Quick Answer
AI invoice processing cuts invoice handling time by up to 93 percent, based on customer results published by ABBYY and SS&C Blue Prism. Certas Energy, the largest independent fuel distributor in the UK, went from up to 30 minutes per invoice to about 1 minute using ABBYY OCR plus a Blue Prism digital worker. Companies without automation average $12.88 and 17.4 days per invoice, according to Ardent Partners State of ePayables 2024.
- PAPCO cut manual AP labor by roughly 50 percent with ABBYY FlexiCapture. The energy solutions supplier handles more than 600,000 invoices a year from over 1,000 vendors, cut daily processing time by up to 6 hours, and moved invoice approval from an hour a day to a couple of minutes.
- Structured e-invoicing removes OCR, not the approval work. Italy routes every domestic B2B and B2G invoice through the Sistema di Interscambio (SDI) as structured XML, but matching an incoming e-invoice against a purchase order or delivery note and routing exceptions to the right approver stay manual at most companies.
The Invoice Processing Bottleneck
Companies without automated accounts payable processes spend an average of $12.88 to process a single invoice and take 17.4 days to get it approved, according to Ardent Partners' State of ePayables research. Best-in-class AP teams save more than $10 per invoice and cut roughly two weeks off that timeline. That gap between average and best-in-class is most of the business case for intelligent document processing (IDP), and it's why the category has moved from a back-office experiment to a board-level line item.
Invoice Automation, By the Numbers
Two named, published customer results are below. Neither is hypothetical, and neither comes from us. They're what ABBYY and Blue Prism have put their own names behind.
The Hidden Costs Behind the Sticker Price
The $12.88-per-invoice average above only counts direct processing labor. Duplicate payments, missed early-payment discounts, and staff hours spent chasing down exceptions by hand all sit on top of it. Those costs show up in the AP team's time, not on an invoice.
AI Document Processing: How It Works
Modern IDP systems combine multiple AI technologies for end-to-end automation:
OCR + Computer Vision
Extracts text from any format - scans, PDFs, photos, emails
NLP Understanding
Interprets context, not just characters - understands "total due"
Machine Learning
Learns from corrections, improves accuracy over time
Validation Rules
Cross-checks against POs, contracts, and master data
Case Study #1: Certas Energy — 93% Faster With OCR and RPA
Certas Energy, the UK's largest independent fuel distributor, has around a thousand suppliers and an accounts payable team that used to key in data from hundreds of thousands of invoices a year by hand. Fuel volumes shift slightly in transit, so even the matching logic had to tolerate small variances between the goods-received note and the invoice. That's not the kind of judgment call you automate carelessly.
Certas paired ABBYY's OCR with an SS&C Blue Prism digital worker (internally nicknamed "Sir Lancelot") that logs into the ERP, matches invoiced items against the goods-received note, tolerates the expected in-transit variance, and completes the reconciliation.
Certas Energy Results
| Invoice Processing Speed | 93% faster |
| Per-Invoice Time | Up to 30 min → 1 min |
| Turnaround | Same day received |
| Annual Volume | Hundreds of thousands of invoices |
The AP staff who used to spend their days keying in data now spend it on the invoices the digital worker flags as genuinely ambiguous. For a business handling variable fuel quantities, that's exactly where a person should be spending judgment.
Case Study #2: PAPCO — From an Hour a Day to a Couple of Minutes
PAPCO, an energy solutions supplier in the eastern United States, processes more than 600,000 invoices a year from over 1,000 vendors. Its AP staff were, in ABBYY's own description, working overtime and weekends just to keep up.
PAPCO Results
| Daily Processing Time | Cut by up to 6 hours/day |
| Labor Reduction | ~50% |
| Invoice Approval Time | 1 hour/day → a couple of minutes |
| Annual Savings | $30,000 |
| Annual Volume | 600,000+ invoices, 1,000+ vendors |
PAPCO scans incoming paper through network copiers, and ABBYY FlexiCapture extracts, classifies, and routes the data to the right AP staff for verification. The team went from working weekends to closing out invoices inside a normal week, and PAPCO has since extended the same approach to freight and fuel invoices.
Italian SME Context: Fatturazione Elettronica
Italy's mandatory electronic invoicing system routes every domestic B2B and B2G invoice through the Sistema di Interscambio (SDI) in structured XML format. That structure is exactly the kind of input intelligent document processing handles well, since the data is already machine-readable before OCR even enters the picture.
Where AI still adds value for Italian SMEs is on the two sides SDI doesn't automate by itself: matching an incoming e-invoice against a purchase order or delivery note, and routing exceptions to the right approver. Those steps stay manual at most companies regardless of the mandate, because SDI standardizes the invoice format, not the approval workflow built around it.
Platform Comparison: Document Processing AI
| Platform | Best For | Key Strength |
|---|---|---|
| ABBYY Vantage | Complex documents | Strong OCR on complex layouts |
| Basware | Enterprise AP | End-to-end AP, supplier network |
| Kofax | High volume | RPA integration, workflow automation |
| Custom AI Solution | Unique formats | Tailored to specific document types |
How to Build the Business Case
Start from your own numbers, not a sample calculation: invoices per year, minutes of handling per invoice, the share of invoices that hit an exception, and the early-payment discounts you miss today. The Ardent Partners averages above describe the gap between manual and automated AP teams in general; they are a benchmark, not a forecast for yours. What the build costs depends mostly on how many document types and suppliers it must handle and how deep the ERP integration goes.
Where Invoice Automation Fits in a Broader Finance Stack
Invoice processing is rarely the only finance workflow worth automating. If you're evaluating this alongside other back-office work, our finance and accounting automation guide covers the wider toolset, and the workflow automation guide for small businesses is a good next stop if invoice processing is your first automation project rather than your fifth.
Before signing anything, run the numbers with our AI ROI calculator against your own invoice volume and current cost per invoice. The Ardent Partners averages above are a starting point, not your number.
Implementation Best Practices
Ready to Eliminate Manual Invoice Processing?
Get a free assessment of your document processing opportunity. We analyze your invoice volume, current costs, and recommend the optimal AI solution for your business.
Request Free Assessment →Key Takeaways
- ✅ 93% faster processing at Certas Energy, verified via SS&C Blue Prism
- ✅ Roughly 50% less manual labor at PAPCO, verified via ABBYY
- ✅ Industry average without automation: $12.88 and 17.4 days per invoice (Ardent Partners)
- ✅ Structured e-invoicing like Italy's SDI removes the OCR step but not the matching and approval work
Sources: SS&C Blue Prism: Certas Processes Invoices 93% Faster, ABBYY: Automated Invoice Processing Case Study (PAPCO), Bottomline, citing Ardent Partners' State of ePayables 2024
Key statistics (2025)
Further reading
Frequently asked questions
AI Solutions6 min2025-12-03
Share this article
