Almost every engineering team has adopted AI, and almost every developer will tell you they feel faster. But ask whether it's actually working, where it's paying off, and what it's costing, and most leaders can't answer. Adoption has outrun our ability to measure it, and there's no playbook for leading the transformation.
Anastasia Zamyshlyaeva, VP of Engineering at GitKraken, has been closing that gap from the engineering leader's chair. When existing tools couldn't tell her whether her developers were really adopting AI, or what to do about it if they weren't, her team built a way to see it, proved it on their own engineering org first, and then put it in the hands of engineering teams everywhere. Her argument: treat your engineering workflow like a product, measured with rigor and improved deliberately instead of on gut feel.
Seeing where AI works is only half the job. The other half is scaling it, moving a developer from babysitting one agent to orchestrating several in parallel, without the cost or the chaos getting away from you. She'll get honest about what's hard: why the obvious metrics mislead, how to tell real leverage from work that just looks busy, and what it takes to move a team from experimenting to running agents at scale.
