AI boosted developer output across a 250-engineer organization – a measured +14% productivity gain. But only +6% reached delivery. The other 8 points evaporated somewhere between "code written" and "code shipped." This is a field study of what happened next: how the team traced the leak to code review, what two changes they made (neither touched their AI tooling), and which numbers they're now watching to see if the gap closes.
Phase two – the before/after numbers on review time and deployment frequency - lands in Q4 2026. If you want them when they publish, [connect with Anita on LinkedIn. And if "coding got faster, everything else got slower" is happening in your org right now, we'd love to compare notes: Let's meet.
