01Overview
Relay models a common situation: an operations team spending hours each week exporting data from several tools, reconciling it in spreadsheets, and producing reports and invoices by hand. The concept automates the routine end to end, with checks where human judgment is still needed.
02The challenge
The process involved four systems, two spreadsheets, and a lot of copy-and-paste. Errors were common, reports were late when someone was away, and nobody could be sure which spreadsheet was the latest version.
03Proposed solution
Scheduled workflows collect data from each system's API, validate and reconcile it in Python, and generate reports and draft invoices automatically. Exceptions — mismatched totals or missing fields — are flagged in Slack for a person to review, and every run is logged.
04Approach
Process mapping
Every step, input, and exception documented before automating anything.
Validation layer
Reconciliation rules catch mismatches and missing data before they reach a report or invoice.
Human-in-the-loop
Exceptions are routed to Slack for review rather than silently skipped or guessed.
Observability
Run logs, failure alerts, and a written runbook for the team.
05Results
As a concept project, there are no client results to report. If this were a live engagement, these are the measures we would agree on up front and report against:
- Hours of manual work per week
- Reporting errors caught before delivery
- On-time delivery of weekly reports
- Exceptions requiring human review