RICE scores a candidate project on four factors: Reach (how many people or transactions it touches), Impact (how much it moves the outcome that matters), Confidence (how sure you are in the first two numbers), and Effort (how much work it takes to ship). Multiply Reach by Impact by Confidence, divide by Effort, and you get one number you can rank against every other candidate on the list.
The Confidence factor is the part most teams skip, and it’s the part that matters most. It exists to stop a project from winning purely because someone was optimistic about its Reach or Impact rather than because those numbers were measured. A high-scoring project built on three guesses should score lower than a modest project built on three facts, and RICE is the mechanism that enforces that discipline.
RICE was built for software feature prioritisation, not for AI projects specifically, so applying it to an AI roadmap usually means swapping in dimensions that predict AI project success, like data readiness or governance readiness, while keeping the Confidence discount intact. The discount is the reusable part. The four original letters are just one way to fill it in.