Bring together a small, empowered team and identify one recurring problem that people can describe and measure clearly. Include someone who does the job, a budget owner who experiences the consequences and an executive sponsor who can help remove obstacles.
I would start by asking a few simple questions. What slows you down? Where do surprises cost us money? What are our “predictable surprises” – moments where information lives somewhere in our organization but the right person finds out too late? Suppliers should have a voice in this discussion, too, because changes to procurement processes affect their workload.
Then establish a baseline. If the problem is a slow approval process, measure how long it takes today and where the waiting happens. If it is inconsistent pricing, identify the relevant transactions and validate that you are comparing equivalent requirements and contexts.
That gives you a concrete basis for choosing an AI use case and judging whether or not it helps.
In the book, Len DeCandia’s First 100 days playbook emphasizes building this foundation of trust and measurable progress. An early project should be manageable enough to show results quickly, with a clear decision about how it can expand if it works.
The goal for the first month is to understand the problem, agree on the desired outcome and give the right people authority to act.