I’ve sat in enough leadership rooms to know the shape of the deck before it opens: a two-by-two matrix, four quadrants labelled things like “quick wins” and “strategic bets”, and a maturity curve borrowed from a consultancy template. Nineteen years ago at Deloitte I helped build decks that looked exactly like this. They’re not wrong, they’re just not enough on their own.
The problem with the matrix
A maturity matrix tells you where a use case theoretically sits. It doesn’t tell you whether anyone on your team can actually ship it, or whether the data it needs exists in a usable form, or whether the person who’d own it has the bandwidth this quarter. Those three questions kill more AI initiatives than strategic misalignment ever does.
A shorter list that actually works
Instead of a matrix, three questions, asked about every candidate use case, in order:
Does the data already exist, in a form a system could use today? Not “could we collect it eventually.” Today. If the answer is no, this isn’t a Q1 project, no matter how strategic it looks on a slide.
Is there one named person who owns this, with time freed up to own it? AI initiatives without a specific owner become everyone’s fifth priority, which means nobody’s first one. If you can’t name the person right now, don’t greenlight the project right now either.
What does wrong output cost you, and who notices first? This determines how much human oversight the workflow needs from day one, which determines the actual build cost, which the matrix never captures.
What this replaces
Not the strategic thinking, just the theatre around it. The maturity matrix exercise can take a leadership team most of a day. These three questions, asked honestly about each candidate use case, take an afternoon and produce a shorter list that’s more likely to survive contact with your actual team and your actual data.
That shorter list is the roadmap. Everything else was slides.