Logistics · Business Process Automation · Custom Software Development
Automating Invoice Processing for a Regional Freight Company
68% Reduction in Processing Time
THE RESULTS
Outcomes That Changed How This Business Operates
Within 90 days of launch:
- Invoice processing time: from 4 days avg to ~18 hours avg (68% reduction)
- Data entry error rate: from 6.2% to 0.8% (87% reduction)
- Staff hours on manual entry: from 400 hrs/month to 80 hrs/month
- Invoices processed per day: from ~30 to ~90
The finance department went from being perpetually behind to being ahead of schedule for the first time in the company's history. The three staff members who were doing data entry are now focused on exception management, vendor relationships, and cash flow analysis, work that actually requires human judgment.
Annualized staff time savings: $210,000. Project cost recovered in under 13 months.
THE CHALLENGE
The Problem We Were Asked to Solve
A regional freight company with 150 employees was processing 2,000+ invoices per month entirely by hand. Three staff members in the finance department spent the majority of their time on data entry: pulling invoice PDFs from email, manually keying values into their ERP, matching them to purchase orders, flagging discrepancies, and routing for approval.
The process was slow (average 4-day processing time per invoice batch), error-prone (roughly 6% error rate on data entry), and fragile (entirely dependent on one senior employee who had been with the company for 19 years). When that employee took two weeks off, the department fell three weeks behind.
The company had looked at off-the-shelf accounts payable automation tools, but none integrated with their aging ERP system without a full ERP upgrade they weren't ready to do. They needed a solution that worked within their existing ecosystem.
The core problem: Manual invoice processing was costing them $210,000/year in staff time alone, before factoring in the cost of errors, late payment penalties, and the management overhead of a department constantly behind.
THE SOLUTION
How We Solved It
We began with a 3-day discovery workshop with the finance team. We mapped every step of their actual invoice processing workflow, including the informal steps that weren't in any documentation (like the sticky-note system the senior employee used to flag disputes). This phase was critical: two previous automation vendors had failed because they automated the documented process, not the real one.
Phase 1: Document Ingestion & Classification. We built a custom Python service using OCR and ML classification to extract data from invoice PDFs received via email. The system achieved 94% field extraction accuracy out of the box, reaching 98% after a 2-week training period with the team's real invoices.
Phase 2: ERP Integration. Rather than waiting for an ERP upgrade, we built a custom middleware layer that spoke to the existing ERP via its (underdocumented) API. This was the hardest part of the project. The ERP vendor's API documentation was three versions out of date. We reverse-engineered the actual API behavior and documented it as a by-product.
Phase 3: Workflow Automation. All extracted invoices flow into a custom approval workflow. Invoices that match POs automatically queue for one-click approval. Discrepancies route to the appropriate manager with context. Disputes flag to the senior AP specialist with all relevant information in one screen.
Phase 4: Exception Handling. We built careful exception handling because we knew the system wouldn't be perfect. Any invoice the system isn't confident about gets flagged for human review. The team sees exactly what the system extracted vs. what's on the document, and can correct it with one click. All corrections feed back into the model.
CLIENT TESTIMONIAL
What the Client Said
We'd been living with this problem for years. What surprised me most was how quickly they understood our actual process, not just what we told them, but what we actually did. The system went live in week 10 and we haven't looked back.
TECH STACK
Technologies Used
The tools we chose for this project, and why they fit the job.
Ready to Build Something Like This?
Tell us about your project. We'll tell you what it would take.
24-hour response guarantee · Your information stays private · Or email us at hello@vandeni.com