Continuous Product Discovery: 5 Lessons from Teresa Torres
5 lessons from Teresa Torres on continuous discovery, customer context, and testing assumptions in the age of AI.
AI is changing how quickly product teams can build. Ideas that once required significant engineering effort can now be prototyped and tested in a fraction of the time.
But faster development raises an important question: Just because we can build something, does that mean we should?
That question was central to a recent conversation between Teresa Torres, author of Continuous Discovery Habits and founder of Product Talk, and Jack DeVita, Head of Customer Success at Enterpret.
Their discussion explored what AI changes about product development, what remains essential, and how teams can make better decisions when building becomes easier.
Here are five lessons for product teams navigating this shift.
1. Building faster doesn't mean building better
AI-powered development tools are making it easier to turn ideas into working products. But the ability to build more doesn't necessarily mean customers want more.
In fact, adding features without understanding the problems they solve can make products harder to use.
Teresa emphasized that customers generally want enough functionality to accomplish their goals effectively, not every possible feature a team could imagine.
The challenge for product teams is maintaining the discipline to build only what meaningfully addresses customer needs.
That starts with the fundamentals of continuous discovery: Define the outcome you're trying to achieve, understand the customer's circumstances, identify their needs, and explore solutions that address those needs.
It's a straightforward approach in theory. In practice, teams often jump directly to solutions before fully understanding the problem.
The takeaway: Faster development should create more opportunities to learn, not simply more opportunities to ship.
2. Customer feedback is a signal, not the full story
Product teams receive feedback from countless sources: support tickets, feature requests, sales conversations, and customer interviews.
These signals can help teams identify recurring frustrations and potential opportunities. But there's an important difference between knowing what customers are requesting and understanding why they need it.
Teresa explained that much of the feedback teams receive is missing essential context.
A customer might request a feature, but that request alone doesn't reveal what they're trying to accomplish, what's preventing them from succeeding, or how the feature would fit into their workflow.
Without those details, teams can make assumptions about the problem and end up building solutions that miss the mark.
This is where understanding customers' actual experiences becomes critical.
What are they trying to achieve? What obstacles are they encountering? What does their environment look like? Who else is involved?
The answers help teams move beyond individual requests toward a clearer understanding of the underlying opportunity.
The takeaway: Customer feedback can reveal where to investigate. Understanding the context behind that feedback helps teams determine what to build.
3. AI can support discovery, but it can't replace real customer understanding
AI creates new possibilities for product discovery, from synthesizing research to helping teams process large volumes of customer information.
But the conversation highlighted an important distinction between using AI to support research and relying on it to replace real customer evidence.
Teresa discussed the growing interest in synthetic users and AI-generated research participants, including the limitations of treating simulated responses as reliable substitutes for actual customer experiences.
Her perspective reinforced the importance of grounding product decisions in evidence from real people.
AI-generated responses may help teams explore ideas or formulate questions, but teams still need to understand how customers behave in their actual environments.
Direct customer conversations remain valuable because they reveal circumstances, motivations, and constraints that may not be apparent from a feature request or summarized feedback.
The opportunity is to use AI thoughtfully while remaining clear about what the available evidence can and cannot tell us.
The takeaway: AI can help teams work with customer information more efficiently, but it shouldn't replace learning directly from customers.
4. Roadmaps should reflect what teams are learning, not just what they plan to ship
Traditional roadmaps often organize work around features, projects, and delivery deadlines.
But committing to solutions too early can leave teams with little room to respond when new evidence challenges their assumptions.
During the conversation, Teresa explored how teams can use a now, next, later roadmap to balance immediate priorities with longer-term opportunities.
This approach helps teams distinguish between work they're ready to pursue and opportunities that still require investigation.
It also creates space for discovery before development begins.
Teams can explore customer needs, identify potential solutions, and test important assumptions about future opportunities without committing prematurely to a specific implementation.
The goal isn't to eliminate planning. It's to make planning more responsive to what teams are learning.
The takeaway: A useful roadmap provides direction while leaving room for evidence to shape decisions.
5. Assumption testing is about reducing risk, not proving you're right
How do product teams know when they've tested an assumption enough?
Teresa challenged the idea that discovery is simply about validating or invalidating an idea.
Product decisions rarely come with absolute certainty. Even when research supports an assumption, new information can change how a team understands the opportunity.
Instead, Teresa encouraged teams to think about assumption testing as a process of progressively reducing risk.
Teams identify the assumptions that could most seriously undermine a solution, gather evidence to address them, and reassess which uncertainties deserve attention next.
As confidence in one area increases, another assumption may become the most important risk to investigate.
This approach helps teams avoid spending too much time seeking certainty about a single question while overlooking more consequential unknowns.
It's particularly relevant as AI makes implementation faster. When teams can build quickly, there's even more value in identifying the risks worth investigating before committing to a solution.
The takeaway: Discovery isn't about proving an idea is right. It's about reducing uncertainty enough to make a better decision.
When building becomes easier, understanding matters more
AI is changing how product teams work, but the fundamentals of building successful products remain remarkably consistent.
Teams still need to understand their customers, identify meaningful opportunities, and make thoughtful decisions about where to invest.
What changes is the speed at which they can explore possibilities, test solutions, and turn ideas into products.
That speed creates an opportunity to make discovery more effective, provided teams don't confuse the ability to build with evidence that something should be built.
The teams that succeed won't necessarily be those shipping the most features. They'll be those that consistently connect what they're building to the problems their customers actually need solved.
Because when building becomes easier, knowing what matters becomes even more important.
This article summarizes highlights from an October 29, 2026 conversation between Teresa Torres and Jack DeVita, hosted by Enterpret. Learn more about Teresa's work on continuous discovery at Product Talk.
