What are we trying to improve?
Identify a specific problem and decide what a useful result would look like. Be clear about where AI may support work and where human judgment remains necessary.
What information is involved?
Understand the data a proposed use case would access, where it would go and who is responsible for approving its use. Review the organization’s existing policies before proceeding.
Who remains accountable?
Agree who owns the use case, who checks outputs and who can pause or change the process. Discuss how issues will be raised and how performance will be reviewed.
How will we test and improve?
Start with a defined scope and clear review points. Ask how the team will recognize an unsuitable result and what evidence will help decide whether to expand the use case.