Insurers Strengthen Digital Foundations for AI-Native Operations
Writing in Insurance Innovation Reporter on June 3, 2026, Tom Scheel, a managing director at Kyndryl, argues that insurers pursuing AI-native operations must first rebuild their digital foundations before agentic AI can scale across core functions. The commentary contends that while many carriers have moved to cloud systems and digitized customer experiences, most remain early in adopting agentic AI, which orchestrates workflows across underwriting, claims, and service and demands resilient, low-latency infrastructure and integrated data. Scheel points to claims processing, where policy data, claims histories, and adjuster notes often sit in separate systems, warning that AI layered on fragmented records can confidently recommend approvals or flag fraud from an incomplete picture. He frames unified observability and governance built into the architecture, rather than added afterward, as prerequisites for moving from pilots to production.
What happened
In an Insurance Innovation Reporter commentary published June 3, 2026, Tom Scheel, a managing director at Kyndryl, argues that insurers moving toward AI-native operations must first redefine the digital core that supports them. He writes that many carriers have progressed on cloud migration and customer digitization but remain early in adopting agentic AI, which orchestrates workflows across underwriting, claims, and service and places demands on infrastructure that legacy environments were not designed to meet. Scheel cites claims processing, where policy data, claims histories, and adjuster notes often live in separate systems, warning that AI layered on fragmented records may recommend approvals or flag fraud from an incomplete view.
Editorial analysis
The argument reframes modernization as an infrastructure-and-resilience challenge rather than a software-replacement exercise, emphasizing hybrid cloud, unified observability, and governance built into the architecture instead of added after deployment. This is vendor commentary rather than independent reporting, but it aligns with a broadly observed pattern in enterprise AI: model performance tends to improve only after brittle data pipelines and unclear ownership boundaries are resolved. Useful signals of real progress include investments in real-time data platforms, integrated policy and claims data, and explicit governance and traceability for AI-driven workflows.
Key Points
- 1Kyndryl's Tom Scheel argues, in an Insurance Innovation Reporter commentary, that agentic AI in insurance needs resilient, low-latency infrastructure and integrated data before it can scale across underwriting and claims.
- 2Siloed policy, claims, and adjuster systems leave AI working from partial customer views, which can produce inaccurate approvals or fraud flags.
- 3Editorial analysis (generic industry): the piece reframes modernization from a software-replacement task to an infrastructure, observability, and governance challenge.
Scoring Rationale
This is a substantive but vendor-authored commentary on the infrastructure, data, and governance gaps insurers face before scaling agentic AI, useful to practitioners designing production systems in the vertical. It is opinion and analysis rather than original reporting or a product or research milestone, so it rates as a solid, niche industry item.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
