Problem
AI initiatives almost never die. They stay alive by inertia, even when they no longer deliver value.
When no one can cut them, the system fills with zombie projects: they consume hours, budget, and credibility without delivering better decisions.
Thesis
Without an operational kill‑switch, the portfolio becomes politics. A clear criterion eliminates theater and protects focus.
Callout — A kill‑switch is not an emotional veto. It is a governance protocol that defines when continuing is no longer rational.
Framework
Three death signals that can be measured without noise:
- Real adoption: % of teams using the system at 30 days. If it does not exceed a minimum threshold, the value is hypothetical.
- Reversal cost: hours/month and euros trapped in an initiative that does not scale.
- Net impact: real improvement in an operational KPI, not in intermediate outputs.
| Signal | Operational metric | Minimum threshold |
|---|---|---|
| Real adoption | % of teams using the system at 30 days | Defined before the pilot |
| Reversal cost | hours/month and euros trapped | Must not grow for two consecutive cycles |
| Net impact | real improvement in an operational KPI | Must be sustained for 2 cycles |
Case (anon): a services company reported 40% savings by automating reports. The kill‑switch showed that the savings came from eliminating manual tasks, not from AI. The project was closed and effort redirected to a process with real adoption.
Another false signal: “positive feedback” without sustained use. If the team applauds the pilot but does not use it after 30 days, there is no value, only narrative.
Anti-example: keeping an initiative because of sunk cost (“we’ve already invested too much”) even though reversal cost rises each week.
Position: Continuing by inertia is not optimism, it is operational debt.
Breathing: What breaks first is not the budget, it is confidence that the system knows how to decide.
When NOT to activate the kill‑switch: when the hypothesis has not been tested and no agreed metric exists.
Quick checklist before cutting
- Did adoption exceed the agreed threshold?
- Is reversal cost under control?
- Is net impact sustained for two consecutive cycles?
Life signal: when an initiative improves an operational KPI for two consecutive cycles without inflating hidden costs. If you cannot define that signal before starting, it is not an initiative, it is an experiment. And experiments are not “kept alive”: they are tested, measured, and cut.
A well‑designed kill‑switch reduces internal politics because it eliminates the “endless debate”. If the protocol is public, the closure stops being a fight and becomes a system consequence.
Protocol (3 steps)
- Define the minimum threshold for adoption, impact, and reversal cost by initiative category.
- Assign an owner with authority to close without massive consensus.
- Review the portfolio bi‑weekly: if a signal falls below the threshold, freeze or cut it.
Related:
- Decision Quality: the KPI that replaces speed
- Decision Rights Map: who decides what in an AI system
- 10 errors that sink AI initiatives in mid‑size companies
Next step
If today you cannot explain why an initiative remains alive, schedule a diagnosis at contact.
Translated from the Spanish original with AI assistance and reviewed for accuracy. Read the original in Spanish.