- 44% of underwriters say senior judgment leaving the industry, without being passed on, is their top fear, ahead of AI itself
- Coaching and knowledge-transfer technology gets just 8% of firms’ investment over the next 12 to 18 months, the lowest of any category surveyed
- Only 15% say their firm has found a way to capture what its best underwriters know
- Two in five say that judgment today is captured poorly or only lives in a colleague’s head
- Notably, worry that AI will replace the underwriter’s job entirely keeps falling: 18% called it an urgent issue in 2024, 11% in 2025, just 5% today.
AI is helping commercial insurance underwriters reduce administrative work, but fewer users report improvements in decision quality, according to the Underwriting Edge 2026 report from hyperexponential. The study surveyed 350 senior commercial property and casualty underwriters across the US and UK in June.
Independent research firm Coleman Parkes conducted the survey. Hyperexponential, or hx, sells AI underwriting software to insurers, giving the company a direct commercial interest in the market examined by the report.
Among underwriters already using AI, 51% said its greatest contribution was reducing time spent on manual administration. Only 21% identified improved decision quality as the technology’s main benefit.
The findings suggest current AI deployment is concentrated heavily on workflow efficiency. Underwriters, however, identified several information and context problems that affect the quality of decisions made during the submission process.
Inconsistent submission data was cited by 44% of respondents as a major obstacle to better decisions. Pressure to bind business quickly followed at 38%, while 35% pointed to insufficient context on how similar risks had performed previously.
Manual data entry between systems was the most frequently cited workflow problem, selected by 42% of underwriters. These issues leave teams spending time assembling information while also making decisions with incomplete historical or portfolio context.
When respondents compared faster systems with tools providing more context and reasoning, the majority generally preferred additional context. The strongest result involved live portfolio information, where 73% favored a live book signal over an automatically refreshed report.
Underwriters also showed interest in tools that explain how similar risks performed and how a new risk would affect the existing portfolio. This places decision support closer to the point of underwriting rather than limiting AI use to submission processing.
The report also identified a significant concern around the loss of experienced underwriting judgment. Forty-four percent named senior expertise leaving the organization without being transferred as one of their three largest concerns.
That ranked ahead of concern about AI making decisions too early in the underwriting process, cited by 40%. Another 37% pointed to the speed at which new and emerging risk classes are developing.
Investment priorities don’t closely match those concerns. Only 8% of respondents said their organizations planned to invest in coaching and knowledge transfer over the next 12 to 18 months, placing the category last among ten areas examined.
Workflow automation received substantially more attention. Fifty-two percent identified end-to-end automation as an investment priority, while 48% cited submission ingestion.
The problem appears more pronounced among US carriers. Thirty-seven percent of US underwriters said senior expertise primarily remains in individuals’ heads, compared with 26% of UK respondents.
Across both markets, 40% said underwriting knowledge is documented poorly or isn’t documented at all. That creates a risk that experienced employees leave without transferring the reasoning behind established underwriting practices.
Underwriters were also reluctant to use AI as a substitute for human coaching. Fifty-three percent said AI should either remain outside junior coaching entirely or be available only when specifically requested.
Respondents described a similar boundary around final underwriting decisions. Their preferred future model gives AI responsibility for tasks such as data preparation, triage and drafting while experienced underwriters retain control of complex risks and final approval.
That preference was broadly consistent across seniority levels. AI was viewed primarily as a source of information and recommendations, with the ultimate decision remaining with the underwriter.
The survey therefore shows a gap between where insurers are investing and where senior underwriters see one of their largest operational risks.
Automation and submission processing are receiving substantial funding, while formal knowledge transfer remains a relatively small investment area.
That distinction could become more significant as experienced underwriters retire or leave carriers. The quality and consistency of future underwriting decisions will depend partly on whether their expertise is captured in a form that other underwriters can use.
