Commercial rates are sliding just as AI agents and automation flood underwriting workflows, creating a dangerous mix of softer pricing, fragile infrastructure and a shrinking talent bench that could leave carriers exposed to mispriced risks and eroded profitability in 2026.
Send’s latest underwriting trends report warns that nearly every major commercial line except excess casualty has tipped into soft market territory in North America. Property, construction, cyber, marine and professional liability are all described as soft or softening, barring a major catastrophe that could still jolt pricing. At the same time, underwriters are under pressure to preserve the strong combined ratios achieved in the Covid era hard market, even as pricing power fades and competition intensifies.
In that environment, the risk isn’t just lower rates. It’s the temptation to lean on simplistic rules and automated declinatures instead of judgment, straining already fragile relationships with brokers and clients. The report warns that as prices move down, it becomes easy for carriers to fall back on short term decisions and a “computer says no” mentality that might protect a quarterly number while damaging long term distribution ties.
Andy Moss (pictured), co‑founder and CEO of Send, sees the same fault lines up close. For him, the stumbling block is often not the tools themselves, but the starting point. “For a lot of insurers, the first challenge is knowing where to begin or how to measure the value automation will deliver,” Moss previously told Insurance Business. “Even with a solid plan, getting it right is hard.”
That uncertainty keeps many companies anchored in outdated workflows. Moss has warned that Excel remains a persistent bottleneck for managing rules and rates, and that automation depends on “good data, standardized processes, and consistent rules” to function properly. Offline processes, version‑control issues and duplicated or missing entries simply don’t scale – and if underwriters don’t trust what comes out of a system, they go around it. “If this doesn’t happen, users will revert back to email or manual processes, and the benefits disappear,” he said.
The report makes it clear that many organizations still haven’t fixed those basics. It describes a market that wants instant responses, clean data flows and auditability, while still depending on patchwork cores and decades of accumulated technical debt. Experimentation around AI and point solutions has outpaced the hard work of cleaning legacy data and agreeing on standards. The result: systems that look modern at the front end but are stitched to brittle, inconsistent backbones at the back.
That’s especially worrying given how quickly AI is being embedded into underwriting decisions. The 2026 report says agentic AI is moving from theory into live workflows, handling everything from intelligent intake and triage to external research and training scenarios. Vendors and carriers are already talking about cutting the time spent on an underwriting task by up to three-quarters and doubling individual productivity.
On paper, that sounds like the efficiency story every chief underwriting officer wants to hear in a soft market. In practice, if those agents are ingesting inconsistent data and opaque rules, they can accelerate bad decisions as easily as good ones. The report admits that even the most sophisticated tools are only as strong as the information they’re learning from; if formats are inconsistent, fields incomplete and systems incompatible, automation amplifies noise rather than insight.
Moss has consistently cautioned against “automating for automation’s sake.” In his view, insurers need to “automate with purpose” – be clear about the value they expect, which steps should be automated, and “where a human should stay in the loop.” Some processes will never make sense to hand over entirely, and the prize is freeing underwriters to focus on higher‑value work like complex risk analysis and broker relationships, rather than burying them in faster low‑value tasks.
All of this is happening against a backdrop of rising macro uncertainty. The report leans on World Economic Forum and Swiss Re analysis to underline the point that most chief economists expect global trade to split further into competing blocs through 2026, while climate related losses continue to produce some of the costliest events on record. The fact that overall insured catastrophe losses in 2025 still allowed for strong underwriting results can create a dangerous complacency in a softer market.
Then there is the talent problem. The report devotes an entire trend to how the underwriting workforce is being reshaped – and not always in healthy ways. Between looming retirements and the automation of junior tasks, the market risks losing the apprenticeship ladder that used to teach underwriting from the ground up. Industry leaders quoted in the paper warn that while judgment can’t be handed over to a model, it can be allowed to disappear if too much early career work is stripped out and mentoring isn’t rebuilt around the new tools. One talent leader argues that without deliberate pathways, mentorship and knowledge transfer, the industry will “hollow out the pipeline” just as volatility rises, with women and emerging leaders hit hardest.
Put together, the picture is far from the clean “AI-powered efficiency story” the market likes to tell. Rates are softening across most major lines. Geopolitics and climate are making the risk landscape more uncertain, not less. Meanwhile, many carriers are still trying to bolt advanced automation and agentic AI onto data and process foundations that were never designed for this level of speed, scrutiny and connectivity, while the pool of experienced talent slowly drains away.
For Moss, responsible use of technology still comes back to balance. “The insurance industry needs to meet profitability goals without losing sight of the fact that insurance is built on trust and relationships,” he said. Automation, when done right, is there to support that balance – aligning company, broker and client interests so that data‑driven discipline and client centricity work together, not in opposition.
