Come back with three processes that give you grief
Prioritising valuable organisational problems keeps the intervention space wider than a portfolio framed around AI use cases.
When organisations ask me where they should start with AI, I increasingly avoid asking them to identify AI use cases. Instead, I ask something much less sophisticated: Come back with three business processes that give you grief. The word grief is deliberate. I want the things people already know are difficult. Orders go missing. Reports take days to assemble. Information is copied between spreadsheets. Customers wait while somebody finds an answer. Work sits in queues. Errors create rework. Everything depends upon one particular person knowing what to do. These are not AI opportunities.
They are signals that something about the work deserves attention. The next question is value. A process can be extremely irritating while being commercially insignificant. So ask whether improving it would affect something that matters: Revenue. Capacity. Cost. Lead time. Quality. Customer experience. Risk. Now we have changed the object being prioritised. Instead of a portfolio of AI use cases, we have a portfolio of valuable organisational problems. That distinction matters. An AI use case already contains a proposed solution. Once we describe a problem that way, we naturally begin asking how AI might solve it.
Start with the work and the intervention space remains much wider. We might stop the activity. Simplify it. Redesign it. Automate it conventionally. Assist somebody with AI. Delegate part of it to an agent. Or combine several of these. A Power Automate workflow is not an inferior AI agent. Removing an unnecessary activity is not a failure to innovate. The right intervention is the one the work needs. There is still a place for asking what new technology makes possible. If we are exploring new products, services or business models, technological capability can be a useful starting point. But if the objective is improving the business we already have, I would start somewhere else. Not: Where can we use AI? But:
Where does valuable work currently give us grief? That question tends to lead somewhere much more interesting.