Start with repetition
A useful starting point returns often enough to be observed. It has a recognisable beginning, a handoff, and an outcome a team already cares about.
Make the boundary visible
Do not begin with every task a business could automate. Define the specific work, the inputs it uses, and the point at which the responsible person needs to decide.
Agree what would count
A pilot is more useful when the team agrees in advance which evidence matters: time, response speed, approvals, exceptions, volume, or another concrete operating measure.
Worked example: prospect research
Suppose a sales owner repeatedly opens several sources to decide whether an account fits the team’s criteria. Bound the pilot at a research brief for an agreed batch. Name the approved sources, required fields, reviewer, and definition of an accepted brief. Sending outreach is a separate decision. This is an illustrative scope, not a customer case study.
A selection checklist
Compare candidate workflows on repetition, input availability, clarity of ownership, reversibility, and ability to measure the outcome. Prefer work with an observable review point. Defer a workflow if permission to use its inputs is unclear or nobody can decide whether the output is correct.
Write the acceptance rule first
Agree how many missing or unsupported fields make a brief unusable, what evidence each claim needs, and how review time will be recorded. A pilot that produces more drafts while increasing total review time has not yet earned expansion.
Put the guide into practice
Explore a lead research pilot, or explore the workflow pilot examples to find a scope your team can assess.