No 3-year roadmaps.

I keep seeing AI and digital transformation in the same job title lately. It made me think about which habits carry over.

The one that doesn’t is the three-year roadmap.

Digital transformation could run on one. The technology was mature and the payoff was predictable. If you were moving to the cloud or a new ERP, you knew what the software could do and roughly what it would save.

AI isn’t there yet. Models get better and cheaper every few months. Vendors change their pricing, get bought, or ship the feature you were halfway through building. In Australia, the privacy rules on automated decisions change in December.

So I plan at two speeds.

The foundations get planned in years. Data you can trust and find. Guardrails on what AI can touch, and who checks it. One cloud platform to store data and to build, deploy and monitor models. An approved AI assistant for everyone, because otherwise they’ll use their own. And people who know how to use all of it. Every use case runs on these, whichever ones win.

The use cases get planned by the quarter. A short list of workflows, ranked on value, feasibility and risk, each big enough for the CFO to notice. Measures are agreed before anyone builds. Every quarter, the results reset the ranking. Some get more money. Some get stopped. Most never start, and that’s on purpose.

Plan everything in years and you lock in stale use cases. Plan everything in quarters and you get a pile of pilots.

Commit to the foundations. Let the evidence move the use cases.