The idea
A working demo is persuasive, and it is easy to conclude that most of the work is done. What the demo has shown is that the idea is possible. It was run by the people who built it, on data they chose, with nobody depending on the answer.
Production is the opposite on every count. Getting there is different work, and much of it is deciding things, not building things.
Four stages, four exits
- Prototype. A narrow version, measured against a baseline. It ends with a go or no-go.
- Pilot. Real users on live data, with access control. It ends with evidence of real use.
- Production. Security review, monitoring, and a team that owns it. It ends with a system your team operates.
- Scale. More teams and sites, with governance and cost controls. It ends with a pattern you can repeat.
What leaders have to decide
Engineers can build each stage. They cannot make these decisions for you:
- What result moves the project forward, and what result stops it.
- Who decides, and by when.
- Which team will run the system after launch.
- When security and IT are brought in. The answer should be early.
When a project stalls, it is usually because one of these was never decided.
A made-up example
A company has an internal assistant that has been shown to leadership several times over many months. Everyone likes it. Nobody outside the project team uses it, no one has said what would make it official, and IT has not reviewed it. It is called a pilot. It is a prototype with an audience, and it will stay one until someone names the exit.
Exercise · about 10 minutes
Pick one AI project in your organization, or one you are considering. Write down which stage it is really in, and one sentence describing what would end that stage.
One thing to take to your team
For every AI project, ask two questions: which stage is it in, and what ends that stage?
For more depth, read our whitepaper Why AI pilots stall before production.