By Teresa Torres · producttalk.org · @ttorres on X · LinkedIn
Teresa Torres announces an expanded partnership with Vistaly to bring AI-assisted discovery tools into its opportunity solution tree platform. The core problem she describes is that many teams conduct customer interviews but struggle to synthesize them into actionable structure. Her proposed approach synthesizes each interview individually first, then synthesizes across interviews, producing a draft opportunity solution tree that humans then refine. She explicitly rejects one-click tree generators built on made-up data, arguing that the tree's value lies in aligning teams around real customer evidence. The post also recruits alpha partners to help shape the product.
01Key takeaways
- Synthesize each customer interview on its own before looking for patterns across interviews to keep important context.
- Treat AI-generated opportunity solution trees as drafts that you review, refine, and reorganize with your own judgment.
- Build opportunity trees from real customer interview data, not from assumed or invented market descriptions.
- Combining expert human judgment with AI analysis tends to outperform either working alone.
- Stalled research is often a synthesis problem, so structure the step from recordings to opportunities deliberately.
02Key sections
- The synthesis gap
- Many teams interview customers but stall before turning recordings into insight. The bottleneck is the cognitively demanding step of extracting moments, identifying opportunities, and organizing them.
- Two-step synthesis approach
- The tool analyzes each interview separately to preserve nuance, then synthesizes across interviews. This contrasts with AI tools that skip straight to cross-interview analysis.
- Draft, not a finished answer
- The output is a starting draft the team reviews and reorganizes. Torres distinguishes this from one-click tree tools that generate trees without real customer data.
- Human plus AI collaboration
- Her experiments showed AI caught opportunities she missed and vice versa, supporting a model where AI produces the draft and experts supply judgment.
- Alpha partner call
- Teams can apply with three related customer interviews, with a preference for story-based interviewing and a range of experience levels.
03From the post
“Most product teams know they should be interviewing customers. And more teams than ever are actually doing it. That's the good news. The bad news? Many teams are struggling with what comes next. I've written before about the challenge of interview synthesis—going from a stack of recordings”
04Frameworks mentioned
Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.