AI assistants now write a growing share of the code your team ships, but most review processes and quality gates were designed for human authors. Reviewing more code faster isn't a strategy. In this lightning talk, Albert Franquesa introduces Control, Evidence, Trust, a simple model for leading teams through this shift: enforce your standards before code is written, measure what AI actually contributes and what it gets wrong, and use that evidence to decide where agents can safely do more and where humans must stay in the loop. Trust in AI code isn't a feeling. It's earned with control and proven with evidence.
This talk has been presented at TechLead Conf London 2026: Adopting AI in Orgs Edition, check out the latest edition of this Tech Conference.






















