How AI Will Replace Insurance Claims Adjusters: Agentic Adjudication and the $300B Industry Reckoning 2026-2030

On April 22, 2026, Google announced Gemini Enterprise as the central nervous system of its new agentic taskforce for enterprise automation. A week earlier, Microsoft's industry blog declared 2026 the year agentic AI moves from bottlenecks to breakthroughs in insurance. And in the background, Allianz's Project Nemo — a seven-agent claims pipeline that clears low-complexity claims in under five minutes — quietly graduated from Australian pilot to global template. Allianz reports an 80% reduction in claim processing and settlement time in the lines where Nemo runs.

Goldman Sachs's April 2026 displacement note named the three occupations carrying the highest immediate substitution risk in the U.S. economy: telephone operators, bill collectors, and insurance claims clerks. Telephone operators have already been hollowed out. Bill collection is mid-collapse. Claims adjusters — 331,000 of them in the United States, with a median wage of $77,860 — are next.

This is a Human After Replacement (HAR) Series deep-dive. We're not asking whether AI will reshape claims adjudication. The question is which of the three displacement curves the industry actually rides — the gentle 7% attrition the Bureau of Labor Statistics modeled in 2024, the 25-30% headcount compression carriers are now privately modeling for 2028, or the catastrophic 50%+ cliff that lands if regulators sign off on closed-loop agentic adjudication faster than the workforce can rotate.

What an Insurance Claims Adjuster Actually Does

A claims adjuster sits between the policyholder and the insurer at the moment of loss. The job decomposes into a fairly stable bundle of tasks that has barely changed since the 1970s: take the first notice of loss (FNOL), verify policy coverage, determine fault and liability, estimate damages, investigate fraud signals, negotiate settlement, and authorize payout. Field adjusters add a physical-inspection layer — climbing roofs, walking flooded basements, photographing wrecked vehicles in tow lots.

The Bureau of Labor Statistics divides the occupation into two SOC codes. Claims adjusters, examiners, and investigators (SOC 13-1031) make up the bulk of the workforce: roughly 286,000 jobs at a median annual wage of $77,860. Insurance appraisers, auto damage (SOC 13-1032) is a smaller category of about 14,000 specialists who estimate physical-damage repair costs. Add in approximately 31,000 claims clerks classified separately, and the at-risk population approaches 331,000.

The BLS counted on demographics — an aging adjuster workforce and modest attrition — to drive a ~7% contraction by 2034. That projection was finalized before the agentic-AI wave of 2025-2026. It does not capture Allianz's Project Nemo, Microsoft's February 2026 industry guidance, the Lemonade-pioneered fully-automated FNOL stack, or the Allstate, AIG, and Zurich pilots that surfaced in Q1 2026 trade press. It is, in BLS's own language, a "lagging indicator" of what the labor market is about to do.

Why Claims Adjustment Is the Perfect Substitution Target

Three properties make a job vulnerable to large-language-model and computer-vision substitution: the work is digitally mediated, it is rules-bound but pattern-rich, and it generates a paper trail that trains the next agent. Claims adjustment scores at the ceiling on all three.

Digitally mediated: Modern claims travel as structured data — policy numbers, loss codes, photographs, telematics streams, repair-shop estimates, medical-bill PDFs. A senior adjuster spends 70-80% of her workday inside a claims-handling system (Guidewire ClaimCenter, Duck Creek, Sapiens) clicking through screens that an agent can navigate just as easily.

Rules-bound but pattern-rich: Coverage interpretation is governed by policy language, state insurance code, and a thick layer of case law. These are precisely the corpora foundation models ingest best. The genuinely ambiguous cases — the 5-10% involving novel exclusions or contested liability — still require human judgment. The other 90% do not.

Self-training paper trail: Every closed claim is a labeled training example. Carriers sit on decades of structured FNOL → investigation → adjudication → outcome data. An adjuster who closed 1,200 claims in 2025 generated 1,200 fully-annotated examples for the model that will replace her.

The augmentation anchors on the right side of that table are real. They are not, however, large enough to absorb 331,000 jobs.

The AI Arsenal Already Deployed on Adjusters' Desks

The most striking thing about AI in claims is not what is coming. It is what is already in production at scale.

