AI Agents for Automated Claims Management
Design and program management for 5 specialized AI agents automating customer inquiry processing at a statutory health insurance fund. UiPath Maestro with LLM integration, compliant-by-design with social data governance and EU AI Act classification from day one.
Show details
Background
A large statutory health insurance fund processes hundreds of thousands of customer inquiries manually each year. Processes do not scale, error rates are high, and the skilled labor shortage is intensifying. A board commitment fixed a PoC deadline.
Challenge
- Social data (SGB V) requires special governance -- AI cannot simply access customer data
- Three organizations (insurer, IT provider, AI implementation partner) must be coordinated
- Budget exceeded, capacity limited -- yet the board deadline must be met
- No AI expertise in the organization -- neither technical nor regulatory
My Contribution
Role: Program Manager and AI Architect -- the central steering authority between business, IT, and implementation partner.
- Designed an AI agent network: 5 specialized agents with defined capabilities, autonomy levels (semi-autonomous to autonomous), and data flows
- Established compliance-by-design: social data governance, GDPR conformity, and EU AI Act risk classification embedded into every architecture decision
- Sharpened PoC scope: from a vague "AI agents project" to a concrete MVP scope with 5 measurable capabilities and clear success criteria
- Steered implementation partner: validated effort estimates, specified demo requirements, built QA processes
- Resolved stakeholder conflicts: aligned differing expectations (board, business, IT, partner) into a shared roadmap
Technologies
UiPath Maestro, DocumentUnderstanding, LLM Integration, Agentic AI, RPA
Result
- PoC phase active and on track despite budget and capacity constraints
- 5 agent blueprints serving as the foundation for the insurer's entire AI strategy
- Compliance framework reusable as a template for future AI initiatives
- Governance structure across 3 organizations operational -- weekly check-ins with clear deliverables