You Know Who Hates AI? Insurance Claims Adjusters
Introduction: The Quiet War Between Adjusters and AI
Picture a claims adjuster's day in 2015. You're in a rental car at 7 a.m., coffee in hand, driving to a three-car pileup on the interstate. You photograph skid marks, measure crumple zones, and interview a shaken driver who keeps apologizing for something that wasn't her fault. You call the body shop, negotiate a repair estimate, and write a report that a claims manager will actually read. By Friday, you've settled eight claims—each one involving a human being who needed something from you.
Now imagine the same job in 2025. A policyholder submits a claim through an app at 2 a.m. An AI chatbot asks a few questions and logs the loss. Computer vision algorithms analyze the photos the policyholder uploaded from their driveway. A fraud detection model scores the claim a 4.3 out of 10 on a risk scale nobody fully explains. If the score is low enough, the claim is approved and paid out without a single human looking at it. If it's flagged, it lands in your queue with a note: "Verify damage. AI confidence: 68%."
You didn't drive anywhere. You didn't talk to anyone. You're not even sure why the system flagged this particular claim.
This is the quiet war between insurance claims adjusters and artificial intelligence—and it's not really about the technology. It's about who gets to make decisions, who gets to talk to people, and whether a profession built on judgment and empathy is being reduced to cleaning up after algorithms.
Here's what we'll cover: what adjusters actually do, how AI has infiltrated every stage of the claims process, why adjusters are pushing back, the real-world costs of AI errors, the numbers behind the tension, the regulatory battleground, and where this is all heading.
What Exactly Do Claims Adjusters Do?
Before we dive into the conflict, it's worth understanding the job itself. Claims adjusters investigate insurance claims to determine what the insurer owes. They work in auto, property, workers' compensation, liability, and specialty lines. Their core responsibilities break down into four buckets:
Investigation. When a claim comes in, the adjuster gathers facts. They inspect damaged property, interview witnesses, review police reports, consult medical records, and sometimes hire experts—engineers, accident reconstructionists, or medical reviewers—to help establish what happened.
Assessment. Once the facts are gathered, the adjuster determines coverage. Is this loss covered under the policy? What's the actual cash value of the damaged property? How much will repairs cost? What's the depreciation? This requires reading policy language carefully and applying it to a specific, often messy situation.
Negotiation. Adjusters don't just write checks. They negotiate with body shops, contractors, medical providers, and sometimes lawyers. They push back on inflated estimates. They explain coverage limits to policyholders who are upset, confused, or grieving.
Settlement. Finally, the adjuster issues payment or denies the claim—and when they deny it, they need to explain why in a way that holds up to scrutiny, both internally and potentially in court.
The human skills here aren't incidental; they're the core of the job. Judgment matters when policy language is ambiguous. Empathy matters when a family just lost their home to a fire. Contextual understanding matters when a claim involves emotional distress, disputed liability, or a policyholder who is being less than honest. An adjuster who can read a room—or a police report—catches things an algorithm can't.
Why are adjusters essential to the insurance ecosystem? Because insurance is a promise. The policyholder pays premiums in exchange for the insurer's commitment to pay when something goes wrong. The adjuster is the person who makes good on that promise—or breaks it. They're the face of the company at the moment it matters most.
How AI Is Infiltrating the Claims Process
AI hasn't arrived all at once. It's crept in piece by piece, starting with the most routine tasks and working its way toward the ones that require judgment.
Claim intake. Chatbots and virtual assistants now handle the first point of contact for many claims. Lemonade's "Jim" is the most famous example—it asks policyholders a series of questions, cross-references their answers with their policy, and can approve simple claims instantly. But this isn't just a Lemonade phenomenon. Major carriers including Allstate, Progressive, and State Farm have deployed AI-assisted intake systems that gather initial information before a human gets involved.
Damage assessment. Computer vision has made huge strides in analyzing photos and drone footage. An adjuster used to climb a roof or walk a body shop lot; now AI can look at images of a dented fender or a damaged roof and produce a repair estimate in seconds. Some systems can even detect subtle structural issues that a human might miss—though, as we'll see, they also invent damage that isn't there.
