Ownership of failure: who is allowed to say we should kill this

A portfolio lead at a mid-market company told me about the AI project that everyone privately knew was failing and no one had the standing to stop. It was not working. The people close to it could see that. But killing it meant declaring a failure, and no one owned the authority to make that declaration. The champion did not want to admit the project they pushed for had failed. The team did not have the standing to kill their own work. Leadership was not close enough to know it should be stopped. So it ran, month after month, consuming budget and people, not because anyone believed in it but because no one owned the decision to end it. Stopping a project turned out to be much harder than starting one, and the missing piece was ownership of the kill decision.

He named the asymmetry directly. Orgs assign the authority to start projects and forget to assign the authority to stop them. So failing projects do not get killed. They get abandoned slowly, or worse, they keep running on inertia because ending them requires an owner who does not exist. Someone has to own the authority to say this should die, and it has to be someone who can say it without it being a personal defeat.

The authority to stop is an ownership too

Starting a project has clear ownership. Someone decides to fund it, staff it, launch it. Stopping one usually does not, and that gap is expensive. A failing AI project that no one has the authority to kill keeps consuming resources long past the point where the failure is obvious, because killing it is socially and politically costly and no one owns that cost. The people who could see it should stop lack the standing, and the people with the standing lack the closeness to see. This is a real and under-discussed way that AI budgets get wasted. Not on projects that fail fast, but on projects that fail slowly because no one owned the decision to end them.

The teams that manage their AI portfolio well assign ownership of the kill decision up front, alongside the go decision. Someone holds the explicit authority and responsibility to stop the project if it is not working, chosen so that killing it is a legitimate professional call rather than a personal admission of defeat. They define, in advance, the conditions that would justify stopping, so the kill decision is a pre-agreed judgment rather than a fraught confrontation. Owning failure is as important as owning success, and it is the ownership most orgs forget.

Building the kill decision in

Name who can kill it. Assign the authority to stop the project, up front, to someone who can exercise it without it being a personal defeat.
Define the conditions to stop. Decide in advance what failure looks like, so killing the project is a pre-agreed judgment, not an improvised confrontation.
Separate the kill owner from the champion. The person who pushed for the project is the wrong one to declare it dead. Give the kill decision independent standing.
Treat stopping as a legitimate outcome. A killed project that should have died is a success of the process. Make ending things a respected call, not a shameful one.

How we approach it at Density Labs

In our AI Opportunity Assessment ($2,500), we build the kill decision into the plan, because orgs assign the authority to start projects and forget the authority to stop them, and failing AI projects run on inertia as a result. We name who can kill the project, define the conditions that would justify it, and give that authority independent standing. Assigning ownership of failure up front is far cheaper than a project everyone knows is dead that keeps running because no one owns ending it.

Any org can start an AI project. The ones that spend well can also stop one. Name who is allowed to say we should kill this, and define when, before the project you cannot stop starts running on inertia.