The Day-90 decision gate
Make one of three calls against your Day-1 threshold: scale (it beat the bar), pivot (promising but mis-scoped), or stop (it missed). The discipline is in refusing to scale a miss or let a pilot linger undecided.
Why the gate must be binary
A vague “let’s keep going” is how pilots become zombies that consume budget without ever scaling. Comparing results to a pre-set threshold forces a clean scale, pivot, or stop decision.
The most expensive AI outcome is not a failed pilot, it is a pilot that neither succeeds nor dies. Zombie pilots consume attention, licenses, and credibility indefinitely. The Day-90 gate forces resolution: measure against the threshold set on Day 1, then scale it, reshape and re-run it, or kill it. Mature adopters who run this discipline see payback in 6–9 months because they concentrate resources on winners instead of spreading them across the undead.
Frequently asked questions.
What if leadership does not want to kill a pilot they championed?
This is exactly why the threshold is set on Day 1, before anyone is emotionally invested. A pre-committed number depersonalizes the decision, you are not killing someone’s idea, you are honoring the criteria everyone agreed to. Naming a single accountable sponsor also makes the call cleaner than a committee ever could.