If anything, AI has clarified what product managers are really here to do. We're not just writing PRDs or managing backlogs. We're sense-makers. Translators. Stewards of purpose.

In a world where it's easier than ever to spin up software, what matters most is not how you build, but why you build — and for whom.

AI can give you 50 possible versions of something. The PM chooses which one gets built. And just as importantly, the PM knows when none of them are quite right, because the customer is missing from the equation.

Far from making the PM role obsolete, AI makes it even more critical. The tools are faster — so the thinking needs to be sharper.

Closing thought: the real superpower

I still plan to use AI in my product work. It's an incredible tool to brainstorm around and refine my ideas. I also am obsessed with vibecoding. I love that I can go from idea to prototype in an afternoon. But this experience reminded me that speed without clarity is just noise.

I don't think AI will kill the product manager. To me, it has clarified why the role matters in the first place.

Because products need a soul. They need intent. They need to start with real people, real problems, and a clear reason to exist.

Tools can build the product. Product managers give it purpose.