AI prioritisation matrix
An AI prioritisation matrix ranks possible AI applications against each other so an organisation can decide what to do first. It scores each opportunity on value and on feasibility, where feasibility covers the organisation as much as the technology: data, ownership and capacity to maintain.
The problem it solves
After an inventory, most organisations have between ten and thirty ideas and no way to choose between them. Everything sounds plausible and the loudest sponsor usually wins. A matrix removes the argument from the room by making the criteria explicit before the scoring starts.
The two axes
Value. What it is worth in your own numbers: money, errors avoided, or work that becomes possible. Not strategic importance, which is unfalsifiable.
Feasibility. This is where most matrices are too narrow. Technical feasibility is rarely the binding constraint in a mid-sized company. The real questions are whether the data is consistent enough, whether one named person owns the process, and whether anyone will have time to maintain it after launch.
Scoring it honestly
- Score with the people who do the work, not only with management, because feasibility is knowledge the floor has and the boardroom does not
- Use a narrow scale, one to three per axis. Finer scales invite false precision and endless debate
- Write down the reason for each score. Six months later the reason matters more than the number
- Force a ranking. If everything scores high, the criteria were too kind
What to do with the quadrants
High value and high feasibility is where you start, and there is usually less in that corner than people expect. High value and low feasibility is not a rejection, it is a list of what would have to change first, and it is often the most useful output of the whole exercise. Low value goes away regardless of how easy it is.
The matrix feeds the AI roadmap, and the value column is where AI ROI gets calculated.