Commercial rates fell for the first time since 2017. The rate gap underneath didn’t get any bigger. It just stopped being invisible.
For almost nine years, commercial P&C rates only moved one direction. That changed this year. In the first quarter of 2026, average commercial premiums actually fell, the first drop since 2017 and the end of a 33-quarter streak, according to the Council of Insurance Agents & Brokers’ Q1 2026 market index.
The picture isn’t uniform, though. Commercial auto kept climbing, up 5.8% in Q1 alone, its 59th consecutive quarterly increase, with loss ratios running above 100% most years since 2014, according to AM Best’s numbers in that same report. Property, workers’ comp, cyber, and D&O moved the other way, as more capacity came into the market and competition picked up, which IMA Financial Group’s Q2 outlook calls out directly. Fitch settled on “neutral” for the year overall, though that word is really just an average of two different stories depending on the line in question.
None of this changes how a rate gap behaves, just how visible it is. In a hard market, a small gap between a filed rate and the current advisory number rarely gets noticed, since rates are already going up regardless a missed circular just means a slightly smaller increase than was possible. When the market softens, that cushion goes away, and the same gap shows up directly in loss ratio instead. Deloitte’s 2026 outlook calls this the start of margin pressure across the industry less room, in practice, for anything to go unnoticed.
A lot of carriers still keep this kind of monitoring filed under compliance, next to license renewals and training records. A loss cost change that isn’t matched by a filed rate is a pricing decision either way, made or not.
Regulators are circling the same gap
Roughly half of US states have adopted the NAIC’s Model Bulletin on AI use in insurance, and its companion evaluation tool is mid-pilot right now, running through September 2026 across a dozen states. Colorado’s own AI law, delayed once already, is due to take effect June 30, 2026. Nothing here is finalized, but it’s moving in one direction: a rate gap with no clear link back to the bulletin that caused it is a harder thing to explain in an exam than it used to be.
More people isn’t quite the fix
Adding a person is the usual response another analyst, another task on someone’s calendar. It doesn’t hold up for long, mostly because reading was never really the bottleneck. Matching a bulletin against every class code, state, and renewal date it might touch is slow, manual work, and in most shops that work lives inside one or two people who’ve done it long enough to move quickly. If one of them leaves, nothing announces itself as broken. Things just start taking longer, and it’s a few quarters before anyone connects why.