Fraud detection. Insurers have poured money into machine learning models that flag suspicious claims. These systems analyze patterns across thousands of claims—the timing of a loss, the language used in a report, the policyholder's history—and assign a risk score. The problem? Fraud models are notoriously prone to false positives, and adjusters end up spending hours investigating claims that are perfectly legitimate.
Claim triage. AI systems now sort claims by urgency and complexity. A minor fender-bender with clear photos and no injuries gets routed to a fast-track, low-touch process. A multi-vehicle accident with injuries and disputed liability gets flagged for a senior adjuster. This sounds sensible in theory, but it means the AI is deciding which claims deserve human attention—and which ones don't.
Document processing. Natural language processing tools extract data from medical records, police reports, and repair invoices. They can transcribe doctor's notes, pull key dates and figures, and populate claims management systems automatically. This saves time, but it also introduces a new failure point: if the AI misreads a document, the error propagates through the entire claim.
Key Takeaway: AI in claims isn't one thing—it's a stack of tools layered into every stage of the process, from first notice of loss to final settlement. Each tool was sold as a way to make adjusters more efficient. Each one has also shifted decision-making power away from them.
Why Adjusters Are Pushing Back
The resistance from adjusters isn't Luddism. It's a set of concrete, legitimate concerns about their jobs, their profession, and the quality of outcomes for policyholders.
Fear of job loss. The numbers are stark. McKinsey estimates 45% of adjuster tasks are automatable. The Bureau of Labor Statistics projects a 6% employment decline by 2032. A 2023 survey by the National Association of Independent Insurance Adjusters found that 58% of adjusters believe AI will replace them within a decade. When your profession is told, year after year, that a machine can do what you do, you pay attention.
Deskilling. This is the more insidious concern. Even when adjusters keep their jobs, the nature of the work changes. AI triage systems decide which claims are "simple" and route them away from human review. AI damage assessment tools generate estimates that adjusters are expected to rubber-stamp. Over time, adjusters lose the reps—the day-in, day-out practice of assessing damage, reading policies, and negotiating—that build expertise. The profession deskills from the inside, and the people who are supposed to catch AI errors become less capable of doing so.
The "black box" problem. Many AI systems in claims are opaque. The adjuster sees a risk score or a fraud flag, but not the reasoning behind it. When a claim is denied based on an algorithm's output, the adjuster is expected to defend that decision—without being able to explain it. This is professionally untenable. You can't advocate for a decision you don't understand.
False positives and errors. AI systems make mistakes, and when they do, adjusters clean up the mess. The Wharton study found a 12% error rate in AI image recognition for auto damage. That means roughly one in eight claims requires a human to catch and correct an algorithmic error. For adjusters, this isn't a hypothetical concern—it's a daily reality. They're not doing their own work; they're doing the AI's work over again.
Loss of client interaction. The entry-level claims that AI handles—the fender-benders, the minor water damage, the stolen laptops—were traditionally how adjusters built relationships with policyholders. They were also the farm system for the profession. New adjusters cut their teeth on simple claims before graduating to complex ones. With AI swallowing the simple claims, new adjusters have fewer learning opportunities, and policyholders have fewer human touchpoints with their insurer.
Key Takeaway: The adjuster's pushback isn't just "I might lose my job." It's "The job I was trained to do is being replaced by a process that makes worse decisions, and I'm the one held accountable for them."
The Real-World Consequences of AI Errors
Let's get specific about what happens when AI gets it wrong. These aren't hypotheticals.
Case study: Hail damage and shadows. In 2023, a major U.S. auto insurer used satellite imagery to identify hail damage and automatically deny claims where the imagery showed no damage. The problem? The AI was misidentifying shadows as hail damage—and, more consequentially, missing actual damage that fell outside its training patterns. After adjusters flagged thousands of erroneous denials, the insurer reversed course. But the damage to policyholder trust was done. Customers who had been told their claims were denied based on "objective satellite data" had to be re-contacted and told the system was wrong.
