01

Follow a real task from start to finish

Choose work your team repeats and observe a few real examples. Record where the input arrives, the systems involved, the judgment required, and what a finished result looks like. Include the exceptions; they often reveal the true scope.

02

Look for a reviewable result

Drafts, summaries, classifications, and prepared records are often easier starting points than actions that are hard to undo. Pick a result a person can inspect before it affects a customer or a business record.

03

Write down the baseline

Record how often the task happens, how long it takes, and the amount of rework it creates. Use a representative sample rather than your best or worst day. This is what you will compare the new workflow against.

04

Define a useful pilot

Name an owner, limit the input types, and decide what success means. A pilot needs a stop condition as well as a success condition. Start with a small set of real examples, review the outputs, and expand only when the team is comfortable.

Make it specific to your business.

Bring one recurring task. Let’s explore whether AI belongs in the workflow.

Find your starting point