Allianz — Project Nemo

Allianz launched Project Nemo in Australia in mid-2025 and reached full operational deployment in under 100 days. Nemo's architecture is seven specialized agents — coverage, fraud, subrogation, severity, repair-cost estimation, payment, and customer-comms — that hand off through a shared workflow controller. For low-complexity claims (food spoilage, travel delay, simple auto), the entire pipeline executes in under five minutes from FNOL submission to human payout authorization. Allianz reports an 80% reduction in cycle time and is explicitly positioning Nemo as the blueprint for other lines.

Lemonade

Lemonade's "AI Jim" claims bot has been processing renters and homeowners claims end-to-end since 2017. The carrier's public benchmark — 2 seconds from FNOL to approved payout on simple claims — is no longer experimental. Lemonade's loss ratio has converged with traditional carriers, removing the last industry argument that "you need humans to control losses."

Tractable, Mitchell, CCC Intelligent Solutions

The auto-damage-appraisal sub-segment has been substituted first and hardest. Tractable, Mitchell, and CCC each operate computer-vision systems that ingest claim photos and produce repair estimates that match or exceed field-appraiser accuracy. Major U.S. carriers — State Farm, GEICO, Progressive — now route a substantial share of glass and minor-collision claims through photo-only adjudication. The 14,000-person auto-appraiser segment is arguably already past the point of no return.

Roots, Agentech, Nurix

A new layer of insurance-specific agentic platforms — Roots, Agentech, Nurix, Hi Marley — sells "digital coworkers" directly to mid-market carriers that can't afford to build internal stacks. The pitch is uniform: 70-80% reduction in straight-through-processing time, 30-50% reduction in cycle costs, and "your adjusters can focus on complex claims." The unspoken corollary is that you need fewer adjusters to handle the complex claims that remain.

These numbers are vendor-reported and should be treated with the usual skepticism. Even discounted by half, however, they describe an industry crossing the productivity threshold at which human-headcount math stops penciling.

Anatomy of a Modern Claim: The Seven-Agent Workflow

To understand why this displacement curve is steeper than past automation waves, walk through what an agentic claim looks like end-to-end. The patterns below are synthesized from Allianz Nemo, Microsoft + Cognizant reference architectures, and the Roots / Agentech vendor stacks.

Agent 1 — Intake & FNOL: Customer voice or chat input. The agent extracts policy number, date of loss, loss type, and severity signal. It cross-references the policy in force, validates premium status, and auto-populates the structured claim record. Time: 30-60 seconds. Human adjuster equivalent: 15-25 minutes.

Agent 2 — Coverage Determination: The agent retrieves the binding policy document, parses exclusions and endorsements against the loss description, and outputs a coverage decision with cited policy language. Time: 60-90 seconds. Human equivalent: 30-45 minutes.

Agent 3 — Severity & Reserves: Based on loss type, geography, and historical claim distributions, the agent sets initial reserves. For auto claims, computer vision on damage photos produces an estimate within ~10% of body-shop quotes. Time: under 2 minutes. Human equivalent: 60-90 minutes plus a field visit.

Agent 4 — Fraud Triage: The agent runs the claim against fraud signals — provider blacklists, claim velocity, known fraud rings, image authenticity (EXIF, generative-image detection), text-pattern analysis of the FNOL narrative. Suspicious claims get routed to SIU. Clean claims proceed. Time: under 30 seconds.

Agent 5 — Subrogation & Liability: The agent reviews police reports, witness statements, and prior-claim records to determine fault attribution and identify subrogation opportunities. Time: 2-3 minutes.

Agent 6 — Settlement & Payment: For straight-through-processing candidates, the agent generates a settlement offer, communicates it to the claimant, captures acceptance, and triggers payment. Time: minutes.

Agent 7 — Customer Communications: Throughout the workflow, the agent handles claimant updates by SMS, email, or app, with escalation to a human when sentiment analysis flags frustration or when a regulatory notice is required.

A human adjuster reviews the closed file, signs off on payout above threshold, and intervenes on flagged claims. The headcount math — even at conservative agent throughput — implies that one human can supervise 30-50 simultaneous agentic claim files versus the 80-150 active claims a modern adjuster carries. The bottleneck collapses from minutes-per-task to humans-required-per-thousand-closures.

Auto and homeowners are running ahead. Casualty and liability — bodily injury, professional liability, complex commercial — remain the human preserve, but they are not safe forever.

