Insurers Cut 21% of Adjusters, Shift to Agents

An editorial note published by the Insurance Journal on September 21 revealed a number that the insurance sector has been avoiding: among the employee evaluations compiled by the Glassdoor Employee Confidence Index, claims adjusters are the most critical group regarding artificial intelligence, with 98% negative mentions. In the same survey, employment in the segment fell 21% over twelve months, and entry-level positions for adjusters dropped by 50%.

American official data aligns with this trend. Employment in financial services and insurance in the United States fell from 6,733,400 jobs in August 2025 to 6,651,100 in August 2026, a loss of approximately 82,000 positions, according to the Bureau of Labor Statistics. In August alone, there were 7,400 fewer jobs, of which 6,300 were in insurance companies and related activities. Claims adjusters accounted for 13,100 of the losses for the year, or 17.5% of the total 74,800 cuts in the carrier segment.

Automation Already in Production

Allianz built Project Nemo, a multi-agent system for handling perishable food claims that reportedly reduced processing time by 80%, according to the insurer. Each agent manages a step, from receiving the notice to recommending payment, with cases being escalated to humans only as exceptions. Lemonade reports that by the end of 2025, 96% of claims notifications were received without human intervention, and 55% of claims were resolved end-to-end automatically.

The most explicit cut came from brokerage. In an internal memo dated May 20, Acrisure's CEO, Greg Williams, announced the elimination of 2,250 positions, about 11% of the workforce of 19,000 across 24 countries, attributing the decision to the use of AI, data, and automation to reduce manual work.

The Counter-Argument is Named

Robert Hartwig, a professor at the Darla Moore School of Business at the University of South Carolina and former chief economist at the Insurance Information Institute, believes the causality between job cuts and AI adoption is murkier than it appears. The Federal Reserve has been cutting interest rates since September 2024, down to 3.63% in August 2026, and the most significant employment drop follows more than a year of loosening. Median organic growth in American retail brokerage slowed to 5.9% in the first quarter of 2026, the weakest pace since 2021. Part of what is attributed to algorithms is tight margins.

Martha Gimbel, executive director of Yale's Budget Lab, succinctly contends that we do not have a good track record in predicting how technological changes fit into the labor market. Researchers from Brookings measured American occupational composition since the launch of ChatGPT and did not find the sharp dislocation that headlines projected.

There is also a statistic that weakens the thesis of the profession's collapse, and it deserves to be highlighted: senior positions for adjusters remain about 80% above 2017 levels. What shrinks is not the occupation but the base of it. The distinction that the shallow debate ignores is precisely this. Automating the receipt of a claims notice is a structured problem solved through data extraction and rules. Arbitrating coverage, setting reserves, and sustaining a denial in the face of litigation is not. Lemonade's 55% refers to small, straightforward, low-severity claims. The correct measure is not how many claims the machine closes, but which layer of the pyramid it removes.

United States, Germany, India, and Brazil

In the United States, the cut is measured and visible: BLS, Glassdoor, and memos like Acrisure's in Grand Rapids. In Germany, Allianz is shifting work without cutting jobs at the same ratio, which changes the makeup of the workforce before changing its size. In India and the Philippines, where shared service centers process claims for American and British insurers, the effect is second-order and more brutal: outsourced work there has been exactly the entry-level jobs that are now becoming agents. In Brazil, exposure is in claims operation centers and BPOs servicing local and foreign insurers, and there is a regulatory difference: Susep requires decision traceability in coverage denials, and an agent that decides without an auditable trail creates liability, not efficiency.

The Bureau of Labor Statistics projects a 6% decline in the employment of claims adjusters between 2025 and 2035. This projection assumes a funnel that continues to develop personnel. With entry-level positions dropping 50% in a year and postings for the role reducing by 55% since the post-pandemic peak, compared to a 36% drop in the general labor market, the constraint by 2032 will not be an excess of adjusters. It will be the absence of professionals with ten years of experience to review what the agent has decided.