Problem
The same information was being retyped between an inbox, a spreadsheet, and a storage folder. A missed step was only noticed downstream, when somebody went looking for a record or an attachment that had never been filed.
Our role
Design and build of the flow itself, including the failure handling.
Approach
The flow is a sequence of webhook- and API-driven steps: an incoming message is received, the details are extracted, a record is written, the attachment is filed, and the right person is notified. Each step is idempotent, so a retry does not create a second record for the same message.
Failures are surfaced rather than swallowed. If a step cannot complete, the record keeps the state it reached and the failure is reported, which means the recovery path is a specific step rather than a manual re-entry of the whole item. The HTTP calls and database writes are separated so the notification can fail without losing the record — the same ordering used in the contact pipeline on this site.
Outcome
The flow runs end to end, and failures are visible instead of silently skipped. Time saved has not been measured, so no figure is claimed here.
Notes
Still in progress. The specific implementation is being generalised before it is published, and this entry will stay open until that write-up is finished.