Computer Vision Eats the Field Adjuster

If there is a sub-population to watch first, it is the 14,000-person auto damage appraiser segment. Field appraisal was, on paper, the most automation- resistant slice of claims work — it required physical presence at the loss site. Computer vision dissolved that moat in three steps.

Step one was photo-based first-loss estimation, deployed industry-wide by 2022. State Farm's Drone Pilot Program, GEICO's Photoclaim feature, and Progressive's Snapshot Claims app trained customers to photograph their own damage. Tractable and Mitchell turned those photos into estimates.

Step two was the smartphone-as-LIDAR transition. Modern iPhones and Android devices generate 3D depth maps that match what trained appraisers infer from visual cues. Repair cost models calibrated on 10+ million claims now beat field appraisers on cycle time and tie or beat them on estimate accuracy.

Step three, currently underway, is autonomous estimation for property damage — roof condition assessed from drone or satellite imagery, water intrusion mapped from interior 360° captures, structural questions escalated to engineering specialists rather than carried by generalist adjusters.

A field appraiser's career used to require a 3-5 year apprenticeship. That apprenticeship is now a 30-day onboarding for an estimator who reviews CV outputs and handles exceptions. The total-employed-population math runs in one direction.

Fraud Detection: The Justification That Sells the Board

Insurance fraud costs the U.S. industry an estimated $308.6 billion per year (Coalition Against Insurance Fraud, 2022 estimate, widely accepted). That is the single biggest line item AI vendors point to when they sell agentic claims to carrier executives. The pitch: if our model recovers even 1% of that fraud, it pays for the entire stack and the headcount reduction is gravy.

Modern fraud-detection agents combine four signal classes:

1. Network analysis: Are the same medical providers, repair shops, and claimant phone numbers showing up across claims? A pattern that takes a senior SIU investigator weeks to assemble takes the agent under a second.
2. Image forensics: Was this damage photo generated by an image model? Has it been re-used from a prior claim? Does the EXIF metadata match the claimed loss date?
3. Behavioral signals: Telematics data, claim-narrative analysis, sentiment during call recordings.
4. Cross-carrier consortium data: ISO ClaimSearch, NICB, and emerging carrier consortia feed shared fraud pattern libraries.

The genuinely sophisticated fraud rings still defeat the models, and human SIU investigators retain a strong moat. But the routine soft-fraud middle — the 70-80% of fraud volume — is exactly where the agents win.

The Regulatory Cliff

Three regulatory questions determine which displacement curve the industry actually rides. Carriers know this. State insurance commissioners know it. The NAIC's 2025 Model Bulletin on AI Use by Insurers and the EU AI Act's classification of insurance pricing and claims as high-risk both signal where the line will land. The questions:

1. Can an algorithm legally adjudicate a claim? In nearly every U.S. state, the answer today is "only with human sign-off." That sign-off is the bottleneck the agentic stack is designed to dissolve. Watch for state- by-state rule changes that move from "human in the loop" to "human on the loop" — supervisory rather than per-claim approval. New York's Department of Financial Services is the bellwether.

2. What disclosure does a denied claimant get? The EU AI Act and emerging state laws require carriers to disclose when an automated system contributed to an adverse decision. That obligation creates a class of "adjudication-explainer" roles — typically filled by senior adjusters — that provides a temporary hedge against the steepest displacement curves.

3. What is the bad-faith liability? A carrier that uses an agent to deny a meritorious claim faces extra-contractual damages in most U.S. jurisdictions. Plaintiffs' bar is sharpening its tools for AI-adjudication bad-faith litigation. The legal exposure is the single biggest reason agentic stacks still ship with conservative human-review thresholds.

A reasonable base case is fragmentation: California, New York, and Washington land restrictive; Florida, Texas, and most of the Southeast land permissive; the EU is restrictive; the UK and Asia-Pacific are mixed. Multi-state carriers will run two stacks and the headcount-math will be averaged across them.

Five-Year Displacement Timeline

The 2028 number — 22% to 32% headcount decline by year-end — is the one to watch. That is the consensus across McKinsey's insurance practice, Bain's claims benchmarks, and the Microsoft + Cognizant agentic-AI adoption study. Carriers privately model toward the higher end of that range and publicly disclose the lower end. Your bias should be toward the private numbers.

The Human Edge: What Adjusters Still Do Better

This is a HAR Series article, not a doom piece. The question is not whether humans add value in claims — they clearly do — but which humans, in which roles, at which scale.

