What belongs on an AI roadmap?
Put initiatives in an order that reflects learning and dependencies. A first workflow may need a data connection, a named owner, an approval path, and a measurable baseline before any agent is built. A good roadmap exposes the conditions for progress, not just target dates. AI readiness tests whether a workflow can start, and AI use case prioritization explains why it comes before another candidate.
Include technical and organizational work together. Data access, user training, AI governance, evaluation, and support are dependencies even if they do not look like product features.
How does it stay useful after the first pilot?
Attach a decision point to every milestone. The team should know what evidence supports moving from prototype to limited use, from limited use to production, or from production to wider adoption. A roadmap earns its place by making it acceptable to stop weak work early.
The AI strategy provides the direction. An AI business case defines the value and cost to check. The AI operating model assigns who acts on the evidence when the plan needs to change.