What should insurers automate with AI first?

Insurers should begin with high-value, lower-risk workflows where AI can accelerate work without independently making consequential customer, claims, or underwriting decisions. Good starting points include software development support, unit testing, document and knowledge retrieval, data-quality monitoring, internal reporting, and operational workflow assistance. Each use case should involve human oversight and demonstrate measurable value.

At Digital Insurance 2026, speakers indicated that AI is not simply a switch an insurer can "turn on." Syed Hussain, AVP of Information Technology at Berkshire Hathaway Specialty Insurance, described a pragmatic approach: start small in development and testing while building workforce capability and governance before broader deployment. "We're just taking it one step at a time," he said.

Production-ready AI requires continuous data-quality controls, because real operational data is less predictable than the clean data typically used in prototypes. Insurers should therefore treat early automation as a way to develop reliable controls as well as demonstrate value.

For CIOs, operations leaders, and innovation leaders, the practical first step is to create a prioritized AI-use-case backlog and score each candidate against five criteria: business value, customer or employee impact, data readiness, regulatory or operational risk, and the consequence of an incorrect output. Select one workflow with clear ownership and a human review point, establish baseline measures for quality, time, cost, and adoption, and use the pilot to strengthen data controls, change management, and governance before expanding into higher-stakes decisions.

Join Digital Insurance to compare low-risk, high-value AI starting points with peers and learn how leading carriers are building the governance and data foundations required to scale responsibly.