AI is accelerating in insurance – are you ready?

For decades, the insurance industry moved at its own deliberate pace - a world of actuarial tables, paper applications, and adjusters dispatched to inspect fender-benders in suburban driveways. The joke in technology circles was that insurers would be the last to change. Nobody is laughing now. 

Artificial intelligence has not crept into insurance. It has crashed through the door. In the past 18 months alone, a major British insurer cut the time to resolve a complex liability claim by 23 days. A German giant built and deployed a seven-agent AI claims system in under 100 days. An American insurtech automated 55 percent of its claims from start to finish, settling some in seconds. And one of the world's largest carriers - Nationwide - announced it was committing $1.5 billion to technology, with 20 percent explicitly earmarked for AI. 

Read next: Is the future of AI for insurance? – don’t miss it 

The question the industry asked three years ago - "Can we trust this?" - has been answered. The question now is far more urgent: How fast can we scale it before competitors pull too far ahead? 

"AI is now central to how insurers operate, affecting underwriting, claims, customer experience, and fraud detection," said Guy Gresham, a global capital markets board advisor and former director of investor relations at BNY. The shift, he added, is turning carriers from product providers into something closer to real-time risk managers. 

Is It Accelerating? Yes - and Faster Than Most Expected 

The numbers make the case plainly. The AI insurance market was valued at $8.63 billion in 2025. It is projected to reach $59.5 billion by 2033, growing at a compound annual rate of more than 27 percent. Industry spending on AI is expected to grow by more than 25 percent in 2026 alone. Accenture found that 86 percent of insurance organizations - regardless of size - plan to increase AI spending this year, with generative and agentic AI topping the investment list. 

Consumer sentiment is shifting just as sharply. In 2025, only 20 percent of policyholders said it was a good idea for their insurer to use AI to improve services. In 2026, that figure nearly doubled, to 39 percent, according to Insurity's annual AI in Insurance Report. Resistance is easing. The share of consumers who said they were less likely to buy from an AI-using insurer fell from 44 percent to 36 percent in a single year. 

A Grant Thornton survey of 950 executives, conducted earlier this year, found that 52 percent of insurance leaders are already reporting AI-enabled revenue growth, while 62 percent say the technology is improving their decision-making. More telling still: a McKinsey analysis found that early AI leaders in insurance are generating roughly six times the total shareholder returns of their AI-laggard peers. Six times. The gap is not narrowing - it is widening. 

What is driving the acceleration? Three forces are colliding at once. The technology itself is maturing faster than almost anyone predicted, moving from simple automation to genuinely autonomous decision-making. The competitive environment is intensifying, with AI-native insurtechs forcing established carriers to respond or cede ground. And consumer expectations - shaped by daily experience with AI tools outside insurance - are now running well ahead of what most carriers can deliver. The pressure is coming from every direction simultaneously. 

That said, the acceleration is uneven, and the gap it is creating is real. A BCG study found that only 7 percent of insurance AI initiatives move beyond pilots, and a separate survey found that 82 percent of insurers believe AI will define their industry's future, while only 14 percent have fully integrated it into their financial operations. This is what analysts are now calling the "operational divide" - and it is hardening by the quarter. As Christopher Frankland of InsurTech360 told Insurance Business, the hardest problem is not building an AI pilot - it is getting it out of the lab. "When you think about AI pilots, the question is: how do you take something from POC to production and scale it effectively? That's perhaps the toughest piece of this whole journey," he said. 

Underwriting: From Days to Seconds 

No function has felt the pace of change more acutely than underwriting. The core task - assessing risk accurately and pricing it correctly - turns out to be precisely the kind of pattern-recognition problem that machine learning excels at. What used to take hours of document review is increasingly a matter of seconds. 

Zurich North America is one of the clearest examples of a carrier moving at speed. In late 2025, Zurich integrated AI-powered aerial imagery and roof-condition scoring from Nearmap directly into its U.S. Middle Market underwriting platform, giving underwriters access to frequently updated, high-resolution property intelligence that on-site physical inspections often cannot match. At the same time, Zurich's teams deployed Sixfold, an AI tool that distills complex submission packets into structured summaries, compressing hours of manual document review into minutes. By December 2025, Sixfold had processed more than one million underwriting submissions across more than 40 lines of business, achieving an 89 percent average user adoption rate. 

