Akur8 and the IFoA Keep Actuaries in the Loop as Insurance Jobs Fall by 95,000

Akur8's new pricing agents can assemble a rating structure, check variable usage before a model is fit, and lift rating logic out of a spreadsheet, yet users "approve changes before anything is executed" (Akur8, September 2026). Five days earlier the Institute and Faculty of Actuaries urged every AI project to keep an actuary in the loop. Both place the actuary at sign-off, downstream of the build.

They landed on an industry that keeps getting smaller. U.S. insurance carriers and related activities employed 2,930,400 people in September 2026, a preliminary figure 95,400 below the February 2025 peak, and the losses have accelerated through 2026 as the hard market cools (BLS, October 2026). The approval seat both releases promise looks secure for now. The build work that feeds it is what the agents take over.

Key Takeaways

About 6,900 jobs a month left insurance carriers and related activities in the first nine months of 2026, against roughly 1,900 a month across 2025, taking seasonally adjusted employment to 2,930,400 in September's preliminary estimate (BLS).
51,600 of the losses through August came at carriers and 40,200 at agencies, brokerages and related services, a near-even split across both sides of distribution since February 2025 (BLS).
19,470 actuaries worked in the sector in May 2025, so the 95,400 decline is nearly five times its entire actuarial workforce, and reporting on the jobs data places most losses in entry-level claims and rules-based roles (BLS; Insurance Business).
4.8% renewal premium change at Travelers Business Insurance in Q2 2026, down from 12.9% in Q3 2023, moves the pricing question from how much rate to where it goes, the structure work Akur8's agents automate.
82% of pricing actuaries told hyperexponential's 2025 survey they lack key skills such as coding, while the IFoA's lightest form of actuarial involvement is a few days of independent review.

What Akur8's Agents and the IFoA Manifesto Each Commit To

Akur8 introduced Agents for pricing at ITC Vegas on September 29, the first release of a line the company says will extend to reserving and life (Akur8, September 2026). The agents build pricing structures and formulas and check variable usage before modeling begins. They also pull filed rates into rating analyses and port spreadsheet rating logic into the platform through what Akur8 calls a reviewable process. Akur8's own statistical engines still fit the models, for a platform the company says more than 3,000 actuaries use daily across 350-plus customers.

Chief executive Sam Falmagne set out the division of labor: the engines "continue to do the mathematical computations, actuaries continue to make the decisions" (Akur8, September 2026). Every output can be inspected and edited before it executes. Calling these agents autonomous overstates the design; the release describes an approval-gated assistant, and none of its listed tasks changes a rate on the agent's own authority.

The IFoA's manifesto, titled Why involve actuaries in your AI projects?, was written by president-elect Alex Waite and published September 24 (IFoA, September 2026). It addresses executives, chief risk officers and regulators. It sorts AI into predictive models, with pricing as its first example, generative systems, and agentic systems that chain tools and decisions with a degree of autonomy. Its central request is that deployers ensure an actuary sits in the loop.

The document adds no practice requirement to the Actuaries' Code. It reads as a pitch to buyers, down to an FAQ on why a firm should pay actuarial fees, and it sizes involvement anywhere from a few days of independent review to an embedded team member. Practitioners appear to have settled into the reviewer role: the share of pricing actuaries fearing replacement by AI fell from 80% in 2024 to 49% in 2025 in hyperexponential's Coleman Parkes survey of 350 specialty and commercial practitioners (Reinsurance News, November 2025).

Where the 95,400 Jobs Went and Why Pricing Workloads Grew

The Bureau of Labor Statistics series for insurance carriers and related activities peaked at 3,025,800 in February 2025 and has fallen in every month of 2026 (BLS, October 2026). The 2025 decline averaged about 1,900 jobs a month; the first nine months of 2026 averaged about 6,900. The detailed split, which runs a month behind, shows carriers down 51,600 and agencies, brokerages and related services down 40,200 from February 2025 through August.

