Undisclosed Top-10 Commercial Insurer

A Fortune 500 commercial insurer ranked among the top 10 in the U.S. had built its accounts payable operations on a foundation of siloed legacy systems that could not exchange data automatically. Staff spent significant time on manual re-keying to bridge the gaps between platforms. Their existing robotic process automation (RPA) layer — which works well on structured, predictable inputs — broke down when confronted with the reality of vendor invoices: handwritten notes, rubber stamps, fax artifacts, and wildly inconsistent layouts ranging from single-page handwritten slips from small vendors to multi-page formatted documents from large corporate suppliers. The result was chronic SLA misses, a growing backlog of manual review queues, and measurable customer frustration.

The insurer augmented its existing Blue Prism RPA platform with Hyperscience's intelligent document processing (IDP), which applies proprietary machine learning to classification and data extraction tasks that rule-based automation cannot handle. Rather than replacing Blue Prism, the integration made the two systems complementary: Blue Prism's digital workers route incoming accounts payable invoices to Hyperscience, which classifies each document, extracts the relevant fields — even from messy handwriting, stamps, and degraded fax-quality images — and returns clean structured data for downstream processing. Critically, the ML models are self-adapting; when vendors change their invoice layouts, data center staff can retrain the system themselves without waiting on IT or development resources, embedding flexibility directly into the operational workflow.

The implementation delivered a material reduction in manual invoice processing volume, allowing staff previously assigned to data entry to be redeployed to higher-value work. SLA compliance improved, directly addressing the customer satisfaction issues that had prompted the initiative. Beyond throughput gains, the project produced two notable structural improvements:

• Auditability: the internal audit team can now trace every extracted data point back to its source document — a capability that was previously unavailable.
• Operational self-sufficiency: data center employees can adapt the system to new vendor form layouts independently, eliminating a recurring IT bottleneck.
• Data accessibility: information previously trapped in unstructured documents became available for downstream analysis for the first time.

• RPA alone cannot handle unstructured documents; pairing it with an IDP layer that uses machine learning fills the gap without replacing existing automation investments.
• Designing for business-user self-service — allowing non-technical staff to update document models — is as important as the initial accuracy gains; it determines long-term adaptability.
• Full auditability of extracted data is a compliance asset, not just an operational nicety, and should be treated as a first-class requirement during vendor selection.
• Winning internal audit and compliance teams over early can accelerate enterprise adoption of document automation.