Case study: Demographic bias in flood risk. A property adjuster in Florida reported that an AI-driven flood risk model flagged a homeowner's claim as "high risk" based on the neighborhood's demographics, not the actual condition of the property. The AI's risk score led to a lower payout offer. The adjuster, who knew the property and the neighborhood, manually overrode the system after an on-site visit. But the case illustrates a broader problem: AI models trained on historical data can encode historical biases—including redlining patterns—into their risk assessments.
Case study: Medical records misread. A workers' compensation adjuster used an AI transcription tool to process medical records. The system misread a doctor's note about "chronic pain" as "no pain." The claim was denied based on the AI's output. The policyholder appealed, a human reviewed the original records, and the denial was overturned. The cost wasn't just financial—it was the weeks of delay, the stress on an injured worker, and the erosion of trust in the system.
The cost of errors. These mistakes have real consequences. Financially, they result in unnecessary payouts (like the Texas case where an AI totaled a repairable car, costing the insurer $15,000), improper denials that lead to lawsuits, and regulatory fines. Operationally, they create a hidden workload—adjusters spending time correcting AI errors instead of handling claims that genuinely need human attention. And reputationally, each error is a story a policyholder tells their friends, family, and social media followers about how their insurance company screwed them over.
The Numbers Behind the Tension
Let's lay out the data that's driving both the industry's enthusiasm for AI and adjusters' anxiety.
McKinsey (2023): AI could automate up to 45% of adjuster tasks. This isn't the same as eliminating 45% of jobs—tasks get redistributed—but it's the number adjusters see in headlines.
Deloitte (2024): 62% of insurance executives report that AI has reduced claims processing time by at least 20%. For insurers, speed is a competitive advantage. For adjusters, speed isn't the only metric that matters—accuracy and fairness matter too.
Bureau of Labor Statistics (2023): Employment for claims adjusters is projected to decline 6% from 2022 to 2032. That's a net loss of roughly 5,000 jobs in a profession that employed about 84,000 people. The BLS specifically cites automation as a contributing factor.
NAIIA (2023): 58% of independent adjusters believe AI will replace their jobs within a decade. This is a perception gap worth noting—if adjusters believe they're replaceable, they're more likely to leave the profession, which becomes a self-fulfilling prophecy.
Juniper Research (2021): AI-powered claims processing will save insurers $1.3 billion annually by 2025, up from $170 million in 2020. That's real money—and it's money that comes from somewhere. Some of it is efficiency; some of it is labor cost.
Wharton (2024): AI image recognition tools misclassified 12% of auto damage claims. This is the number adjusters cite when they argue that AI isn't ready for prime time.
Key Takeaway: The economics of AI in claims are clear—it's faster and cheaper. The question is whether the savings justify the errors, the bias, and the loss of human judgment.
The Regulatory and Ethical Battleground
The fight over AI in claims isn't just happening in break rooms and union halls. It's happening in state legislatures and regulatory bodies.
State laws on bias audits. Colorado and Illinois have passed laws requiring insurers to audit their AI systems for bias. These laws are a direct response to concerns that algorithms could discriminate against protected classes—intentionally or not. The laws require insurers to document how their models work, test them for disparate impact, and report the results to regulators.
NAIC model guidance. The National Association of Insurance Commissioners has issued model guidance on AI transparency. The guidance calls on insurers to ensure that AI systems are explainable, that they're tested for bias, and that humans maintain oversight of automated decisions. The NAIC guidance isn't legally binding, but it shapes how states regulate.
The risk of algorithmic bias. The concern isn't that insurers want to discriminate. It's that AI models trained on historical claims data can learn historical patterns of discrimination. If an insurer's historical data reflects redlining or other biased practices, the AI will reproduce those patterns—at scale. This is the same problem that's emerged in hiring, lending, and criminal justice, and insurance is no exception.
Calls for "human-in-the-loop" requirements. The most concrete demand from adjuster advocacy groups is for regulations requiring human review of AI decisions before they become final. This isn't a radical ask. It's already the standard in high-stakes domains like credit decisions, where the Equal Credit Opportunity Act requires adverse action notices to include specific reasons—reasons that a human must be able to articulate.
Key Takeaway: The regulatory landscape is shifting, but it's uneven. Some states are moving toward meaningful oversight; others are taking a hands-off approach. The result is a patchwork of rules that insurers must navigate—and that adjusters can use to demand accountability.
