A mid-sized FinTech client was burning 40 person-hours a week on a Friday reconciliation cycle: matching payouts, exceptions, and ledger updates across three systems that didn't speak to each other. The work itself wasn't difficult — it was high-stakes copy-paste in a hurry.
The cost we kept finding
Almost everyone in operations could do the work, but only two senior people had the context to handle the exceptions. So the queue stacked behind them every week, and the rest of the team waited.
warning What we measured first
Average reconciliation cycle: 40 hours/week.
Exception rate: ~6%.
Median time-to-resolve a single exception: 26 minutes.
What we built
A single Python pipeline that pulls from all three systems on a schedule, runs the matching logic in one place, and emits a Friday-morning digest of exceptions only — not the whole job. The team now reviews exceptions before standup instead of chasing them after lunch.
Outcome 90 days in
Reconciliation cycle down to 10 hours/week (75% reduction). Processing volume up 120% because the team finally had headroom. The two senior people got their Fridays back.