Product discovery has always been slow and resource-heavy. The average discovery loop takes about two weeks and roughly 45 hours of PM time per cycle, and 20% of loops run more than twice before reaching a decision. The traditional five-step process (Observe, Define, Ideate, Test, Learn) is sound in theory, but the manual aggregation and synthesis required at each step consumes time that could go toward actual learning.
At PR_D_CT DAY 2025 in Ghent, Pontus Gifvas, co-founder of Swedish product discovery company Nalvin, walked through how AI agents are reshaping each phase of the discovery loop. Grounding his talk in established frameworks like Opportunity Solution Trees, Jobs-to-be-Done, and Dual Track Development, Pontus showed four specific processes where AI augments the loop: turning unstructured data into insights, refining and validating hypotheses, making product decisions with confidence, and closing the feedback loop after release.
Original Presentation Slides
Download the slides from this talk as presented at the gathering.
