Yardstick is organized into three suites you can adopt in order:
Observe (measure): measure AI on real outcomes in the units each vertical respects, such as cost per merged PR for dev tools or resolution rate for support. Activity metrics lie, outcomes do not.
Optimize (improve): a score is only useful if you can move it. Optimize reads what Observe measures and tells you how to raise it, the prompt to tighten, the model to swap, the experiment to run, then closes the loop.
Treasury (govern): once outcomes are measured, money can follow them. Set allowances per team, and reclaim unused budget and reallocate it to the teams producing the most value.
The long-term vision, Autopilot, is an agent that runs the whole loop within policy, fully logged and reversible.