Read next: How Novacore embeds AI across underwriting to sharpen profitability 

The Hartford is taking the concept even further - moving AI out of the underwriting office and into the field. Dan Campany, who leads risk services at the carrier, told Insurance Business that proliferating IoT and telematics devices are allowing insurers to understand risk as it unfolds in real time. "That allows us to understand risk in the moment," Campany said. In commercial auto, telematics and in-cab cameras detect drowsiness and distracted driving, triggering immediate alerts before an accident occurs. In property, water sensors flag leaks early. "Whether you find and mitigate the leak in a minute, an hour, a day, or a couple of days matters a lot," he said. AI then makes sense of the data at scale. "AI gives us the ability to find patterns in that granular data, analyze that data, and then focus the human beings on the things that are the outliers," Campany added. The result is a shift from reactive payer to proactive risk partner - a fundamental redefinition of what an insurer actually does. 

These tools are enabling what practitioners are now calling "continuous underwriting" - a model in which pricing and risk exposure adjust dynamically rather than at annual renewal. Insurers are moving from static contracts to living ones: policies that evolve alongside the insured's actual behavior and circumstances. Embedded insurance - travel coverage bundled at the flight booking, e-bike protection at the rental checkout - depends on this kind of real-time risk assessment. Legacy systems simply cannot do it. AI-native platforms can. 

Climate risk is adding urgency to the underwriting overhaul. Legacy models built on historical weather data are breaking down as extreme events grow more frequent and severe. As Megan Kuczynski, founder and CEO of ClimateTech Connect, told Insurance Business, real-time tools have moved from optional to foundational. "Technology, the ones that you just cited and the ones that are emerging, are really critical to precision underwriting," Kuczynski said. California wildfires and severe convective storms made the first half of 2025 the second costliest on record. Carriers that cannot reprice risk dynamically are not just leaving money on the table - they are absorbing losses they never should have written. 

Claims: The 23-Day Gap and the 3-Second Settlement 

The claims experience is insurance's moment of truth - the test that determines whether a customer stays or walks. For years, that test produced failing grades. Delays, opaque decisions, and the sense of fighting for what one was owed drove complaints and attrition. The AI transformation of claims is now rewriting that story, and in some cases the results are startling. 

Consider the contrast between two very different carriers, both getting it right in very different ways. 

Lemonade, the New York-based insurtech built from the ground up around AI, has its claims bot, AI Jim, handling first notices of loss for 96 percent of claims without human intervention. As of year-end 2025, 55 percent of all Lemonade claims are fully automated from start to finish. In the most straightforward cases, a claim is filed, assessed, and paid in seconds. Sean Burgess, the company's chief claims officer, told Claims Journal that Lemonade's pet product handles more than half of claims instantly, with a net promoter score above 75 - "unheard of in the claim space." The company operates at roughly 2,300 customers per employee, a ratio that would be structurally impossible for any carrier relying on traditional manual processing. 

Read next: Everyday AI use rises, but P&C customers still want humans on the hook – Insurity 

Then there is Aviva, the U.K.'s largest general insurer. Aviva undertook a wholesale AI transformation of its motor claims operation in partnership with McKinsey's QuantumBlack unit, deploying over 80 AI models across every stage of the claims lifecycle. The results were disclosed to investors: liability assessment time for complex cases fell by 23 days, claims routing accuracy improved by 30 percent, customer complaints dropped by 65 percent, and net promoter scores climbed more than sevenfold. The company told investors the transformation saved more than £60 million - roughly $82 million - in 2024 alone. 

Allianz added a third model with Project Nemo, launched in Australia in July 2025. Seven specialized AI agents handle coverage verification, weather validation, fraud screening, payout calculation, and audit to process food spoilage claims arising from storm-related power outages. Resolution time dropped by 80 percent, from several days to hours or minutes. A human professional approves every payout. Nemo was built and deployed in under 100 days. Allianz is now extending the framework to travel delays, straightforward auto claims, and other high-frequency lines. 

