Start with the work, not the tool
An automation request often starts with a solution: connect these systems or let AI prepare the report. That is a possible direction, but not yet a shared understanding of the problem. First ask where work gets stuck, who has to resolve it and what should happen next. If people describe different problems, that difference is worth resolving before anything is built.
Choose one recurring situation. At a shift handover, for example, an unfinished task might have no named owner. The improvement needed may be a clear transfer of responsibility, not another dashboard. Describe the current steps simply enough for the person doing the work to recognise them.
Make better observable
Words such as faster and easier need a more specific description beside them. Does better mean less retyping, fewer rounds of clarification or finding the next responsible person without asking around? Agree on one main outcome and the things that must not get worse. Faster processing is not a gain if correcting mistakes quietly becomes someone else’s job.
Gather a small, appropriate baseline and use the same method during the trial. Record what was observed and what remains an assumption. A small trial can help choose the next step without demonstrating a wider business impact.
Agree on the boundaries too
Decide who checks the result, when to stop and how to return to the previous way of working. If the solution makes suggestions, distinguish them from approved decisions. Choose a bounded step with a checkable result for the first trial. Finish the conversation with a short agreement about the problem, intended change, responsibility and evaluation. Choosing a tool becomes easier once those decisions are clear.
A useful next step
Discuss a scoped engagement