What is an AI roadmap?

An AI roadmap is the order in which an organisation will put AI to work, laid out over time: which process first, which next, what each one needs before it can start, and who owns it. It turns an AI strategy, which says where AI pays off, into a plan that a budget cycle and a change process can follow.

Most roadmaps fail for the same reason: they list technologies or themes instead of processes with owners. A line that says "generative AI in customer service, Q2" cannot be budgeted, built or stopped. A line that says which process, which system, who owns it and what it saves can.

What goes on each line

A usable roadmap line answers six things: the process (handling purchase invoices, triaging tickets, forecasting demand), the system it lives in (AFAS, Exact, the WMS, TOPdesk), the owner inside the organisation, the cost and the expected saving with the assumption stated, the dependencies (data that must be cleaned first, an interface that does not exist yet, a policy decision), and the AI Act risk class. The quarter comes last, because it follows from the dependencies rather than from ambition.

Five questions that let a line survive a change advisory board

In an organisation with its own IT function every line eventually becomes a change. It goes through faster when it already answers what the board will ask: where does it run and who hosts it; which data does it read and write, and on what GDPR basis; what happens when it is wrong, and who notices; how is a new model or prompt version tested before release; and who carries it after go-live, your IT department, your ICT partner or the supplier that built it. A line that cannot answer these is not ready for its quarter.

Roadmap, strategy and audit: how they relate

The strategy says where AI pays off and which rules apply. The roadmap puts those candidates in order over time. The audit is the work that produces both, on your own data and processes. A roadmap without an audit behind it is a wish list; an audit without a roadmap is a report. The deliverable of our AI audit and strategy is exactly this: a ranked list with cost, saving and payback per candidate, the first build priced, in one to two weeks for €2,500.

How often to revise it

Every quarter, and whenever a line is built or dropped. Two things move a roadmap more than anything else: a line whose data turned out not to be reachable, which pushes everything that depended on it, and a first system that runs well, which makes the next line in the same system much less expensive. A roadmap that has not changed in a year is no longer being used.