Read next: CFC pilots agentic AI system that underwrites in seconds 

The broader trend connecting all three is what the industry calls "straight-through processing": automated intake, automated damage assessment, automated payment triggers, with human judgment reserved for cases that genuinely require it. But as José Luis Bernal, chief digital, data and innovation officer at MAPFRE USA, told Insurance Business, speed is only part of the story. "AI is bringing much more predictive predictability into the premiums," Bernal said. "It can be done much more accurately and, with a new technology, much more firm and being explainable to the end customer." The transparency dimension matters as much as the velocity - customers who understand how their claim was handled are far less likely to dispute it. 

The $308 Billion Problem That AI Is Starting to Solve 

The fraud numbers are staggering. Insurance fraud costs the United States an estimated $308.6 billion annually, according to Forbes Insurance Fraud Statistics for 2025. That is not a rounding error; it is a structural drain on the industry that traditional detection methods have never been able to plug. 

The reason is straightforward. Static rule-based fraud detection was always playing catch-up. Sophisticated fraud rings learned the rules and worked around them. AI is changing the equation. Allianz deployed an AI system called Incognito across its motor, home, and new application lines. The system analyzes at a pixel level - identifying distortion and manipulation in submitted images, videos, and documents that no human reviewer would consistently spot. By Allianz's estimates, it led to a 29 percent increase in fraud detection rates. 

Other carriers have pushed AI fraud detection even earlier - to the moment of underwriting. AI-powered vehicle image recognition now scans photographs submitted during policy inception for pre-existing damage, tagging it before coverage attaches. When a claim arrives for that damage later, the flag is automatic. 

As Bernal noted, effective fraud detection is not a single-algorithm problem - it requires a blend of machine learning, graph databases, and shared industry datasets. "Right now, those third customers are paying the premium of the fraudulent customers," he told Insurance Business. "Thanks to AI, now we can detect much more accurately." Real-time fraud scoring now operates at many carriers on every incoming claim, routing suspicious cases to investigators while clean claims flow through without delay - a combination that catches more fraud while simultaneously accelerating the experience for honest claimants. 

The Governance Gap: Speed's Shadow 

Speed without accountability is a different kind of risk. As the pace of AI deployment accelerates, the governance infrastructure required to manage it is struggling to keep up. 

Grant Thornton's survey found that 44 percent of insurance executives said governance or compliance challenges had contributed to AI project failure or underperformance. Only 24 percent said they were confident their AI controls could survive an independent audit today. Regulators are moving to close that space. New York's Department of Financial Services enacted Circular Letter No. 7 in July 2024, requiring insurers to establish governance frameworks and explain clearly how AI factors into underwriting and pricing decisions. As of early 2026, 23 states and Washington, D.C. have adopted the NAIC's model bulletin on AI use in insurance, and a national AI evaluation tool is being piloted across 12 states. 

The challenge of algorithmic bias is not theoretical. Poorly designed models can perpetuate the discriminatory patterns embedded in historical data, denying coverage or charging elevated premiums to underserved populations in ways that are genuinely difficult to detect. PwC's 2025 Responsible AI Survey found that 58 percent of executives believe responsible AI practices improve ROI - a recognition that governance is not just a compliance cost but a competitive asset. 

The accountability question is equally unresolved. When an AI system recommends a coverage denial that turns out to be wrong, who bears responsibility? Aviva's approach - AI recommendations, human decisions - is widely cited as a model. But even Aviva required years of cultural and organizational work to make that accountability meaningful in practice. For carriers rushing to scale, the temptation to treat governance as an afterthought is a genuine danger. As Lynn A. O'Leary, global COO of Intact's specialty lines group, told Insurance Business, the technology itself is rarely the hardest part. "The biggest challenge is the speed of adaptation and adoption," O'Leary said. "You're asking businesses and individuals to work in different ways, in new ways, and to get out of their comfort zones." The carriers that get governance right are the ones that bring their people along from the start - not the ones that deploy first and manage the fallout later. 

