No forty-slide AI strategy. We take one process, make it measurably better and build from there. Three phases, each with its own deliverable, and you can stop after any of them.
Software on an unclear process turns into expensive custom work. So we start by measuring: where the time goes, which steps are routine and which need a person. Only then do we decide what AI may do and which software fits.
In two weeks we map one core process: close, forecast or reporting. We talk to the people doing the work and follow a real cycle.
We use AI to check and reconcile figures, summarize information and prepare data. The controller keeps the last step: review and approve.
Where Excel hits its limits, we implement a planning platform. Sometimes better-structured Excel with Power BI is the answer, sometimes UNA or Vena. The software has to fit the process, not the other way around.
The person who understands the process also builds the solution. No translation to another team.
A thirty-minute call every week with an open-items list. What is done is live. What is open has an owner.
For every phase we agree what we measure: days, hours, versions, differences. That number decides whether we continue.