How Adjusters Are Adapting and Fighting Back
Despite the grim headlines, adjusters aren't passive victims. They're adapting, organizing, and in some cases, fighting back.
Upskilling. Many adjusters are learning to work alongside AI rather than against it. This means understanding what AI tools can and can't do, knowing how to interpret AI outputs, and developing the skills to override AI when it's wrong. The adjusters who thrive in the hybrid model are the ones who treat AI as a tool, not a threat—or at least as a threat they've learned to manage.
Focusing on complex cases. As AI takes over routine claims, adjusters are shifting toward the complex cases that genuinely require human judgment: disputed liability, catastrophic losses, fraud investigations, and claims involving emotional distress or nuanced coverage questions. These are the claims where AI falls short and where experienced adjusters add the most value.
Professional advocacy. Groups like the National Association of Independent Insurance Adjusters and various state associations have been vocal about the need for transparency in AI decision-making. They're pushing for standards that would require insurers to explain AI outputs, test for bias, and maintain human oversight. Some of this advocacy is self-interested, but much of it is genuinely focused on protecting policyholders.
Unionization and collective bargaining. In 2024, a group of adjusters in Texas formed a collective to demand "human-in-the-loop" requirements after an AI system incorrectly totaled a repairable car, costing the insurer $15,000 in unnecessary payouts. This is part of a broader trend of white-collar workers organizing around AI concerns. Adjusters are realizing that individual complaints don't change corporate behavior—collective action does.
Becoming AI supervisors. The most forward-looking adjusters are positioning themselves as the humans in the loop. They're the ones who review AI outputs, catch errors, and override bad decisions. This isn't the job they were trained for, but it's a job that needs doing. The adjuster of the future isn't the person who climbs the roof; it's the person who knows when the AI's roof assessment is wrong.
Key Takeaway: Adjusters are fighting back not by refusing to use AI, but by insisting on a role in how AI is deployed. The goal isn't to stop automation—it's to ensure that automation serves policyholders rather than harming them.
The Future: A Hybrid Model or a Tug of War?
So where is this heading? The most likely path isn't total automation or total resistance. It's a hybrid model, but the terms of that hybrid are still being negotiated.
The likely path. AI will handle the routine, high-volume, low-complexity claims: the minor fender-benders, the small water damage claims, the straightforward property losses where the photos are clear and the coverage is unambiguous. Humans will manage the exceptions: the claims that AI flags as suspicious, the ones with complex coverage questions, the ones involving significant injury or disputed liability.
What this means for the adjuster role. The adjuster's job shifts from investigator to supervisor and exception handler. Instead of doing the initial damage assessment, they review AI assessments. Instead of triaging claims, they handle the ones the AI couldn't resolve. This is a significant change, and it requires different skills: data literacy, the ability to interrogate AI outputs, and the judgment to know when to trust a model and when to override it.
Potential outcomes. The optimistic scenario is that AI handles the boring stuff, adjusters focus on the cases that need human judgment, claims are processed faster, and costs go down. The pessimistic scenario is that AI errors go uncorrected, bias gets baked into automated decisions, adjusters become rubber-stampers with eroded skills, and policyholders suffer. The outcome depends on how insurers implement AI—and on whether adjusters have a voice in that implementation.
The importance of human oversight. Every serious analysis of AI in insurance concludes that human oversight is essential. The question is whether that oversight is meaningful or performative. A "human in the loop" who rubber-stamps 95% of AI decisions without review isn't providing oversight. A human who has the authority to override AI, the training to know when to do so, and the time to actually review claims is providing oversight.
Key Takeaway: The hybrid model isn't a compromise—it's the only model that makes sense. The question is whether insurers will invest in making the human half of the hybrid work, or whether they'll treat "human oversight" as a checkbox on a regulatory form.
Conclusion: Striking a Balance Between Efficiency and Empathy
The conflict between AI and claims adjusters isn't about technology. It's about values.
