Once in place, does AI still need work?

Many AI projects are budgeted like conventional projects: a build phase, a go-live, then light maintenance. The project team leaves, and the tool is supposed to run on its own.

That is not the case. Data changes, uses shift, suppliers update their models, regulations become clearer. You have to measure results, review samples, adjust instructions, sometimes revisit the level of autonomy or stop a use that has become pointless. It is the last step of our method, and it leads back to the first: what you observe in production redraws the map.

• The running cost is often forgotten.Supervision, adjustments and subscriptions weigh every year, beyond the launch.
• The business inherits a tool it does not know how to develop.The skills left with the project team.
• The best ideas come after going live.It is through use that you discover what more could be done, or done differently.

1. 01Include in the initial budget the time and resources to keep each use alive.
2. 02Name a business owner for the use, trained to monitor and adjust it.
3. 03Review each use after six months: what works, what is drifting, what is changing.

• How do you know whether an AI use is drifting over time?→
• How do you keep up with a technology that changes every month?→
• Who should govern AI in the company?→
• How do you measure the return on investment of an AI project?→