Short, honest essays from the work — diagnosing why good systems stall, and what it takes to make them stick.
A deployment designed right, built right, signed off by everyone — and stalled at the wall when the site decided, at go-live, that it wasn't ready.
Part 2 · PrescriptionThe fix isn't more conviction. It's an owner with authority and a gate that can say "not now" — Organizational Commissioning, the org's own IQ/OQ/PQ.
The empirical close, written after the first feature goes live: what readiness actually took, and whether the "operators can't handle it" story held up.
Your employees are already using AI — through personal accounts, outside your visibility. It's a structural failure, not an employee compliance problem.
Part 2 · The Failed FixRoughly 10% of training transfers to the job. Without a sanctioned place to practice, AI upskilling decays. The sandbox is the missing bridge.
Part 3 · The ModelCentralized governance can't scale across domains it doesn't understand. A named, dual-fluency steward embedded in the domain is what holds it together.