What happened
WIRED reports that insurance claims adjusters had the most negative AI sentiment among American workers in a Glassdoor analysis: 98 percent of reviews mentioning AI were critical. Adjusters told WIRED that AI misclassified claims, generated inaccurate summaries and sometimes shifted the burden of correcting errors onto human workers.
WIRED reports that Glassdoor research found 98 percent of reviews by insurance claims adjusters that mentioned AI were negative, making the profession the most critical of AI in the analysis described by the article. The reviews reportedly complained about leaders forcing AI systems onto workers and customers despite perceived errors. Glassdoor senior economist Chris Martin told WIRED that he initially did not expect the result and interpreted it as evidence of a broader reckoning in the profession. The supplied material does not include the underlying Glassdoor dataset, methodology, number of reviews or comparison figures for other occupations, so AI Understanding has not independently confirmed the 98 percent finding.
WIRED reports that Ahmad Jackson, who previously worked in claims for a major insurer, described an AI system used for initial loss reporting. According to Jackson, the system sent a sudden influx of misclassified claims to adjusters, who had to reroute them. He also told WIRED that AI-generated hallucinations appeared in claims summaries and that adjusters sometimes passed those errors to claimants or their attorneys before discovering the problem. Jackson said the system added work and that he later left the company. WIRED also quotes claims executive Geoffrey Conrad describing widespread exhaustion from AI being imposed on workers.
WIRED reports that the claims industry is also facing employment pressure. The article says the Bureau of Labor Statistics projected in 2024 that the number of U.S. claims adjusters would decline by 18,900, or 5 percent, over the following decade. WIRED further reports that BLS data showed sector employment falling 21 percent between May 2025 and May 2026, while Glassdoor reported a 50 percent decline in entry-level postings since 2025. The article attributes part of the decline to technology and describes insurer efforts to automate claim intake, image-based damage estimates, medical-record summaries and payout processing. Those figures are not independently confirmed in the supplied material.
WIRED places those complaints alongside examples of current or intended automation. The article reports that Lemonade’s proprietary chatbot, AI Jim, handled initial reports 96 percent of the time by the end of 2025, while automation handled roughly 55 percent of all claims. Lemonade spokesperson Paul Staats told WIRED that automation is intended to let employees concentrate on complex claims requiring empathy and expertise. State Farm representative Justin Tomczak likewise told WIRED that the company wants to give employees better tools. The article also reports that adjusters see some limited value in AI, including administrative tasks such as extending a rental-car booking.
Why it matters
The report connects workplace resistance to practical failures in a high-stakes process. Insurance claims affect payments after accidents, property damage and medical events, so inaccurate automation can create confusion for policyholders while leaving employees responsible for explaining or correcting the result.
WIRED’s reporting shows why automation in insurance is more consequential than a routine back-office software change. Claims processing can determine whether people receive money after a home fire, vehicle accident or medical event. The article reports that AI may analyze photographs and videos, summarize hundreds of pages of medical records and process uploaded documentation before generating a payout. When such systems omit a detail or misread a document, the result can affect a customer’s payment. The supplied article does not provide audited examples of specific underpayments or wrongful denials, so those outcomes should not be treated as established beyond the accounts attributed to WIRED’s sources.
The report also highlights an accountability problem. WIRED reports that customers may not know an AI system caused confusion or misinformation and may instead blame the adjuster who communicates the result. That can leave human employees responsible for repairing errors they did not create, while customers may have difficulty challenging an automated process. Sandy Avina, a former adjuster who now consults in the insurance industry, told WIRED that adjusters have limited faith in AI output and described how a smudge in an attorney’s document or a missing point in a medical summary could lead to an incorrect payout. These are reported concerns, not independently measured error rates.
Employment effects add another layer. WIRED reports that entry-level claims postings declined sharply while insurers and startups promoted automation as a way to handle more work with fewer people. Chris Martin of Glassdoor told WIRED that reviews become more anti-AI when workers sense layoffs, suggesting that negative sentiment may reflect job insecurity as much as dissatisfaction with software. That interpretation is Martin’s analysis as reported by WIRED; the supplied material does not establish that AI caused the reported employment decline, nor does it separate AI-related reductions from other economic, organizational or regulatory factors.
The article presents a more limited view of what successful deployment might look like. WIRED reports that adjusters see value in narrow administrative uses, while company representatives describe AI as a tool for freeing employees to focus on complicated cases. Conrad’s account to WIRED argues that claims work can require empathy and judgment after traumatic events, including a total-loss house fire. The public-interest issue is therefore not simply whether insurers use AI, but whether automation preserves meaningful human review and support when customers face high-stakes or emotionally difficult claims.
What to watch next
Key questions include whether insurers can show that AI-assisted claims processing improves accuracy, how often humans review automated decisions and whether workers or customers can identify when AI contributed to an error. The supplied report does not independently establish the scale of these failures or the performance of specific systems.
Insurers should be pressed for evidence about how their AI systems perform in real claims, not only whether they reduce processing time. WIRED reports claims of misclassification, hallucinated summaries and missing information, but the article does not provide system-level error rates, independent audits or a breakdown by claim type. Useful disclosures would include how often automated outputs are corrected, how often payouts are changed after human review and whether error rates differ for medical, property or vehicle claims. None of those measures is supplied in the source.
The role of human review is another central issue. WIRED reports that some systems are intended to automate initial reports, image analysis, document review or payouts, while adjusters worry they are training their replacements. Future reporting should establish whether employees can override an automated recommendation, whether they have enough time and authority to investigate it, and whether customers are told when AI materially contributed to a decision. The supplied article does not say what appeal, disclosure or audit procedures any insurer currently provides.
Workforce data also warrants scrutiny. WIRED reports a 21 percent employment decline in the sector over one year and a 50 percent fall in entry-level postings since 2025, but the supplied material does not show the underlying datasets or prove that AI was the cause. Follow-up coverage should examine whether experienced adjusters are being redeployed, whether entry-level pathways are disappearing and whether productivity gains translate into better service, lower premiums or simply fewer jobs. Companies should distinguish planned automation from broader staffing changes.
Finally, the article suggests that automation may be more defensible for limited, low-consequence administrative work than for decisions that determine compensation after a loss. WIRED reports that Jackson found AI useful for routine rental-car extensions but objected to inaccurate claims summaries and broader replacement of human judgment. What to watch is whether insurers narrow systems to tasks with clear safeguards or expand them toward fully automated payouts. The source provides no independent assessment of Lemonade, State Farm or any other insurer’s current accuracy, customer outcomes or safeguards.