Choosing Your Starting Workflow

If you read your industry chapter and recognized more than one workflow, that's normal — most operations have several. The mistake to avoid is trying to fix all of them at once.

Three filters for picking a starting point

Volume. Pick a workflow that runs often enough to generate a measurable signal within weeks, not quarters. A process that happens twice a month won't tell you much about performance for a long time, however painful each instance is.

Cost of failure. Weigh both the cost of the workflow breaking today (backlog, errors, staff turnover in that role) and the cost of getting automation wrong (regulatory exposure, customer trust). The best starting workflows have real cost today and a bounded, well-understood risk if something needs correcting.

Data and system access. A workflow where the relevant information already lives in a system you can connect to — even imperfectly — will move faster than one that depends on paper records or tribal knowledge that hasn't been documented anywhere.

What a good first workflow looks like

In our experience, the strongest first engagements share a shape: a well-defined process with a clear start and end point, a backlog or error rate that's visible on a report someone already looks at, and a team that's motivated to fix it because they feel the pain directly — not because leadership mandated it. It does not need to be your biggest problem. It needs to be a problem you can solve completely, measure honestly, and point to as proof before expanding further.

From one workflow to an AI-native operation

The pattern we see repeatedly: the first Digital FTE proves the model on one workflow, the team that owns it becomes the strongest internal advocate for the next one, and within a few cycles the organization has a working, learned methodology for evaluating where AI belongs — rather than a single pilot that never scaled.

That's the shift Part One described as moving from "tool-first" to "workflow-first" thinking. It compounds.