AI for Insurance Meets a Claims Adjuster Backlash
Insurance claims adjusters are pushing back on AI for insurance, according to recent reporting that surfaced a sharp rise in worker complaints during 2025 and 2026. The reaction matters because claims is one of the clearest tests of whether AI reduces operating cost or simply moves error-handling onto staff. According to WIRED’s report on Glassdoor research and adjuster interviews, the strongest resistance is coming from the people expected to work alongside the systems every day.
Claims adjusters are turning on AI
The headline finding is unusually stark. Glassdoor found that claims adjusters who mention AI in reviews are negative about it 98 percent of the time, a signal that this is not routine workplace grumbling but a workflow problem spilling into employee sentiment.
That statistic lands because claims work sits at the intersection of customer stress, time pressure, and documentation complexity. Initial loss reporting, routing, document review, estimate preparation, and customer communication all depend on clean handoffs. If AI gets the first step wrong, the cost does not disappear. It cascades.
One review cited in the reporting complained about leaders “forcing error-prone AI” onto teams and clients. Former claims worker Ahmad Jackson told WIRED that AI was “getting things wrong” and adding more work onto adjusters after misclassified claims had to be rerouted by hand. Geoffrey Conrad, a claims executive in Alabama, described the mood more bluntly: there is “an AI fatigue.”
This is why AI deployment services in insurance cannot be judged on intake speed alone. A claims funnel can look faster at the top while quietly creating delay, rework, and customer frustration downstream.
Why claims workflows break when AI is rushed
The operational issue is less about whether models can summarize text or classify incidents in isolation. It is whether those outputs hold up inside a live claims workflow, where a wrong label can send a case to the wrong team, create an inaccurate summary for an adjuster, or mislead a claimant on next steps.
In practical terms, the failure pattern described by adjusters is familiar across AI workflow automation projects. The model handles a seemingly structured task, confidence appears high, and then edge cases pile up in the exception queue. Humans end up validating, correcting, rerouting, and explaining errors to customers. The software processes volume, but the people absorb the ambiguity.
For insurers, that creates a hidden cost stack:
- more touches per claim when intake is misclassified
- more supervisor escalation when summaries are incomplete or wrong
- more customer friction when adjusters repeat or reverse prior guidance
- more distrust from staff who are measured on outcomes they do not fully control
This is also where enterprise AI integrations get harder than demos suggest. Claims systems connect policy data, photos, adjuster notes, third-party estimates, medical records, and payout workflows. Each integration point is another place where a small AI error becomes a larger operational one.
The labor-market signal behind the backlash
The worker reaction is arriving alongside a real shift in the profession. The Bureau of Labor Statistics outlook for claims adjusters and examiners projected a decline of 6 percent from 2025 to 2035. Glassdoor also cited government labor data showing that adjuster employment has fallen sharply in the past year.
Glassdoor’s economist Chris Martin’s analysis said entry-level adjuster job postings have fallen 50 percent since 2025. That matters because early-career roles are often where firms build future claims judgment. If AI business automation removes training-ground work too quickly, insurers may reduce the very pipeline needed for complex claims later.
That is a more important signal than the backlash headlines alone. AI implementation services may reduce time on routine tasks, but they can also hollow out skill development if companies automate before they define what humans still need to learn.
The tension is not unique to insurance, but claims makes it visible. In most sectors, weak automation creates internal inefficiency. In claims, it affects payouts, customer trust, and regulated decision paths.
Lemonade shows the appeal of full automation
There is a reason insurers keep pushing forward. The upside is real when the workflow is narrow enough. AI for insurance can speed up first notice of loss, extract data from uploads, summarize long records, and route simple claims much faster than manual handling.
Lemonade’s claims approach is the clearest example of the industry’s automation ambition. The company has said its bot handles initial reports most of the time, and SEC-filed materials indicate automation supported roughly 55 percent of all claims by the end of 2025. For simple, well-documented claims, that kind of model can reduce wait times and operating cost.
But full automation depends on disciplined inputs and tightly bounded scenarios. A clean photo set is not the same as a disputed liability case. A straightforward renters claim is not the same as a property loss with incomplete documentation. Even in AI-heavy models, the hard part is not just classification. It is exception handling.
That is why the comparison many insurers should make is not human versus machine. It is narrow automation with strong controls versus broad automation with weak handoffs.
Why human-plus-digital is still the safer default
More traditional carriers have leaned into a blended operating model. State Farm’s digital claims positioning emphasizes a mix of human and digital support rather than a pure automation promise.
That looks less dramatic on paper, but it matches how claims actually behave. The best use cases for AI integration services in claims are repetitive, structured, and easy to verify: intake routing, document extraction, summarization with review, duplicate detection, and status updates. The hardest cases still require judgment, customer communication, and the ability to notice when a file does not fit the pattern.
The non-obvious operator lesson is that failure usually starts at handoff design, not model quality alone. If a claim can be kicked from bot to queue to adjuster without a clear reason code, confidence score, and rollback path, the adjuster inherits a mess. If those controls exist, even imperfect AI process automation can still be useful because humans know what to trust and what to challenge.
A more mature deployment approach is to treat claims AI as a routing and drafting layer first, not as an autonomous decision-maker everywhere. That lowers the cost of mistakes and makes performance easier to measure claim by claim.
What insurers should do before scaling AI in claims
The near-term takeaway is not that AI for insurance has stalled. It is that the operating model matters more than the headline capability.
Insurers scaling AI automation implementation in claims should start with a few practical rules:
- Instrument the workflow before expanding automation so teams can see where misroutes and rework begin.
- Define which claim types are safe for automated handling and which must trigger human review.
- Track customer-facing error rates, not just internal speed metrics.
- Build rollback criteria before launch, so a failing workflow can be narrowed quickly instead of defended too long.
- Protect early-career learning paths, especially where routine claims build future judgment.
The companies that get value from AI deployment services in claims will not be the ones with the most automation in the shortest time. They will be the ones that decide, in advance, where software is reliable, where people stay accountable, and how exceptions move without creating more work than they remove.
What to watch next is whether insurers respond to the backlash by slowing deployments or by tightening the design of claims workflows. The stronger signal will come from operational metrics: fewer reroutes, fewer summary errors, and fewer adjusters saying the AI created a second job.
Related reads
Martin Kuvandzhiev
CEO and Founder of Encorp.io with expertise in AI and business transformation