Let the constraint set the order, not the biggest number.

When two teams bring two AI priorities, the one constraint holding the rest back decides what comes first, not the bigger number. Start from the goal both priorities are meant to serve.

The reason is in how work runs. It moves as a chain, so change one step and the bottleneck moves. A large gain upstream of an unchecked step ships unchecked work faster.

Agree the evidence standard before the debate: what counts, how it’s measured, and what result triggers which decision. Then let the constraint pick the first move, one step up from where you are, with the rule that placed it printed on the plan.

The order is a leadership decision. BCG found 72 percent of CEOs now say they’re the main decision-makers on AI. With the rule on the page, both teams can see why the order is what it is.

Constructed example

At Orrin & Vale, the constraint holding the rest back is verification design: no AI-assisted output in Content has a designed check. So Content goes first, and Performance follows at the first gate.

So: before two priorities compete, write down the goal they both serve and the rule that will order them. Let the constraint pick the first move, and give the other priority a gate where it comes in.

Sources checked October 2026