The most common request we get is “we need an AI strategy”. That is rarely what is missing. What is missing is a list of work done by hand, how often, and by whom.
Make that list first. Then pick the process that is the most boring, the most frequent and the least judgement-based. It can almost always be automated, and the gain is measurable before you have spent a single meeting on strategy.
Measure before, not after
Time the process as it is now. Without that number nobody can say whether the automation was worth it, and the next project becomes a discussion about feelings.