The Workforce: New Jobs, New Anxieties, New Demands 

Perhaps no issue generates more internal anxiety than the question of what AI means for the people who do the work. The concern is not abstract, and the demographics make it sharper still. The share of insurance professionals aged 55 and older has grown by 74 percent over the past decade. Within the next 15 years, approximately half the current workforce is expected to retire, leaving an estimated 400,000 positions unfilled. AI is partly a response to that looming gap. It is also accelerating the transformation of the jobs that remain. 

The most instructive evidence comes from the carriers that have gone furthest. At Aviva, deploying 80 AI models across claims did not eliminate the workforce - it transformed it. Adjusters moved away from data assembly and routine routing toward the complex, contested, and emotionally charged cases where human judgment genuinely matters. Employee engagement scores more than doubled. At Zurich, underwriters using Sixfold described the experience as liberating: hours of document assembly compressed to minutes, leaving time for the analysis and relationship-building that experienced underwriters actually came to do. 

At Lemonade, the implications run deeper still. Sean Burgess, the company's chief claims officer, has been blunt about what the future looks like. "One day in the claims space," he told Claims Journal, "you may be a frontline claims adjuster, but in the future, you could be an artificial intelligence trainer, or an artificial intelligence supervisor." 

Christopher Frankland of InsurTech360 frames this as the rise of the "bionic" workforce - not humans replaced by algorithms, but humans supercharged by them. "I look at the bionic agent as this future version of the workforce where people aren't being replaced by AI and technology, but are being augmented by intelligence across every step of the insurance journey," Frankland said. "It's really the idea of supercharging how people work today using the right tools at the right time." The carriers that are getting this right are not treating AI as a headcount reduction tool. They are treating it as a capability multiplier - and building workforces that know the difference. 

Grant Thornton's survey found that only 7 percent of insurance executives believe their workforce is fully ready. Skills requirements for AI-exposed roles are evolving 66 percent faster than other fields, according to PwC's Global AI Jobs Barometer. The challenge is not simply retraining - it is reimagining what insurance jobs look like when the routine cognitive work has been substantially automated. 

What Comes Next: Agentic AI and the End of Annual Renewal 

The frontier is already visible, and it is moving fast. The next phase of AI in insurance is agentic AI: systems of specialized agents capable of independently planning, deciding, and executing complex, multi-step workflows with minimal human intervention until the moments that genuinely require it. 

Allianz's Project Nemo is an early, deliberately bounded version of this. Zurich's ambitions run considerably further. Its Agentic AI Hyperchallenge in 2025 drew more than 1,000 participants from 40 countries, producing 218 prototypes across 17 use cases, with five now moving into production - including systems for travel claims processing, motor liability assessment, and internal knowledge management. At the industry's major technology conferences in Q4 2025, executives consistently described agentic AI not as a future possibility but as the immediate next deployment priority. 

Underlying all of it is a more fundamental shift in what insurance actually is. The annual renewal cycle - a structural feature of the industry for more than a century - may give way to continuous, dynamic coverage that adjusts in real time as risk changes. Policies that learn. Premiums that move. Protection that follows the insured through daily life rather than sitting in a filing cabinet until something goes wrong. 

"Risk will be priced in real time," said one insurance executive quoted in Insurance Thought Leadership's mid-2025 industry survey. "Protection will be embedded, not bolted on." 

That vision is not uniformly available today. Legacy data architectures, regulatory complexity, and the hard work of organizational change all impose real constraints. But the constraints are loosening - faster, it now appears, than even the optimists expected. 

The operational divide will define the next decade. Aviva's claims transformation, Zurich's underwriting overhaul, Allianz's agentic pilots, Nationwide's $1.5 billion commitment, and Lemonade's AI-native model are not isolated bets. They are the leading edge of an industry in rapid motion. The carriers pulling ahead are not doing so by running better pilots. They are doing so by turning pilots into production, production into culture, and culture into competitive moats that grow harder to cross every year. 

The machines are not replacing the industry's judgment. They are, however, profoundly reshaping where that judgment is required - and what it will take to exercise it well.