Create an account to save sources and durable progress.
Product Talk · Teresa Torres
articleplaybookSource excerpt
About Product TalkListen to this episode on: Spotify | Apple PodcastsAI evals feel like a brand-new skill — but the hardest part is a question product teams have always had to answer: what does "good" actually look like?Petra Wille and Teresa Torres dig into Teresa's recent
This is a limited feed-provided excerpt, not the full original work.
Apollo layer
Treat AI evaluation as a product-definition problem first: align on what a good customer outcome looks like before choosing metrics or testing outputs.
Without an explicit standard for quality, teams can collect evaluation data yet still disagree about whether the product is improving. A shared definition of success focuses experiments and decisions on customer value.
For one AI-powered workflow, ask: “What observable outcome would make a customer call this result good, and what would clearly make it unacceptable?” Turn the answers into a short evaluation rubric.
Private notes
© Apollo Advisors° ABN 34 346 108 139