Automation Process Hunter — Find Work to Automate

Refund workflow automation discovery

Evidence and unknowns

Known from the request: staff read refund emails, check order data, compare requests with policy criteria, and issue refunds in a store platform.
Unknown: request volume, handling time, eligibility rules, exception and dispute rates, refund limits, fraud exposure, system capabilities/API access, audit logs, access controls, customer impact, and current measured outcomes. No ROI or time savings can be calculated.

Process map to validate

1. Receive and classify request.
2. Identify order and verify evidence.
3. Apply the current refund policy.
4. Resolve missing information or exceptions.
5. Authorized staff issue or decline refund.
6. Notify customer and record outcome. Confirm every step and owner against the actual process.

Candidate work

Safe pilot

Use a permissioned, representative historical sample with outcomes reviewed by policy owners. In shadow mode, compare suggested categorization and checklist results with staff decisions; do not issue refunds or send messages. Measure false eligibility, missed exceptions, reviewer time, disagreement rate, and customer-risk cases. Set thresholds and a stop condition before testing.

Next steps

Confirm process owner and policy version; collect privacy-safe samples and baseline measures; map access and integration evidence; classify exceptions; review controls with finance and support; then decide whether a bounded assisted pilot is justified.