Zurich launched a business challenge to validate the use of agentic AI in travel claims processing. BotmasterAI developed Claims-PRO as a Proof of Concept, successfully processing the entire test batch provided.
Test batch successfully processed
Automatic validation against policy
Transparent natural-language reasoning
Validated, enterprise-ready prototype
The Problem
Zurich Insurance launched a business challenge aimed at technology providers: how to automate travel claims processing using agentic AI.
The context: Zurich receives over 200,000 travel reimbursement requests every year globally. Every claim has to be manually validated against the policy terms, consuming valuable time and internal resources.
The brief: build a multi-agent system (MAS) capable of automatically validating non-medical travel claims against policy, operating quickly, accurately, and at scale.
The technical challenge: the system needed to process heterogeneous documentation, validate policy coverage, identify missing information, autonomously decide whether to reimburse, reject, or request further documents, and provide transparent, verifiable reasoning for every decision.
*This case study describes a Proof of Concept developed for a business challenge launched by Zurich Insurance. The Claims-PRO system was successfully tested on data provided by Zurich, but is not currently deployed in production.
The Solution
BotmasterAI took on the challenge and developed Claims-PRO, a specialized multi-agent system deployed on the enterprise-grade BotmasterAI platform.
Claims-PRO is built on a crew of specialized AI agents, each with distinct roles and reasoning capabilities expressed in natural language in real time. A Crew Manager agent oversees the claim analysis and assigns tasks to the specialized agents based on their expertise.
The system processes the supplied documentation, automatically validates it against the policy terms, identifies information gaps, and determines the outcome: reimburse, reject, or request additional documents. Every decision comes with transparent reasoning visible in real time.
Claims-PRO includes long-term memory: when new documents are submitted days after the initial processing, the agents incorporate the new information without reprocessing the previous steps, improving efficiency.
Human-in-the-loop is built in by design: in ambiguous cases, the system automatically generates email templates to request clarification from the customer, and waits for human validation before proceeding.
The prototype was developed in a matter of weeks and tested on the full batch of claims provided by Zurich for the challenge.
it works
n your context? 
Tell us about the operational problem you want to solve, we’ll analyze feasibility and ROI together, and recommend the agents best suited to your case.
