Success Story
Agentic AI invoice processing has caught $165M in duplicate or incorrect invoices before a single payment went out at a major construction and property investment company. The same solution has processed more than $475M payments automatically, and provided a complete, invoice-to-ERP audit trail. It's also cut manual effort per invoice from 30 minutes to under 5 minutes, an 83% reduction in processing time.
Every invoice from builders, subcontractors, and property operations arrived in a shared inbox, printed automatically, and landed in a stack by the office printer. A team sorted each one into folders by hand, ready for review. From there, the accounts payable team keyed the data into the enterprise resource planning (ERP) system and matched it against one of more than 2,000 ledger codes.
The process broke down in more than one place:
Agentic AI invoice processing uses artificial intelligence that plans and carries out multi-step tasks on its own, such as classifying an invoice, extracting its data, and choosing the correct ledger code, without needing a person to prompt every step. People step in only for the decisions that need judgment, like approving a payment or resolving a genuinely ambiguous line item. The result is faster processing, cleaner financial data, and a complete audit trail on every invoice.
Ashling designed and built an intelligent accounts payable solution on UiPath, powered by agentic AI.
Every invoice arrives by email, often with 10 or more supporting documents, such as subcontractor statements, insurance certificates, and timesheets, across 12 different document types. UiPath Maestro's system of action orchestrates the whole process from that first e-mail trigger, initiating UiPath IXP to read and classify each document and extract the data required to substantiate invoice claims.
Builders and property managers are notified within seconds and can review, approve, reject, or approve a partial payment from their own device, on site, instead of waiting for a Friday office visit.
That info is passed back to the finance team where generative AI assist in intelligently recommending General Ledger Coding before it's entered into the ERP. Every supporting document and comment is attached to that record, giving the team a complete, traceable audit trail. A live dashboard gives leaders real-time visibility into exactly where every payment stands.
Time per invoice has dropped from about 30 minutes to under 5, an 83% cut, across more than 10,000 high-value invoices and roughly 2,500 invoices a month, averaging about $50,000 each. The solution has processed more than $475M in approved payments, and its quality checks have caught and rejected around 3,000 duplicate or incorrect invoices worth roughly $165M, before they ever reached payment.
The solution's real-time visibility also gives builders and property managers a clearer view of budget against spend, building in the accountability that was missing before. With payments processed and paid closer to their due date, the company holds onto capital longer and manages cash flow with more control.
The next phase adds a confidence check to the accounts payable team's first review. High-confidence invoices will skip that initial step entirely and route straight to the builder for approval, reserving the team's attention for genuinely complex cases, like invoices carrying multiple purchase orders with different values. The goal is to keep expanding what agentic AI invoice processing handles on its own, so people spend their time only where judgment is actually needed.
