A simple checklist before AI enters a workflow.
A practical set of questions for any team weighing an AI use case — designed to keep purpose, data, review, and accountability visible from the start.
Purpose
Write down the specific problem. If the use case cannot be described in plain language, it is probably too vague. A strong description identifies the user, the task, the expected benefit, and the quality standard.
Avoid adopting AI because it appears modern. Adopt it because it helps a defined workflow become clearer, faster, more consistent, or more useful without weakening responsibility.
Data and privacy
Identify what information the use case touches. Is it public, internal, confidential, personal, sensitive, or commercially important? The answer affects permissions, storage, access, and review.
Teams should also ask whether outputs create new records, and whether those records need to be retained, corrected, explained, or deleted.
Review and accountability
Assign an owner. Decide who reviews outputs, how errors are reported, and what happens when results are uncertain. A process without ownership can create risk even when the technology seems simple.
The checklist should be proportionate. The goal is enough structure to protect quality and trust, without making low-risk learning impossible.
Monitoring and escalation
Decide in advance how the use case will be watched over time, who can pause it, and how a person raises a concern. Monitoring turns a one-off decision into a process that can be trusted.
Good escalation is simple: people should know when to stop, ask for help, or switch to a manual route, without needing to interpret complex rules under pressure.