AI is transforming how we work, but when it comes to product discovery, it’s not here to replace the way we think as Product Managers. It’s here to reduce friction.
If you spend hours transcribing interviews or clustering sticky notes, then yes, AI can help. But that doesn’t mean you should outsource your thinking.
Discovery is one of the most human parts of product work. It’s where we listen deeply, test our assumptions, and make sense of messy data. AI might speed up the process, but we’re still the ones who have to decide what matters.
In this article, we’ll look at where AI fits and where it doesn’t across the key stages of discovery, and how to blend it into our workflows without losing the critical thinking that makes discovery work in the first place.
What Product Discovery Looks Like
Before we talk about the AI’s role, let’s quickly revisit what product discovery includes.
Teresa Torres outlines three key activities in Continuous Discovery Habits:
- User research and interviews – Talking directly with users to understand their context and pain points.
- Assumption testing – Identify what needs to be true for your ideas to work.
- Opportunity mapping – Using tools like the Opportunity Solution Tree to map problems, opportunities, and solutions visually.
This is something I often mention when I talk about discovery. We don’t do discovery once and then we’re done. Discovery is an ongoing loop of learning, adjusting, and reducing risk. And AI might help us along the way, but the thinking still has to be ours.
How to Use AI in Product Discovery Without Losing Control
Product discovery happens in layers. And AI can play a different role at each one. But you need to be careful with it. Including AI in product discovery, it’s not about handing over the process. It’s about knowing where automation saves time and where judgment still matters most.
Let’s start with user interviews. This is where AI can offer support without replacing the human touch. Think about generating an interview guide. Tools like ChatGPT or Claude can help you start faster. You can use AI at this stage by feeding it the problem space, and it will give you a draft. But the questions still need to be reviewed and adapted by you. They need to reflect your product, your users, and your tone.
It can’t take your place during interviews. You are the one asking follow-ups, catching and sensing the hesitation of the users you’re interviewing, or building trust.
Then, once the conversation ends, you can use AI again. Transcription tools can scan and capture your calls instantly, and then AI can help summarize or tag key themes.
An example here is using the Jobs-To-Be-Done (JTBD) framework. In this model, you analyse calls based on what users are trying to achieve, what’s pushing them to act, what outcomes they desire, and what anxieties or obstacles they face. These are deeply emotional layers, and while AI can help extract surface structure, interpreting the why behind those jobs still requires your thinking and understanding of that situation.
Now, about assumption testing. Here, AI can help you brainstorm faster. Give it your product idea, and it might suggest 10 testable assumptions in 30 seconds. It can also help you write landing page copy or mock a fake door test. But selecting which assumptions are riskiest and deciding what results are meaningful are decisions that you need to make.
Finally, when you get to mapping opportunities, AI can assist by clustering insights, grouping user quotes, or helping you organize notes visually. But framing the real opportunity, choosing what’s worth solving, and aligning that with your product vision still demands a Product Manager’s thinking.
In all of these areas, AI is a great copilot, not the driver. If you use it wisely, it can make you faster and more focused. But it won’t replace the creative and strategic parts of discovery that make it valuable in the first place.
Let’s compare how discovery might look with and without AI support.
| Task | Without AI | With AI |
| Create an interview guide | PMs write it from scratch | Drafted with a prompt, then edited by the PM |
| Transcribe interviews | Transcribed manually | Transcribed automatically via tools |
| List assumptions | Brainstormed manually | AI generates suggestions that are reviewed with the product trio |
| Cluster insights | Added on sticky notes, spreadsheets, Miro, etc. | AI detects the themes |
| Decide on what to act | Team-led synthesis and framing | AI should not replace this part |
Remember that even with AI you still own the thinking. The tools help you go faster, but only you decide what matters.
Discovery isn’t just about following a set of steps. It’s often messy, and you’re sitting in that space between hearing what users say and figuring out what it really means. This is exactly where overusing AI becomes a trap.
