How a $10B+ SaaS company with 12 global subsidiaries eliminated 95% of manual AP labor using GenAI Document Intelligence — without ripping out NetSuite or Coupa.
By 2024, this $10B+ SaaS company had scaled its business to 12 global subsidiaries across North America, EMEA, and APAC. With growth came operational complexity that their AP team was not designed to handle.
The finance team was processing over 50,000 invoices every month. Each subsidiary had unique invoice formats — PDFs, scanned documents, EDI files, even handwritten forms from some APAC vendors. Languages ranged from English and German to Japanese and Mandarin.
Their existing workflow relied on a combination of NetSuite manually entered data and Coupa for purchase order management. The handoff between the two systems was fully manual — an AP clerk would download invoices from email, attempt to match them against open POs in Coupa, then manually key the validated data into NetSuite. When discrepancies appeared (which was 3.2% of the time), the ticket would sit unresolved for an average of 4 days while the clerk emailed the vendor or procurement team for clarification.
The VP of Finance Operations described it: "We had 2,400 hours of manual labor every month, just on invoice processing. That's 15 full-time employees doing nothing but data entry. And we still had a 3.2% error rate."
Single subsidiary deployment with the highest invoice volume. Flowtaris AI connected directly to NetSuite via REST APIs and Coupa via webhook integrations. The GenAI extraction model was deployed with zero template configuration — it processed all 15 existing invoice formats on day one with 96% field accuracy.
Expanded to 4 EMEA subsidiaries. Multi-language processing validated across German, French, and Spanish invoice variants. Autonomous Workflow Engine activated for exception routing — discrepancies automatically pinged the responsible buyer in Slack with a structured resolution request rather than leaving an unresolved ticket in the queue.
Full rollout to all 12 subsidiaries including APAC markets with Japanese and Mandarin invoice processing. 3-way matching (PO, goods receipt, invoice) automated end-to-end. NetSuite auto-posting enabled for all invoices below $50,000 without human review. Invoices above threshold routed for digital sign-off only.
Vision-Language Model (VLM) trained on financial documents processes 15+ formats across 8 languages without layout templates. Field-level confidence scoring flags uncertain extractions before ERP posting.
Discrepancies trigger structured Slack messages to the responsible buyer with pre-populated resolution options. 73% of exceptions resolved without an AP clerk touching the ticket.
Real-time bidirectional sync between NetSuite and Coupa via REST and webhook. PO status changes in Coupa immediately reflected in NetSuite invoice validation logic.
Live invoice pipeline visibility for CFO and VP Finance — current queue depth, exceptions by subsidiary, daily throughput, and STP rate. No manual reporting.
At the 12-month mark, the CFO office commissioned a formal financial audit of the AP automation deployment. The numbers were verified against NetSuite transaction logs and Coupa approval records.
The most significant outcome was the redeployment of 18 FTEs from data entry to strategic finance work. These team members were reassigned to vendor contract negotiations, early payment discount programs, and treasury optimization — roles that directly contribute to margin improvement.
The $4.5M annual savings figure breaks down as: $3.1M in labor cost reduction (direct and burdened), $900K in early payment discount capture (previously impossible at 4-day cycle times), and $500K in error correction elimination (reversal processing, late fees, vendor dispute credits).
"We went from drowning in manual invoice processing to near-full automation in 90 days. The ROI was visible in month 2 — not at the end of a 2-year transformation. What surprised us most was that we didn't have to rebuild our ERP or change our processes. Flowtaris AI fit into exactly how we already work."
