Every AI programme reaches the same meeting. Someone has a list of candidates, everyone in the room has a favourite, and the loudest preference wins because there is no shared basis for saying otherwise. The list is usually fine. The ordering is where the money goes.
We score instead of argue. Five factors, applied to every opportunity a Scan surfaces, producing a number that appears next to each line in the report. The number is not the point — being able to show why one line sits above another is the point, because that is what survives contact with a board.
The five factors
Why these five and not others
Notice what is missing. Nothing about model sophistication, technical elegance, or how impressive the thing sounds. Those correlate with engineering enthusiasm and not with outcomes.
Notice also that three of the five — readiness, density, ownership — are constraints rather than attractions. That asymmetry is deliberate. The failure mode in this room is not picking a bad opportunity; it is picking a good opportunity that cannot be delivered yet. Loading the score toward constraints catches that.
A ranking that only measures upside will always hand you the most exciting project. That's the one that eats a year.
How the order defends itself
The reason to score before the meeting rather than during it is that the score changes what kind of conversation is possible.
Without one, disagreement is about taste, and taste is settled by seniority. With one, disagreement has to name a factor. Someone who wants the shiny project moved to the top has to argue that its readiness is higher than scored, or that its owner is more committed than scored — and those are checkable claims. Half the time they are right, the score gets corrected, and the ranking is better for it. The other half, the argument ends on its own.
The conversation about the shiny one
It comes up every time, and it is worth handling well rather than winning.
The shiny project is usually genuinely valuable. It scores badly because of readiness or ownership, not because the idea is bad. So the useful answer is not "no" — it is "not first, and here is the smaller thing that makes it possible."
That reframing does two things. It keeps the favourite alive, which matters politically. And it converts a vague ambition into a sequenced plan, which is what the board wanted in the first place. Nobody objects to their idea being second when second comes with a date and a prerequisite that is already funded.
Doing this without us
You do not need a Scan to run this. Take your current list, score each line one to five on the five factors, and multiply value by the geometric mean of the other four rather than summing — you want a low readiness or absent owner to drag the total down hard, not average out.
Then look at the top three and ask the only question that matters: if this works, whose number moves, and when will we first be able to tell?
If you cannot answer that for the project at the top of your list, the list is not the problem.
