AI transformations fail where all transformations fail, in the operating model. The specific AI failure is handing the whole thing to the people who know the tool and calling it adoption. The people who know the work have to lead, with the tool people alongside them.

When growth starts to strain the system, many teams “professionalise” by adding meetings, reporting, and process. It looks mature — and often makes you slower. This post shares a more practical alternative: find the constraint, focus hard, steer with one KPI, remove bottlenecks, and build a cadence that turns learning into action. Keep the physics. Ditch the ceremony.