GitKraken's 2026 State of AI in Engineering report surveyed 554 developers and engineering leaders, and the title says the finding: everyone feels faster, almost nobody can prove it. 84% say AI tools made them more productive, 43% say much more, and fewer than 5% feel slower. 96.4% of teams have adopted the tools. Then the measurement question: 39% of organizations have no way to measure the impact at all, and only 20% measure productivity in any specific way. The rest rely on belief or self-report.
Usage is deep: 34% keep an agent running all day, 47% in enterprises.
I have written about self-reported velocity before and I will not repeat the argument. What I want to say here is who owns the gap. A developer cannot measure the productivity of a team. They can only say how they feel, and 84% of them feel good, which is real information about morale and nothing else. Measuring whether delivery changed is a management job, and the report says four out of five managers are not doing it.
I think that is the most actionable number in the survey. The tools are adopted. The feeling is positive. The missing piece is a leader deciding what outcome the tools were supposed to change, writing it down, and looking at it every month.
Does your team have a number it looks at monthly to see whether the tools are helping, or just a feeling?
Photo source: https://photos.robertstowe.com/colorado