Should you know how to stop an AI project?

An AI project that does not deliver is rarely stopped. It is extended, adjusted, people wait for the next version of the model. Nobody wants to own the decision, because stopping looks like admitting a mistake.

These projects on borrowed time are expensive: budget, team time, and credibility for the projects that follow. Knowing how to stop means having set in advance what justifies continuing: a measurable result, by a given date, at an acceptable cost. When these criteria exist, the decision becomes a shared observation rather than a rebuke. And a well-documented stop teaches almost as much as a success.

• Stopping criteria are rarely set at the start.Without them, every deadline becomes a negotiation.
• Attachment to the project clouds judgement.Those who carried it are the least well placed to decide to stop it.
• A project stopped early costs less than one that drags on.The resources freed up serve the next project.

1. 01For each project, write down before launch what must be achieved, and by when.
2. 02Entrust the decision to continue to someone who did not carry the project.
3. 03When you stop, write down what you have learnt, and share it.

We help decide, project by project, whether to relaunch, redirect or stop, knowing why.

• Why do our pilots never scale up?→
• How do you measure the return on investment of an AI project?→
• Who should govern AI in the company?→
• Is one big project better than several small ones?→