An example I made up
The situation. The chief operating officer of a regional logistics company, with its head office in Penang, has a proposal waiting for her decision. It's a one-year AI project to plan the next day's delivery routes for the dispatch team. The board wants her answer at its next meeting, and she wants a charter she can sign.
What the AI produced. She asks WhitethroatAI, the assistant her office uses for drafting, to write the project charter, with van-hours as the measure of success. It comes back in minutes with one line at the top: "Charter ready: route planning by AI from 1 February, success measured as 12% fewer van-hours by year end. Based on the vendor's trial plan and last year's route data. I am 90 percent sure." Van-hours are the hours the company's vans spend on the road. The charter is complete on the measure she asked for, and nothing in it is wrong.
The check. She doesn't argue with the number. She asks a different question: a year from now, who will say whether this project worked? The charter doesn't answer that. So she phones the four dispatch supervisors, who plan the routes today and whose daily work the project will change. All four ask her the same thing. What happens to the delivery windows they agreed with customers and still set by hand?
The decision. She approves the charter with a second condition, which the supervisors write themselves: no customer loses an agreed delivery window. One supervisor is named in the charter as the person who answers for that condition. She sets a review at six months. She tells the four supervisors first, then writes to the board with the charter as approved.
What happened after. At the six-month review, van-hours were on track. One proposed change to early-morning routes would have moved some deliveries outside the windows customers had agreed. The supervisors' condition stopped it before it went live.
The story, briefly
At twenty-nine I was at Electrolux, and the growth was 241 per cent. I was dismissed within the year.
I won't retell the rest here. I bring it up for one reason: a good number and a good outcome are not the same thing.
What it means for an AI project
An AI project can reach its number and still not be the success the people whose work it changes needed. The charter in my example was right about van-hours. Without the supervisors, the project could have met its target and still broken promises the company had made to its customers.
That's why the start matters most. At the start, before anyone has spent a month on the work, the leader still decides what success means. Later, people look at the number and little else.
An AI assistant can draft a charter on whatever measure you give it. Someone still has to ask the people whose work will change what must stay true while it happens. That is the leader's job. In my example it took four short conversations.
So before you sign, ask who will say whether the project worked. If the answer is only a figure in a report, the people whose work will change haven't been asked yet. They will live with the new way of working. They are often the first to see when a promise to a customer no longer holds.
I have a line for this:
"You can be right about the technology and wrong about the humans. And the humans will always win."
In a project charter, the humans show up as conditions. Write them in before the work starts, and they protect the project. Leave them out, and you meet them later, when it costs more to change the plan.
What I'd write in the charter
I'd keep the number, because a clear target gives everyone something to check. Next to it I'd put one condition, written by the people whose work will change, in their own words. Each line gets the name of the person who answers for it, and a date when we'll look at both. That is the whole addition: one more line, and a name beside each line.
The example is made up and proves nothing about any real project, and a second condition won't make a project succeed on its own. What it does is make clear, at the start, what success has to include and who will say so.
I give keynotes, briefings and workshops on this.