Google and CoverGo warn insurers against fragmented AI adoption

Fragmented adoption of artificial intelligence (AI) could create new operational and governance risks for insurers, a new white paper from Google and insurance technology provider CoverGo has warned.

The report, ‘Insurance in the Agentic Era: From AI use cases to an agentic insurance core operating system’, examines how autonomous AI agents could reshape insurance distribution, underwriting, claims processing, and customer servicing.

The paper argues that insurers must move beyond using AI primarily to automate individual tasks and improve productivity.

Instead, the authors – Maxim Afanasyev of Google and Tomas Holub of CoverGo – propose an integrated operating model in which specialised AI agents can coordinate activities, access policy and claims information, and interact with existing insurance systems.

The approach could have implications for travel and health insurers, where claims assessment, medical documentation, policy verification, and customer assistance frequently involve information exchanged between multiple parties.

Risks of autonomous decision-making

The report identifies several risks associated with deploying multiple AI agents, including conflicting decisions, repeated automated processes, and systems acting beyond their authorised responsibilities.

For example, one AI agent might calculate a claims payment while another determines that the policy has expired, creating conflicting outcomes without an established process for resolving them.

The authors also warn that AI-generated applications and claims documentation could challenge traditional underwriting and verification processes, potentially increasing the difficulty of identifying inaccurate information or assessing risk.

To address these concerns, the report recommends stronger governance, clear decision-making boundaries, and human oversight for higher-risk activities.

These measures include mandatory human approval for significant claims decisions, confidence-based escalation, and audit trails documenting automated decisions.

Modernising insurance systems

Google and CoverGo also outline a five-stage approach to modernising insurance technology, beginning with an assessment of existing systems and progressing through system design, parallel development, phased data migration, and the retirement of legacy infrastructure.

The authors argue that insurers will need more accessible and consistent data to support coordinated AI operations, particularly where information is spread across policy documents, medical records, and claims systems.

The paper presents an integrated AI operating system as a potential future model for the industry, although it does not provide evidence of widespread adoption or independently measured improvements in claims outcomes.

For travel and international health insurers, the implications could extend beyond administrative efficiency. Better integration between claims systems, medical providers, and assistance companies could support faster information exchange, more consistent benefit verification, and improved coordination of complex cross-border cases.

However, decisions involving medical necessity, treatment authorisation, and emergency assistance would require particular care, given the potential consequences of inaccurate or conflicting automated assessments.