AI roadmap
An AI roadmap sets out which AI applications an organisation will build, in what order, and what each one requires in people, data and decisions. A usable roadmap names an owner and a date per step, so it functions as a plan rather than a wish list.
What separates a roadmap from a list
Most documents called an AI roadmap are lists of opportunities with an ambition attached. The difference is dependency and ownership. A roadmap says this step cannot start before that one is done, and this named person decides. Without those two, nothing sequences and nothing gets refused.
What belongs in it
- The order and the reason for it. Not just what comes first, but why: usually a dependency on data or on a decision, rarely on ambition
- An owner per step, by name, inside your own organisation
- What each step requires in people, data and decisions, not in software licences
- A stopping rule. Under what circumstance do we abandon this step, decided before you are emotionally invested
- A revision date, because a roadmap written today is wrong within a year
What does not belong in it
A technology stack. Choosing tools before you have decided which problem to solve is the most reliable way to end up with a licence nobody uses. The tool follows the use case, not the other way around.
Roadmap and readiness
A roadmap is only as good as the diagnosis under it. If nobody established where the organisation stands, the ordering is guesswork. The AI Readiness Scan exists to produce that ordering on evidence, and the prioritisation matrix is the instrument that does the ranking.