W. R. Berkley Reports AI Underwriting Gains

Berkley said the underwriting workbench digitizes activity from intake through quote. Rather than build its own large language model, the company plans to use available tools, add its own processes, and test approaches across its roughly 60 operating businesses before consolidating around the strongest options.

Claims are the second area highlighted on the call. Berkley said the insurer is moving toward straight-through processing where appropriate and noted that approximately half of its claims settle for $5,000 or less. That figure helps explain where automation could be useful, but management did not report a quantified claims-efficiency gain.

The discussion came alongside broader technology and data investment. Management said it believes the company can keep its expense ratio at 30% or better while continuing those investments, but it did not isolate AI's contribution to that target.

The clarification makes the metric more useful: it describes increased submission throughput, not a generic productivity claim. It still should not be treated as a company-wide benchmark or proof that underwriting quality improved. Processing more business can expand capacity and shorten queues, but the economic value depends on which submissions advance, how quickly quotes are produced, and whether portfolio outcomes remain sound.

A stronger follow-up would report deployment coverage, quote turnaround time, bind rates, underwriter capacity, and loss performance. Until W. R. Berkley publishes that detail, the result is best understood as an early operational throughput gain rather than a validated industry benchmark.