Actuaries are a thin slice of that payroll. Occupational data put 19,470 actuaries in the sector in May 2025, 13,500 of them at carriers, out of 26,670 employed nationally (BLS OEWS, May 2025). The payroll decline is almost five times that population, so arithmetic alone puts most of it outside actuarial departments. Reporting on the August jobs data traced the contraction to entry-level claims handling and routine, rules-based roles, with margin pressure cited alongside AI (Insurance Business, September 2026).

Pricing workloads move the other way as the market softens. Renewal pricing disclosed in carriers' own SEC filings has decelerated across commercial and personal lines, as the site's P&C Pricing Pulse tracks. Each row keeps its carrier's own metric definition, so the figures compare over time within a row and never across rows.

Carrier	Segment and disclosed metric	Peak	Q2 2026
Travelers	Business Insurance, renewal premium change	12.9% (Q3 2023)	4.8%
CNA	Commercial, renewal premium change	11.0% (Q2 2023)	2.0%
Hanover	Specialty, renewal price change	12.9% (Q3 2023)	3.6%
Hartford	Personal auto, renewal written price increase	25.7% (Q1 2024)	5.5%
Selective	Standard personal, renewal pure price	27.3% (Q4 2024)	8.9%

At 12.9%, most indications across a commercial book pointed the same way and a filing was largely a rate-level exercise. At 4.8%, margin depends on structure: which classes keep rate, which get some back, which variables earn a place in the model. A carrier that hands back rate evenly, while competitors cut only where their models show margin, loses its most profitable risks to them and keeps the rest.

That structure work is Akur8's task list: variable checks, new pricing structures, competitor filed rates folded into rating analysis. The cycle leaves less room for error. At a 70% loss ratio, each point of loss trend that rate fails to cover adds about 0.7 of a point to the combined ratio, and Hartford's personal auto renewal pricing fell from 25.7% to 5.5% in nine quarters. The agents let a smaller pricing team run more class-plan iterations per filing cycle, the same expense trade the site examined in Chubb's planned 20% headcount reduction.

The Approval Step Runs on Skills Learned in the Build

The manifesto rests its case on six trained skills, and one is to "hunt for the ways models fail: edge cases, drift, feedback loops" (IFoA, September 2026). Pricing actuaries have mostly acquired that skill by doing what Akur8's agents now do: checking variables by hand, rebuilding a rating algorithm from a spreadsheet and finding the factor that will not reconcile, writing the formula that exposes an implausible relativity.

The people who will approve agent output already report a gap. In hyperexponential's survey, 82% of actuaries said they lack key skills such as coding, and 70% said faster, more accurate pricing models would transform their role, up from 39% a year earlier (Reinsurance News, November 2025). Demand runs toward speed, and the shortfall sits in the skills a reviewer uses to interrogate faster output. The site's analysis of agentic pricing pipelines traced how one bad ingestion assumption compounds through development factors into the rate indication with no step along the chain built to catch it.

The arrangement holds in 2026 because the actuaries approving agent output learned on the manual version of the same tasks. The approval gate and the training ground occupy the same hours, however. An analyst who starts on agent-built structures spends the first years of a career signing off on work never assembled by hand.

Akur8 has named reserving and life as the agents' next workflows. A pricing structure error surfaces in the next policy year's loss ratio, where a few days of independent review, the manifesto's lightest option, can plausibly catch it. A mis-selected development pattern in a long-tail reserve surfaces as adverse development across several accident years, and the reviewer approving it may never have selected one.

Akur8 and the IFoA Keep Actuaries in the Loop as Insurance Jobs Fall by 95,000

Key Takeaways

What Akur8's Agents and the IFoA Manifesto Each Commit To

Where the 95,400 Jobs Went and Why Pricing Workloads Grew

The Approval Step Runs on Skills Learned in the Build

Further Reading

Sources