Bad-faith and litigated claims. Once a claim is in litigation, an adjuster's role shifts to expert witness, deposition target, and litigation strategist. That work is irreducibly human and will remain so through this decade.

Catastrophe response. When a hurricane lands or wildfires sweep through, the surge in claims breaks pure-agent throughput. Field adjusters in catastrophe-response roles — mobile teams that travel, inspect, and triage — keep their seats. The total cat-response workforce shrinks because each adjuster supervises more agents, but the role is structurally durable.

Complex commercial and specialty. Energy, marine, aviation, political-violence, environmental — these lines have small loss volumes, high stakes, and stylized investigations that don't fit the pattern libraries. Specialty adjusters at Lloyd's syndicates, AIG's CAT lines, and reinsurer claims desks are not in the substitution zone.

Trauma and total-loss empathy. The adjuster who calls a parent after a fatal-collision claim, or who walks a homeowner through total-loss documentation in a fire, performs a job that is part case management and part grief counseling. Customer-experience research at Lemonade and Allianz both show steep satisfaction drops when this work is delegated to agents, even if the back-office tasks run agentically.

The 20% durable slice is real — and it is also the slice that already pays the highest wages, demands the most experience, and is hardest to enter without a 5-10 year career start in the substitutable middle. Closing the on-ramp is the structural problem the industry has not yet solved.

Survival Strategies for Working Adjusters

If you are a claims adjuster reading this, you are not in a hopeless position — but you do have a finite runway. The adjusters who navigate this transition successfully tend to make one of three moves.

Move up the complexity ladder. Specialize in casualty, complex commercial, environmental, or marine. The training pipeline is longer and the work is denser, but the substitution curve is much shallower. The adjuster who has been processing routine auto for ten years and waits to specialize until 2028 is moving against an industry-wide rotation.

Move into AI supervision. Carriers will need a new role that sits halfway between adjuster and risk-management: the "claims AI supervisor" who certifies model performance, audits adversarial cases, investigates anomalies, and signs the regulatory disclosure. This is real work, well-paid, and a natural pivot from senior-adjuster experience. Most major carriers have not yet built the role formally, which means the people who define it inside their carriers in 2026-2027 will own it.

Move into catastrophe-response or SIU. Field-deployed adjusters in hurricane, wildfire, and earthquake response retain unique value. SIU investigators on complex fraud rings retain unique value. Both roles demand re-skilling and stomach for travel or unpleasant casework, but the substitution curve is the gentlest of any role in the function.

The adjusters who do not make a deliberate move — who keep handling routine auto and homeowners claims while the agentic stack ramps — are the population the BLS curve catches first.

The runway numbers above are rough — within a 2-3 year band on either side — but the rank ordering is robust across carrier strategy decks and consultant reports.

Implications for Policyholders

If you are an insurance customer, the displacement curve has direct implications.

Cycle time will get dramatically faster on routine claims. This is the consumer-friendly headline. A homeowners claim that took two weeks in 2022 closes in two days in 2026 and arguably two hours by 2028 if your carrier runs a Nemo-class stack.

Denial rates may rise on edge cases. Agentic systems are tuned for throughput. Edge-case claims that an experienced human would have worked harder to approve get denied faster. Your appeal becomes more important. The carriers with the strongest customer satisfaction (Lemonade, Amica, USAA) are the ones investing most heavily in human escalation paths alongside agent throughput.

Fraud detection cuts both ways. Legitimate but unusual claims — classic-car restorations, custom-built homes, niche commercial — get caught in fraud-triage filters more often. If your agent flags a claim "under review" beyond a few days, escalate.

Premium implications are mixed. Carriers that successfully deploy agentic stacks see 10-15% reductions in loss-adjustment expense. Some of that flows to policyholders through pricing competition; some is retained as margin. Expect a 2-5 year lag before customer-side savings materialize, and expect them concentrated in lines where agentic substitution runs furthest first (auto, homeowners).

Carrier-by-Carrier: Where the Big U.S. Books Stand

The U.S. personal-lines market is concentrated in roughly fifteen carriers that together write 80%+ of policies. Their agentic-claims trajectories diverge sharply enough that the labor-market implications are not uniform.

State Farm — The largest U.S. personal-lines carrier (auto and home) sits in a methodical-adopter posture. Its Drone Pilot Program for roof inspections has matured; photo-claim adoption is steady. State Farm is unlikely to lead the cliff. It is also too large to absorb the competitive pressure if Progressive and GEICO ramp aggressively, which they are.

