The Challenge
At an Enterprise finance operation, invoice processing looked automated on paper — OCR intake, ERP posting rules, approval workflows — yet AP analysts still spent most of their week chasing exceptions through email and chat. A missing PO, a GST mismatch, a duplicate vendor record, or a GRN that never landed in the system all became the same undifferentiated pile: "stuck invoices."
Nobody owned the boundary between procurement, warehouse, and finance. Vendors called AP; AP pinged buyers; buyers forwarded threads; GRN status lived in a spreadsheet someone updated on Tuesdays. Month-end close triggered a manual sweep that cleared symptoms without fixing why exceptions kept returning. Leadership saw a green "invoices processed" count while vendor escalation volume told a different story.
The organisation did not need another OCR vendor. It needed exception automation — every failure mode named, routed to an owner, tracked with an SLA, and visible in a weekly pack finance could run without assembling slides the night before.
The Approach
We mapped the AP path end to end and listed the top failure modes at each handoff: intake validation, three-way match, tax compliance, approval routing, and ERP posting. Each mode became a category with a single owner — not "finance" as a blob, but a named role on procurement, warehouse, or AP as appropriate.
- Category taxonomy — PO missing, quantity mismatch, price variance, duplicate invoice, invalid GST, GRN pending, approval timeout — no generic "hold" bucket.
- Validation at intake — Block or flag before OCR spend when mandatory fields or vendor master mismatches are obvious.
- Exception queues — Each category lands in a queue with age, vendor, amount, and assigned owner visible in ERP and a daily ops view.
- SLAs by category — Procurement-owned queues target faster turnaround than master-data fixes; breaches escalate in the weekly review.
- Weekly exception pack — Open count, ageing trend, SLA breaches, and repeat vendors — prepared before the ops call, not during it.
How It Worked
When an invoice failed match rules, the system assigned a category automatically where rules allowed — PO not found, quantity over receipt, price above tolerance — and routed to the correct queue. Ambiguous cases went to an AP triage queue with a same-day SLA to classify, so nothing sat unowned.
Procurement saw only PO and buyer-action items; warehouse saw GRN gaps; AP owned tax and duplicate checks. Vendors received structured status from a template tied to category — not free-text replies that contradicted the ERP. Approvers got reminders through the workflow engine, not parallel email chains.
The weekly ops review walked oldest exceptions first, confirmed owners in the room or on the call, and logged decisions before hang-up. When a category repeated — the same vendor GST issue three weeks running — the fix shifted to master data governance, not another manual touch on each invoice.
Results
Invoice cycles became leaner and more predictable. Analysts spent less time chasing threads and more time clearing owned queues. Vendor escalation email volume dropped because status was visible and categorized. Month-end close no longer depended on a heroic pre-close sweep; open exceptions were a managed backlog with trend lines, not a surprise.
Finance leadership opened the weekly pack as the source of truth for AP health — queue depth, SLA performance, and repeat failure modes — instead of debating whether "processed" meant "paid" or "sitting in someone's inbox." Cross-functional friction moved from blame in chat to named categories with owners, which made hiring and training easier too.
The improved process reduced rework from re-keying, duplicate payments caught late, and approval loops that restarted because context lived in email. Savings showed up as fewer hours on exception firefighting and smoother vendor relationships — not a vanity metric on a slide.
Takeaway
AP automation without exception design only speeds up the happy path while hiding pain in inboxes. Name every failure mode, route it to an owner with an SLA, review queues weekly with decisions logged, and align procurement, warehouse, and finance at the handoffs. That is how invoice processing becomes trustworthy end to end.