Despite widespread adoption, trust in AI systems remains a challenge. One answer comes from a layer that often gets too little attention: AI evaluation. This talk explores evaluation layers for agentic AI systems in life science R&D. The approach uses AI to scale expert judgment rather than replace it, grading outputs on how scientifically grounded and plausible they are. In this context, evaluation is as much a product decision as it is a technical one.

Most product teams treat product-market fit as something you find once and move on from. In reality, for any company operating across more than one market, it's closer to a continuous, multi-year process of pivots, one that never fully settles. This panel challenges the "PMF as milestone" mental model and replaces it with a more honest one. Panellists bring the enterprise view: products built centrally often get commercialised very differently region by region, shaped by local regulation and customer expectations rather than the product team's original assumptions. The panel digs into why a product that wins convincingly in one region can fail to scale into another for reasons that have nothing to do with the product itself, and what that means for how teams should actually define and track fit. Attendee takeaways: - Why treating product-market fit as a one-time milestone sets teams up to misread their own results - How regulation and regional customer expectations reshape commercialisation, even when the product stays the same - A more useful way to track fit continuously, rather than declaring it "done"