Progressive — Snapshot Claims and the broader Progressive direct stack have made it the industry's pace-setter on STP rates. Internal documents leaked into trade press in late 2025 suggested a 2027 target of 60% STP across personal auto. That number, if achieved, implies 4,500-6,500 net adjuster-role reductions at Progressive alone.

GEICO — Berkshire Hathaway's discipline shows: GEICO has been slower to deploy agentic stacks than direct-comparison rivals, but the rationale is not technological caution but loss-ratio focus. Expect GEICO to land in the middle of the pack on adjuster headcount but relatively late.

Allstate — Public statements in Q1 2026 earnings calls flagged "agentic claims architecture" as a 2026-2027 strategic priority. Allstate's "Drivewise" telematics infrastructure provides the data substrate for agent-driven settlement. The carrier's prior "QuickFoto Claim" was an early STP foothold. Watch Q3 2026 reporting.

Liberty Mutual — Its "claims-as-a-service" product line for mid-market carriers is one of the more interesting strategic pivots: selling agent-augmented claims handling to carriers that lack scale to build their own. If Liberty Mutual succeeds, it shrinks the industry adjuster headcount even faster, because mid-market in-house adjuster populations get retired during outsourcing transitions.

USAA — A unique case. Its membership-driven culture and customer- satisfaction obsession mean USAA will likely run agent throughput behind the industry leaders while preserving a heavier human-handling layer than competitors. The membership accepts longer cycle times in exchange for the human relationship. Adjuster headcount at USAA may glide more gently than the industry average.

Travelers, Hartford, Chubb — Mid-tier multi-line carriers all publicly using AI in claims; none yet at Allianz Nemo's deployment depth in the U.S. market. Mid-2027 is the likely inflection.

Lemonade, Root, Hippo — The insurtechs were always agentic. They are the proof point that closed-loop adjudication works. Their loss ratios converging with traditional carriers in 2025 was the moment that broke the "you need humans to control losses" objection.

The Reinsurance Angle: The Pressure From Above

Adjuster-headcount conversations focus on primary carriers, but the reinsurance market is the silent accelerant. Munich Re, Swiss Re, Hannover Re, and Berkshire Hathaway Re collectively underwrite tail risk for nearly every primary carrier. Their 2026 treaty renewals included explicit operational-efficiency requirements: claims operations must demonstrate measurable agentic-AI deployment as a condition of favorable reinsurance terms.

That pressure is invisible to the public but enormous in industry boardrooms. A primary carrier that resists agentic claims faces both margin compression (loss-adjustment expense uncompetitive) and capital- cost compression (reinsurance loadings creep higher). The reinsurers have decided. The primaries are following whether they want to or not.

International Perspective

The displacement curve runs at different speeds in different regulatory environments.

European Union — The EU AI Act classifies insurance claims as high-risk, requiring conformity assessments, human-oversight documentation, and transparency obligations. The compliance burden slows deployment but does not stop it. Allianz, AXA, and Generali are running production agentic claims with extensive governance overlays. Expect EU adjuster headcount declines on the order of 18-22% by 2030 — slower than the U.S. but real.

United Kingdom — Post-Brexit, the FCA has taken a lighter-touch approach than the EU. Aviva, Direct Line, and Admiral are deploying agentic stacks faster than continental peers. UK adjuster headcount declines may track U.S. permissive-state outcomes (25-30%).

Asia-Pacific — Japan's stable workforce and conservative carrier culture (Tokio Marine, MS&AD) point to gentle headcount declines. China's Ping An has been operating agentic claims at scale since ~2020 with remarkably aggressive headcount discipline. Australia followed Allianz Nemo's lead within months.

Latin America — The fragmentation of the carrier landscape and weaker regulatory infrastructure means agentic claims arrive late but, when they arrive, displace harder per claim because the human layer is already lean.

Worker-Side Response: Where the Adjuster Workforce Pushes Back

Claims adjusters are not historically union-dense, which limits the collective bargaining response. But several developments to watch:

• Licensed adjuster associations in California, Florida, and Texas have begun lobbying state insurance commissioners for human-approval requirements on adjudication above specified loss thresholds. Florida's bill, currently in committee, would require licensed-adjuster review of any claim denial above $10,000.
• The plaintiffs' bar — particularly the bad-faith litigation specialists — is the de facto worker advocate. Adjuster headcount cuts that produce wrongful-denial spikes generate the litigation exposure that dampens carriers' enthusiasm for the steepest curves.
• AI-supervisor career framing is emerging as the union-equivalent worker strategy. Senior adjusters who frame themselves as essential governance personnel ("you can't operate the agent fleet without me") are negotiating durable roles inside their carriers.