Here’s a personal example from building this blog – Scrum District.
I could’ve asked AI “What should I write for beginner product managers?” And it would’ve given me a list with obvious suggestions: MVPs, backlogs, maybe some agile practices.
But that’s too generic and it skips the discovery work.
Instead, I have spoken with other Product Managers. And I noticed a pattern: they felt unsure how to run product discovery when AI tools were suddenly everywhere. They weren’t asking “What is discovery?”. They were asking “How do I still do it well when AI is part of the picture?”
This insight didn’t come from a prompt, but from listening.
But once I had that insight, AI became useful. So I asked “Based on this pain point – PMs feel lost trying to include AI in product discovery without skipping the thinking – help me come up with questions I can use to explore what they’ve tried and where they’re stuck.”
This is what a healthy discovery workflow looks like. You listen. You think. And you bring in AI when it helps you move faster, not when you’re trying to skip steps.
Let AI help you organize. But don’t let it clean up the thinking too much, or you risk missing the signal in the noise.
If you’re looking for a step-by-step breakdown of the traditional product discovery process (without AI), this article walks you through the logic and structure.
Step-by-Step Flow to Combine AI With Product Discovery
Here’s what a combined workflow might look like where AI supports your work, but it doesn’t lead it.
- Define the problem – You begin by clearly framing the uncertainty or opportunity you’re trying to explore. AI can sometimes help you stress test this framing, but only after you set the initial direction.
- Draft the interview guide – Use AI to generate a first draft of your research questions. You can prompt it with your goals and audience and it will return a rough structure. Then refine it to suit your product context and language, and make sure it really has something to do with what you need.
- Run interviews – This is where your involvement matters most. AI can’t build trust, ask thoughtful follow-up questions, or sense where the user is hesitating. Discovery is emotional and requires human connection. And that is your role as a Product Manager.
Ironically, this is also the step where we should be spending the most time, yet many Product Managers rush through it because of all the admin work they have to do.
But AI can take over the repetitive parts, like transcription or organizing notes, which frees up time to really focus on these conversations and get more from them. - Transcribe – Once interviews are complete, let AI tools handle the transcription. This saves hours. You can also use AI to extract and group answers into potential themes. But always review the raw text yourself to catch the nuance.
- Synthesize – AI can group similar quotes or cluster feedback. This is where tools like ChatGPT, Notion, or Miro AI can help. But then you have to bring your judgment: spotting contradiction, weighing patterns, and filtering the signals from the noise.
- Map opportunities – AI may help you lay out a structure, but choosing what to prioritize and what to drop requires your product strategy lens.
- Test ideas – Use AI to create fake door copy, onboarding flows, or even prototype variants. But then it’s for you to decide what makes a meaningful test and how you’ll evaluate the results.
When you work like this, AI becomes a helpful tool, not a shortcut. You’re still doing the thinking, but now you have more time to focus on what matters most: understanding your users and making smart product calls.
How to Use AI In Product Discovery Without Skipping The Thinking
AI is fast. But that speed can trick you into thinking you’ve done the work. When using AI in discovery, we should slow down at a few key moments to make sure we’re not just accepting whatever the tool gives us.
Whenever using AI, we should ask ourselves:
- Does this insight actually make sense in our market?
- Is it actionable or just a pattern AI grouped together?
- What is missing that AI didn’t surface?
This mindset can help us avoid skipping the messy thinking. For example, AI might suggest a top-clicked article, but if it doesn’t fit my blog vision or solve a real user problem, I won’t consider it.
AI should be a gap filler, not a starting point. We have to first define the user problem by ourselves and then use tools to help us phrase a test or draft an interview guide. We should not use it to define the strategy instead of us.
Final thoughts
AI won’t kill discovery, but it will change how we do it.
You’ll still need to:
- Be curious
- Challenge assumptions
- Make judgment calls
- See through the noise to what matters
But now, you can do these things with a little less admin work, a little faster synthesis, and a lot more focus on user interactions.