Most senior commercial underwriters say AI has saved them time on admin. Far fewer say it has improved the quality of their decisions. That gap shapes how submissions get handled, how risks get priced, and how consistent underwriting decisions are likely to be.
The Underwriting Edge 2026 report draws on a June survey of 350 senior commercial property and casualty underwriters across the US and UK, conducted by independent research firm Coleman Parkes. It was published by hyperexponential (hx), an AI underwriting software vendor that sells technology directly to the underwriting functions the report surveys. That commercial interest should be kept in mind when reading the findings.
Among underwriters who use AI, 51 percent say its greatest contribution has been saving time on manual admin. Only 21 percent say it has improved the quality of their decisions. The two numbers describe a technology that has been pointed at the wrong problem.
What underwriters say damages decision quality is mostly a context problem, rather than a speed one. The three biggest drags they name are inconsistent submission data (44 percent), pressure to bind quickly (38 percent), and no context on similar prior risks (35 percent). Manual data entry across systems tops the workflow blockers list at 42 percent.
Those blockers explain why underwriters want AI applied to context at the moment of decision - how similar risks performed, what writing a risk would do to the book - rather than to autonomous calls. Given a choice between a faster setup and one that carries more context and reasoning, a majority chose context each time. The margin was widest for a live book signal over a report that refreshes automatically - 73 percent preferred context over speed.
The survey also found commercial P&C carriers deploying AI broadly but unevenly, with strategic confidence lagging well behind the pace of tool deployment.
The sharpest finding is the gap between what worries underwriters and where their organizations are investing. Losing senior judgment without passing it on is their single biggest concern, named by 44 percent as a top-three worry. It sits ahead of AI deciding too early at the cost of craft (40 percent) and the pace of change in emerging risk classes (37 percent).
The investment picture runs the other way. Asked which areas of the submission-to-quote lifecycle their function is investing in over the next 12 to 18 months, underwriters ranked coaching and knowledge transfer last of ten. Only 8 percent said their organization is investing there.
End-to-end workflow automation and submission ingestion sit at the top of the investment list, at 52 percent and 48 percent, respectively. The gap has a US dimension: 37 percent of US underwriters say senior judgment mostly lives in people's heads, against 26 percent in the UK. A combined 40 percent across both markets say expertise is captured poorly, or not at all.
Underwriters are also clear about what they don't want, and that is AI filling that gap. Coaching juniors is the task they guard most closely from automation, with 53 percent wanting AI kept out entirely or available only on request.
The picture that emerges is one where underwriting will remain a human judgment function for the foreseeable future, with AI handling data assembly and process, rather than the decision itself. Asked to describe an ideal setup three years out, underwriters consistently described AI taking on data, triage, and drafting while the underwriter keeps complex risks and the final call.
That preference holds across seniority levels. The message throughout the survey is that AI should suggest and a person should approve.
The consistency of underwriting decisions and the depth of context applied to complex risks will depend on how well senior judgment gets captured before it walks out. Analytics investment by P&C carriers is accelerating, with WTW's 2026 analytics survey finding carriers with stronger capabilities achieved combined ratios six percentage points lower than slower adopters. The hx survey suggests that investment is flowing predominantly into process rather than judgment, which is the part that determines how a submission actually gets evaluated.