Historical Comparison: The Radiology Parallel

The closest historical analog to the claims-adjuster transition is radiology, where deep-learning systems for medical-imaging interpretation began clinical deployment in 2017-2018 and triggered predictions of imminent radiologist obsolescence by Geoffrey Hinton and others.

The radiology displacement actually played out as follows: radiologist demand increased through 2025 because imaging volumes grew faster than productivity gains, but the role compressed — radiologists today review AI-generated reads rather than making them from scratch, and trainee programs have shrunk. Compensation for radiology specialists held up; entry-level radiology jobs got harder to find.

The claims-adjuster transition is likely to track that pattern with two important differences. First, claim volumes are not growing — if anything, autonomous-vehicle penetration and risk-engineering tools are pushing them down. So the volume-growth offset that protected radiology demand does not exist. Second, the political constituency for radiologists (the medical-society lobby, hospital systems, malpractice infrastructure) is far stronger than the political constituency for claims adjusters. Less protection, less volume offset, similar mechanics — that is why adjuster headcount likely contracts more sharply than radiology did over the same time horizon.

The Industry's Strategic Pivot

The story carriers tell themselves about agentic claims has shifted in 2026 in ways worth tracking.

In 2023-2024, the pitch was "AI augments your adjusters." The industry's tone-deaf 2024 messaging — "we're hiring more adjusters than ever" — collided with the reality that adjuster postings on LinkedIn were down 28% YoY by Q3 2025.

In 2026, the pitch has become "AI handles the routine; humans handle the complex." This is more honest but understates the math: as the routine share goes from 50% to 80%, the headcount required to handle the complex residual is much smaller.

The next pitch, arriving in 2027-2028, will likely be "AI supervisors oversee agent fleets." That framing acknowledges the structural shift and creates a new (smaller) job category. It is also the pitch that unlocks the closed-loop regulatory regime carriers privately want.

For an honest strategic forecast, see my prediction on insurance claims-adjuster headcount through 2028. For a broader view of which white-collar roles are getting hit first, see the HAR Series analysis on financial analysts. For the parallel collapse in administrative back-office work, the medical-coders displacement deep-dive is the closest analog: same agentic mechanics, same regulatory tension, same 25-30% headcount math by 2028.

The Substitution Frontier

There is a specific moment in every occupation's automation history when the substitution argument flips from speculative to inevitable. For telephone operators, it was the 1960s rollout of direct-dial. For travel agents, it was the 1996 launch of Expedia. For bank tellers, it was the ATM (which curiously grew the teller workforce for two decades before slowly shrinking it). For claims adjusters, the moment is now.

What makes this transition different from the bank-teller analog is that the agentic stack does not require a parallel infrastructure buildout. There are no machines to install, no real-estate footprint to expand, no consumer behavior shift required. The customer experience is identical or better. The carrier P&L is dramatically better. The regulatory friction is real but bounded. The training data is already collected and labeled. The vendors are already shipping.

That combination of preconditions has produced exactly one prior parallel — radiology — and the radiology displacement story has played out faster than the optimistic 2016 essays predicted but slower than the catastrophic ones. Claims adjusters will probably follow that curve: faster than the BLS thinks, slower than the worst Twitter takes claim, with most of the action concentrated in the 2026-2030 window.

The adjusters who specialize, supervise, or specialize-and-supervise will keep meaningful careers. The adjusters who continue handling routine auto and homeowners claims as if the seven-agent pipeline is not coming for their queue will discover that the queue does not need them. Project Nemo and its cousins are not the future. They are the present, mid-rollout, in your carrier's claims center, right now.

Further Reading

• How AI Will Replace Financial Analysts: The Three Futures of Wall Street
• How AI Is Replacing Medical Coders: Revenue Cycle and Autonomous Coding Displacement
• Prediction: Insurance Claims-Adjuster Headcount Will Decline 25%+ by Q4 2028
• Allianz Project Nemo and the Substitution Frontier — News Analysis
• The Last Claim — Short Story