How to Do Product Discovery with AI

The Discovery Trap

Product managers are under more pressure than ever. The pace of technology is accelerating, leadership wants faster delivery, and somewhere in the middle, product discovery—the thing that actually ensures you’re building the right product—keeps getting squeezed out.

It’s a trap a lot of teams fall into. From the outside, everything looks like momentum. Sprints are full, roadmaps are packed, features are shipping. But inside? People are feeling friction. Business cases that don’t hold up under scrutiny. Stakeholders who weren’t aligned until it was too late. Roadmaps driven by the loudest voice in the room rather than actual customer need.

The fundamentals of product discovery haven’t changed. What’s changed is the environment they’re operating in. And for a lot of teams, discovery has become something you do as a ritual, a checkbox before you get to the "real work,” rather than the confidence-building process it’s supposed to be.

That’s the core tension Adam Davis, CEO of Colab Cohorts, unpacked in a recent webinar, Rethinking Product Discovery in 2026. Teams are moving fast, but underneath that speed there’s a lot of uncertainty. And in an AI-native world, product teams have both an opportunity and a responsibility to do discovery better than ever.

Here’s what modern teams are doing to fix it.

Your Product Discovery Process Probably Has Too Many Steps

One of the most eye-opening exercises Adam shared is deceptively simple: map every step your team takes from insight to launch. Not the idealized version. The real one.

When Colab does this with clients, they typically find somewhere between 40 and 60 steps. Data gathering, analysis, reviews, prioritization sessions, handoffs, approvals. Each step is individually defensible. Collectively, they become a weight that slows everything down and introduces points where things quietly fall through the cracks.

Once you’ve mapped the steps, the next question is the important one: which of these actually need a human in the loop?

Some steps, like synthesizing customer research, aggregating feedback, monitoring churn signals, tracking competitor moves, can be largely automated with the right tooling. Others, like shaping the direction of a product, building the narrative for a business case, or deciding what’s worth building at all, need human judgment at the center. AI can do the heavy lifting, but a person still needs to shape the decisions.

This framing—automate what you can, keep humans in the loop where it matters — is the starting point for a modern product development lifecycle.

The Three Buckets of Great Product Discovery

Adam’s framework breaks discovery into three interconnected areas. Most teams are decent at one or two. The teams pulling away from the pack tend to be strong across all three.

1. Gathering: Go Further Than Customer Interviews

The most common mistake in the gathering phase is over-relying on internal subject matter experts or running a handful of customer interviews and calling it done. That’s a starting point, not a foundation.

Strong gathering pulls from multiple signals: usage data, sales and support conversations, market context, competitive intelligence, and CRM data. The goal is to get close enough to the problem that you stop guessing.

And if you’re in an industry where direct customer access is limited (B2B, regulated environments, enterprise sales cycles), you get creative. Join sales field trips. Spend time in communities where customers hang out. Set up automated signals that bring market intelligence to you rather than requiring someone to chase it down.

2. Analysis: Most Teams Stop Too Early

Gathering data is the easy part. Analysis is where most teams lose momentum. Either teams stop before they’ve converged on anything meaningful, or they confuse symptoms with root causes.

The analysis steps that tend to get skipped are clustering themes into coherent problems, separating what customers say from what’s actually driving their behavior, and quantifying the impact and risk of different paths forward.

This is also where the commercial lens matters more than ever. Product teams are increasingly expected to go deeper on financial analysis: cost to serve, operational expenditure, return on investment. As delivery gets faster and more automated, the analysis stage is where product managers build the confidence that justifies the investment.

3. Framing and Shaping: The Art of the Business Case

You can do everything right in gathering and analysis and still lose in the room if you can’t communicate it. Framing is where discovery meets organizational politics, and it’s a skill that doesn’t get nearly enough attention.

The structural stuff—ROI, market sizing, TAM/SAM—is table stakes. What actually moves decision-makers is narrative tension. Adam frames it around three pillars: 

• the villain (the problem you’re solving) 
• the protagonist (your customer and the evidence behind them)
• And the stakes (what happens if you do nothing in the next 6 to 12 months).

The other thing worth knowing about framing: you don’t win budget in the room. You win it in the one-on-ones before. The more you can get senior stakeholders to co-create the case with you, the harder it is for them to say no to it. People don’t reject decisions they helped shape.

Where AI Actually Helps in Discovery (and Where It Doesn’t)

The honest answer is that AI can do a lot in the discovery process, but the teams getting the most out of it are pairing automation with genuine product craft, not using it as a shortcut around the thinking.

The highest-leverage automation opportunities Adam identified include:

• Aggregating and synthesizing customer feedback across channels
• Analyzing usage data to surface how customers are actually behaving
• Monitoring churn and retention signals
• Tracking competitor moves and market shifts
• Pulling CRM and commercial data into the discovery context

One area that’s generating a lot of interest right now is synthetic data and AI-generated customer personas. The verdict is still mixed. If you have a rich underlying dataset and a deep understanding of your customers, synthetic data can help you validate assumptions and move faster in environments where direct customer access is limited. But it doesn’t replace real conversations. The emotions, the jobs-to-be-done, the friction points that only come through in a live interaction, that’s still where the best product insights come from.

The bigger risk Adam flagged isn’t using AI too little. It’s tools sprawl: moving data between tool after tool during the discovery process, losing context at every handoff, and adding steps instead of collapsing them. The next frontier is AI that stays in the flow of discovery rather than living off to the side.

Discovery Isn’t a Phase. It’s a Loop.

