Why AI governance fails on a Plan-Build-Run operating model, and what executives must redesign
Most organisations that have deployed AI can show an AI policy, a risk assessment and an ethics committee. Four national audit offices have looked behind those documents and found the same thing: the governance exists, and nobody can show what the AI returned.
The cause is the operating model underneath. Governance written once and added on top of Plan-Build-Run cannot keep up with AI systems that change every quarter. Drawing on Evolve: The Operating Model AI Demands, this whitepaper sets out where human oversight belongs, who owns the instructions AI agents follow, how the portfolio decides at machine speed, and the order in which executives change each.
It includes five tools for executives: a classification of AI decisions by reversibility, an instruction ownership map, a four-sentence investment case, four fields for any automation business case, and an AI governance readiness check.
