A few months ago, a PM friend at a fintech startup told me about a challenge her team couldn’t quite solve. They were drowning in customer feedback: thousands of support tickets, NPS comments, and sales call transcripts. Everyone knew there were insights buried inside, but no one had the time to dig through it all. They had a hunch about a feature customers wanted, but no confidence to act.

So she tried something different. She ran all of that feedback through an AI clustering tool. Within an hour, clear patterns emerged: confusion about fees, repeated frustration with onboarding, and yes, demand for the feature they suspected. Instead of spending weeks manually combing through comments, they validated their hypothesis in a single afternoon and shifted into prototyping.

That’s the real promise of AI in discovery. It doesn’t replace your judgment or conversations with customers. It simply gives you leverage. It clears the noise so you can spend more time on what matters.

Why Discovery Often Breaks Down

Product discovery is meant to be the most creative and user-driven part of our jobs. But if we’re honest, it often feels rushed. There’s too much data and not enough time. We listen to a few calls, run a survey, and then fall back on the opinions of the loudest stakeholder. Discovery becomes a checkbox, not a real exploration.

The result? Teams ship features that don’t stick because they never truly understood the problem. AI won’t magically fix this, but it can make it easier to work with the data you already have and uncover patterns that would otherwise stay hidden.

Practical Ways to Use AI in Discovery

Here are a few practical use cases I’ve seen work:

1. Synthesizing customer feedback

You probably already have a mountain of support tickets, survey responses, and call notes. Feeding this into an AI model can help group similar themes and spot recurring pain points. An edtech team I worked with discovered that students weren’t leaving because of price, which was their initial hypothesis, but because onboarding instructions were unclear. That insight reshaped their roadmap.

2. Analyzing competitor reviews

App Store and G2 reviews are a goldmine, but who has time to read thousands? AI can summarize what users love or hate about competitor products. This doesn’t replace customer interviews, but it can surface gaps you might explore with your own users.

3. Exploring “what if” scenarios

Some PMs use AI as a sounding board. For example, inputting a draft feature concept and asking the model how different personas might respond. It’s not real research, but it helps you think through edge cases and frame better questions for actual interviews.

4. Prioritizing discovery backlogs

If you’re like most PMs, you’ve got a long list of “ideas to validate.” AI can help group, cluster, or even rank these ideas based on signals from customer sentiment, feedback frequency, or external data. It won’t tell you what to build, but it can help you focus your limited discovery cycles.

What AI Can’t Do For You (this is the most important part)

This is the trap I see most often: outsourcing discovery entirely to AI. It’s tempting, especially when you see clusters and summaries that feel neat and convincing. But AI can only tell you what customers say, not what they actually mean.

The fintech PM I mentioned earlier didn’t stop with the AI output. She still spoke with a handful of customers, pressure-tested the insights, and aligned them with broader strategy. That’s the difference between using AI as a tool and using it as a crutch.

TL;DR the old rules still apply.

Takeaways for Product Leaders

• Treat AI as your research assistant, not your decision-maker.

• Use it to save time on synthesis, so you can spend more time in conversations with customers.

• Always apply a human check. AI can point out patterns, but only you can decide what matters for your product and your strategy.

Where This Is Heading

By now, most PMs have experimented with ChatGPT or similar tools. The novelty is wearing off, and the real question is: how do we integrate AI into the everyday craft of product management? My view is that in a few years, every PM will be expected to know how to use AI tools in discovery the same way we’re expected to run a decent user interview today. It will become part of the baseline toolkit.

The teams that figure this out early will have an edge. Not because they skip discovery, but because they finally have the time and clarity to do it properly.

AI won’t do discovery for you, but it will help you uncover patterns faster, prioritize smarter, and spend more time on the conversations that actually matter.