AI didn't kill product intuition. It exposed who never had it.
Using AI to validate ideas before they consume sprints, credibility, and roadmap space.
A company I worked at decided to implement an affiliate marketing platform. The president had used it at a previous company. It worked there. So, it would work here.
That was the entire thesis.
He pushed other priorities off the roadmap to make room. Gave high-level requirements. Asked for delivery dates. Engineers built the integrations within a couple of sprints. Fast work.
One problem. Nobody on the technical team had ever touched this platform. Nobody steering the implementation understood how affiliate marketing actually worked at a business-rules level. The president who did understand it wasn’t available to give field-level requirements. He just wanted to know when it would ship.
I was the PM with ties across the company’s tech stack, running CRM full-time but helping steer this alongside a handful of other projects. I could see the shape of the failure forming. Integrations getting built on top of assumptions nobody had validated. Business logic that lived entirely in one person’s head, and that person was in back-to-back meetings asking, “when is this done?” instead of “here’s how this should work.”
The platform launched. The integrations functioned. And the whole thing flopped, because no one could translate complex affiliate rules into a system that actually reflected how the business needed it to operate.
Nothing about that failure was inevitable. Every question that would have killed the bad assumptions was available from day one. Who on this team has domain expertise? Can we articulate the affiliate rules before we build? Is “it worked at my last company” sufficient evidence for a six-figure platform commitment?
Nobody asked. Everyone built.
That project didn’t fail because no one had the answers. It failed because no one had a reason to ask the hard questions out loud.
Product intuition was never about having good ideas
I keep thinking about that project because the failure mode is so common. Not bad ideas exactly, but untested ideas that accumulate momentum before anyone interrogates them. And I think LLMs are quietly changing how the best product folks prevent it.
Everyone thinks product intuition means you can spot a winner. Walk into a room, hear a pitch, and know. That’s movie stuff.
Real product intuition is knowing which ideas are fragile before you’ve invested in them. Strong PMs have always done this. They corner a sharp colleague and say, “poke holes in this for me.” They look for the kill shot early, because finding it at your desk on a Tuesday cost nothing. Finding it in front of leadership costs weeks of credibility. Or worse, it costs sprints of engineering time on something that was never going to work.
What’s changed is the colleague is available at 11pm. It doesn’t get tired of your questions. And it won’t politely nod along because the idea came from the president.
An LLM with the right framing will interrogate your idea the way a skeptical stakeholder would. It won’t know your company’s affiliate rules. But it will notice you haven’t mentioned them.
Who has domain expertise here? What’s the switching cost from the current solution? Where’s the evidence this problem is severe enough to change behavior? Can we articulate the business rules before we start building? These aren’t AI-generated insights.
They’re the questions you already know you should ask but skip because the room has momentum and nobody wants to slow it down.
What this actually looks like
You have a half-formed idea. Instead of opening a slide deck, you open a conversation. You describe it raw and unpolished. Then you ask the model to be the skeptic.
Not to improve your idea. To find what’s wrong with it.
Over a series of exchanges, you build up context: market assumptions tested against pushback, user segments refined, business logic interrogated before a single integration gets scoped. By the time you pitch, you’re not presenting a deck. You’re presenting the residue of an argument you’ve already had.
I’ve started doing a version of this with my own projects. Before I build anything, the idea goes through a structured conversation that forces the uncomfortable questions first. Who specifically has this problem? What do they use today? Would they pay? Most ideas don’t survive. That’s the point.
How to apply this to your next idea
Step 1: Stop yourself before the deck. The moment you feel ready to start building a proposal, that’s the moment to interrogate the idea instead. Excitement is the enemy of scrutiny.
Step 2: Describe the idea like you’d describe it at lunch. No frameworks. No positioning language. Just “I think we should do X because Y, and the people who’d care are Z.” Feed that to an LLM and tell it: “You are a skeptical Head of Product who thinks this is a waste of engineering resources. Find the three assumptions that will break this implementation.”
Step 3: Build a corpus, not a deliverable. Each conversation adds context. Objections you’ve handled, assumptions you’ve tested, business logic you’ve mapped. This becomes your pitch ammunition. Not a polished artifact but a body of evidence that shows your thinking.
Step 4: Separate what survived from what you’re attached to. Some ideas hold up under interrogation. Most don’t. The skill isn’t generating ideas. It’s recognizing which ones earned your time.
Final thoughts
If someone had fed that affiliate platform idea into an LLM on day one and asked, “what are the top five reasons this implementation fails,” domain expertise gap would have been near the top of the list. It wouldn’t have replaced the president’s knowledge. But it would have surfaced the question nobody was asking: does anyone else here actually understand this business well enough to build it?
A PM with weak instincts uses an LLM to generate a proposal and calls it done. A PM with strong instincts uses it to destroy three bad ideas before breakfast and walks into the office with the one that survived.
The difference isn’t the tool. It’s knowing what to do with the wreckage.
What’s your kill-shot question — the one that’s saved you from a bad bet? Reply and tell me. I’ll share the best ones next week.
Mike Watson @ Product Party
P.S. Want to connect? Send me a message on LinkedIn, Bluesky, Threads, or Instagram.