Bain

The talk explains how boards are shifting from asking whether AI speeds up engineering to asking what measurable value it delivered and whether Finance can validate the result. It recommends agreeing on business outcomes, baselines, attribution, and all-in workflow costs before deployment, then tracking both near-term productivity signals and longer-term results. The speakers argue that AI raises throughput only when teams improve the surrounding SDLC: context, routing, verification, decision rights, and engineering practices. They also describe governing consumption by workflow risk and continuously reducing cost per accepted unit while improving velocity or quality.