Finance teams at large enterprises have a dirty little secret: every year, millions of supplier invoices stall in “exception purgatory,” waiting for humans to fix bad data, hunt for missing POs, or re‑key tax codes. One global food‑and‑agriculture leader (we’ll keep the name private) knows this pain all too well: 

• OCR accuracy is stuck at ≈ 20 %, four out of five invoices need manual correction 

• Only 50 % of invoices auto‑match in OpenText VIM, leaving the rest in exception status 
• An exception can take two hours and bounce across six systems before it’s cleared
  

Multiply those numbers by tens of thousands of invoices every month and you get spiraling labor costs, frustrated suppliers, and finance teams too busy firefighting to focus on higher‑value work. 

Why Yesterday’s Tools Hit a Ceiling 

Traditional OCR engines weren’t designed for today’s dizzying variety of invoice layouts, nor for the nuanced decisions buried in exception handling. Rules‑based workflows help, but once data quality degrades, they simply route the same problems faster. 

The business objectives are aggressive, but achievable, targets for any new solution: 

• 70 %+ OCR accuracy 

• 80 %+ straight‑through processing 

• 30 % fewer exceptions 
   

The Big Idea: AI Exception‑Handling Agents 

Qellus proposed an AI‑first approach that lives inside the customer’s existing SAP + OpenText stack, powered by SAP Business Technology Platform (BTP). Instead of  labor intensive intervention six specialized agents collaborate to diagnose and fix each exception in seconds: 

These roles are straight from SAP’s case study showcasing the power of AI Agents. 

Proof That It Works: SAP’s AB InBev Story 

If the concept feels futuristic, consider AB InBev’s SAP case study. In Brazil, the brewer blended process mining with AI exception agents and delivered: 

• 97 % discrepancy resolution 

• Invoice cycle time slashed from 2hours to 2 minutes 

• Automated PO validation and even fraud detection 
   

That success wasn’t led by Qellus, but it shows what’s possible when you combine SAP BTP, Signavio insights, and AI‑driven agents. 

Where We’re Headed Next 

First up: quick‑win sprints (OCR, matching, exception triage) that prove value in weeks, not quarters. Tailored agentic AI solutions for high volume exceptions that prevent straight through processing. 
 
And the best part?  Because everything rides atop SAP BTP, the project leverages already a familiar technology stack, no rip‑and‑replace, no new data silos. 

Three Questions to Ask Your Own AP Team 

1. What share of invoices still hit human touchpoints? 
2. How many systems does an exception travel through today? 
3. If AI could free 30 % of your team’s time, where would you redeploy it? 
   

If the answers make you uneasy, it might be time to explore AI agents inside SAP VIM. 

Ready for an Exception‑Handling Assessment? 

Qellus helps SAP customers modernize invoice‑to‑pay without blowing up what already works. Let’s spend 30 minutes spotting your biggest manual drains and sketch out a pilot you can stand up before year‑end.