Run a customer-data check before automation
A grounded review for deciding which customer signals are needed, allowed, and worth retaining in a retail workflow.
Source: Rookmint-authored AI safety guide
Write the purpose in one sentence
Before selecting a signal, state the decision it should improve and who will use the result. If the purpose is vague, the workflow will tend to collect more data than the decision requires.
Use only what the decision needs
Separate essential inputs from convenient extras. Check whether the team already has an approved source and whether the proposed use matches the customer and internal expectations for that data.
- Remove fields that do not change the next action.
- Limit exports and temporary working files.
- Document the reviewer and the allowed purpose.
Review the output, not just the input
A workflow can use permitted signals and still produce a result that is misleading or hard to explain. Sample outputs with an operator, record corrections, and keep the current process available while the team learns.
Keep going
Turn the guidance into a next step
Keep the first move narrow, reviewable, and owned by someone who can pause it.