Four independent 2026 publications, read together, give one answer to why AI programmes stall. Individual productivity is up, enterprise returns are flat, complexity is eating benefits, and the costs that produce the dip have names. The numbers, with sources, and what a sponsor does with them.
Programs do not slip because work is slow. They slip because decisions are slow. How to measure decision latency, what good looks like, and why it is the first number AI actually moves.
The PMO that survives 2026 is not the one with the most dashboards. It is the one that redesigned its weekly operations around agents and kept judgment for humans. Here is what that looks like in practice.
The PMO is not dead, but the version of it that survived the 2010s is. Here is what the function looks like in 2026, and what stops working if you keep running last year's playbook.
Process without values feels like bureaucracy. Align controls to what the company says it stands for.
A PMO that only reports is a spectator. Here is how to turn it into a room where choices get made.