AI platform for claims assessment in insurance

Starting point

An insurer processes thousands of benefit claims annually. Each claim file contains medical findings, assessments, and physician statements — typically running to double-digit page counts per case. All analysis was performed manually. The existing line-of-business system had been in production for years but was never designed for AI-assisted analysis. The challenge: integrating AI capabilities into a mature system without disrupting live operations — with an architecture that holds up under DORA and the EU AI Act.

What we did

We built an AI platform for claims assessment and developed two integration scenarios — from pragmatic entry point to full system integration — through to a board-level decision document. The platform processes medical documents, automatically classifies over one hundred document types, and extracts structured domain data. In parallel: compliance analysis across five regulatory frameworks, data protection assessment, and responsibility delineation between all parties involved. The decision document was prepared for a board-level committee. Team: 3 people over 12 weeks. Disciplines: product architecture, system integration, regulatory analysis.

Results

What we learned

The real challenge with AI in claims processing is not model quality — it is integration with existing systems. The best extraction results are worthless if the examiner cannot see and correct them in their familiar working environment. Human-in-the-loop is not a feature. It is an architectural decision.