One of the persistent myths in product development is that discovery is something you do at the beginning of a project and then move on from. In practice, the teams building the best products treat discovery as continuous, a loop that feeds every stage of development and keeps getting smarter over time.

What makes this more achievable now than it’s ever been is the tooling. Fast feedback loops, live prototyping, real-time customer signal monitoring—the infrastructure for continuous discovery exists. The question is whether your team has the habits and systems to actually use it.

The walls between product, design, and engineering are also coming down faster than most organizations have adjusted to. Engineers who get early exposure to the discovery context can get ahead of architecture questions. Designers are shifting toward brand and system custodianship while product takes on more of the solution exploration. Go-to-market teams that are included in discovery become genuine partners in launch rather than passengers.

The companies that will separate themselves in the next few years aren’t the ones shipping the most features. They’re the ones that can run discovery continuously, tighten the loop between market signals and product decisions, and land products that drive real commercial and customer outcomes.

The Teams That Win Will Move with Clarity, Not Just Speed

Speed is table stakes in 2026. Everyone is moving fast. The differentiator is moving fast toward the right things…and that only happens when discovery is working.

The product teams that are pulling ahead right now share a few things in common. They’ve mapped their process and gotten honest about the friction. They’re using AI to collapse the low-value steps and free up time for the thinking that actually requires a human. And they’ve built a continuous discovery loop that keeps them close to the customer and the market, even as everything else accelerates.

Productboard Spark was built for exactly this moment. It’s an agentic platform purpose-built for product managers that turns scattered customer and market signals into clear product plans — without requiring manual prompting or losing context between sessions. Spark guides teams through proven PM workflows with prebuilt prompts grounded in a decade of product management experience, so the output isn’t just fast, it’s actually good.

If you’re ready to make discovery a competitive advantage rather than a bottleneck, learn more about Productboard Spark (now in public beta). 

Frequently Asked Questions about Product Discovery

Answers to the most common questions about product discovery, straight from the webinar Q&A and the broader PM community.

What is product discovery and why does it matter?

Product discovery is the process of identifying, validating, and framing the problems worth solving before committing to building a solution. It matters because skipping or rushing discovery leads to compounding failure: fragile business cases, misaligned stakeholders, and products that don’t drive the commercial outcomes they were supposed to. Great discovery is ultimately how product teams de-risk investment and build organizational confidence in what they’re about to ship.

How do you do product discovery with AI?

AI adds the most value in the gathering and analysis phases of discovery. You can use it to aggregate and synthesize customer feedback at scale, surface themes in usage data, monitor competitive signals automatically, and pull commercial data from your CRM into the discovery context. Where AI is less suited to replace humans is in the judgment calls: deciding which problems are worth solving, shaping the narrative for stakeholders, and making the strategic calls that define direction. The best approach pairs AI for automation with human craft for decisions.

When should you skip product discovery?

There are a few legitimate cases: critical bug fixes and SLA-driven work, well-understood UI patterns where the solution is already obvious, and small-scope changes where your confidence is already high. A useful rule of thumb from the webinar…if the investment is less than two weeks and you have solid evidence behind it, you can move without a full discovery process. For anything bigger, skipping discovery is usually just borrowing risk from the future.

How do you involve stakeholders in product discovery?

The most important principle: you don’t win alignment in the room, you build it before the room. Getting senior stakeholders involved early—giving them a role in shaping the business case rather than reacting to it—dramatically increases your odds of getting a yes. For go-to-market teams, involve them early enough to help shape the value proposition, not just launch it. For engineering and design, bring them in once you have a firm view of the problem and the evidence behind it, so they have something concrete to react to and can start getting ahead of technical questions.

What is an Opportunity Solution Tree and how does it help discovery?

The Opportunity Solution Tree (developed by Teresa Torres) is a framework for structuring discovery around a specific outcome. You start with the outcome you want to achieve, identify the problem areas that are preventing it, and map out the solutions you’re exploring for each problem. It’s a way of keeping your discovery focused and traceable—so you can see exactly why a particular solution is on the table and what evidence is behind it. It also makes it easier to prune: if a problem area doesn’t have enough evidence, you can deprioritize it without losing the thread of your overall direction.

What tools are most useful for product discovery?

The most valuable capability is having all of your signals in one place. Productboard is designed for exactly this—centralizing customer feedback, usage data, and market signals so you can move from noise to themes without manual effort. Other tools that come up frequently include Intercom for customer communication signals, Amplitude for usage analytics, and Salesforce or Pipedrive for CRM data. The key is connectivity: the more your discovery tools talk to each other, the less context you lose moving between them.

What is continuous product discovery?

Continuous product discovery is the practice of treating discovery as an ongoing process rather than a project phase. Instead of doing discovery at the start of an initiative and moving on, teams build regular touchpoints with customers, maintain live feedback loops, and keep their understanding of the market updated in real time. The concept has been championed by Teresa Torres and is increasingly achievable with modern tooling. The challenge for most teams is less about knowing what it is and more about building the habits and infrastructure to actually sustain it.

How do you build a compelling product discovery business case?

Great business cases combine structural credibility with narrative tension. The structural elements—ROI projections, market sizing, cost-to-serve analysis—establish that you’ve done the work. But what actually moves people is the story: the problem (the villain), the customer experiencing it (the protagonist), and the cost of inaction (the stakes). The most effective cases also have senior stakeholder fingerprints on them before they go to the room. People are far less likely to push back on something they helped shape.

Watch the full webinar: Rethinking Product Discovery in 2026.