Insurance is built on a promise: when something goes wrong, the insurer will be there. The claims adjuster is the person who keeps that promise. They're the one who shows up at the scene of an accident, who explains a complicated policy to a grieving widow, who negotiates with a contractor to make sure a family's home is rebuilt properly. These are human tasks, and they require human skills—judgment, empathy, and the ability to navigate ambiguity.
AI can make the claims process faster and cheaper. It can process thousands of claims in the time it takes a human to handle one. It can analyze images, flag fraud, and extract data from documents. These are real capabilities, and insurers would be foolish not to use them.
But AI also makes mistakes. It lacks context. It can encode bias. It can't explain its reasoning. And when it gets things wrong, real people suffer real consequences—denied claims, delayed payments, unnecessary stress.
The path forward isn't AI versus adjusters. It's AI with adjusters. That means deploying AI where it works, keeping humans where they're needed, and giving adjusters the authority and training to override AI when it's wrong.
The adjusters who are fighting back aren't trying to stop progress. They're trying to ensure that progress doesn't come at the expense of the people insurance is supposed to serve. That's not a Luddite position. That's a professional one.
The future of claims adjustment depends on whether insurers listen.
Frequently Asked Questions
Will AI completely replace insurance claims adjusters?
No, but it will change the job significantly. AI is most effective at routine tasks like initial data collection, simple damage assessment, and document processing. Complex claims—those with disputed liability, nuanced coverage questions, or significant human impact—will continue to require human adjusters. The Bureau of Labor Statistics projects a 6% employment decline, not a 100% replacement. The adjuster's role will shift from doing the work to supervising AI that does the work.
Why do adjusters specifically hate AI?
It's not the technology itself—it's what it represents. Adjusters are concerned about job security, but more fundamentally, they're concerned about deskilling. AI is shifting decision-making power away from experienced professionals and toward algorithms and non-specialist managers. Adjusters are also frustrated by the "black box" problem: they're expected to defend AI decisions they don't understand and to clean up AI errors they didn't create.
How does AI actually help adjusters?
AI can handle the repetitive, high-volume tasks that consume adjuster time: initial claim intake, basic damage assessment from photos, data extraction from documents, and fraud screening. This frees adjusters to focus on the claims that genuinely require human judgment. AI can also process claims faster, which benefits policyholders who want quick resolution. The key is using AI as a tool, not a replacement.
What are the main risks of AI in claims adjustment?
The main risks are: (1) algorithmic bias, where AI models reproduce historical patterns of discrimination; (2) false positives and errors, where AI flags legitimate claims or misses actual damage; (3) lack of transparency, where AI decisions can't be explained to policyholders or regulators; and (4) deskilling, where adjusters lose the expertise needed to catch AI mistakes.
Are there regulations governing AI use in insurance claims?
Yes, but they're uneven. Some states, like Colorado and Illinois, have passed laws requiring insurers to audit their AI systems for bias. The National Association of Insurance Commissioners has issued model guidance on AI transparency and human oversight. However, many states have no specific AI regulations, and federal oversight is limited. The regulatory landscape is still evolving.
What skills should adjusters develop to stay relevant?
Adjusters should focus on skills that AI can't easily replicate: complex judgment, negotiation, empathy, and contextual understanding. They should also develop data literacy—the ability to understand AI outputs, identify when a model is wrong, and override it effectively. The adjusters who thrive will be the ones who can work alongside AI rather than compete with it.
Does AI reduce the cost of insurance for consumers?
Potentially, but it's not guaranteed. AI can reduce insurers' operating costs, which could lead to lower premiums. However, insurers may also keep the savings as profit. The bigger question is whether AI errors—improper denials, unnecessary payouts, regulatory fines—offset the efficiency gains. The evidence so far is mixed.
How accurate is AI at assessing damage compared to humans?
AI is fast but not always accurate. A 2024 Wharton study found that AI image recognition tools misclassified 12% of auto damage claims. AI is good at identifying clear, obvious damage that matches its training data. It struggles with edge cases, unusual damage patterns, and situations that require contextual understanding. Human adjusters are slower but more accurate on complex claims.
Are you an adjuster or an insurer navigating the AI shift? Share your experiences and thoughts in the comments below—we want to hear how AI is impacting your work.