AI TPA Halves Claims Time — Legacy TPAs’ Response Still Unclear

Funding and Platform Launch

ClaimSorted announced a $13.3 million seed round on Monday, led by Atomico with Eurazeo, Y Combinator, Firstminute Capital, Start Ventures, and a network of insurance veterans participating. The round follows a $3 million pre-seed last October backed by Firstminute, Y Combinator, Precursor Ventures, and Transpose Platform. At just over a year old, the company already works with more than 20 insurers across the U.S., U.K., and EU, serving tens of thousands of policyholders and managing an estimated $3 million in 2024 revenue.

The origin story is blunt. Pavel Gertsberg and German Mikulski, British citizens who arrived as teenagers from Russia and Belarus, founded Fluffy, a pet insurance startup that reached 20,000 customers. Gertsberg said in a Forbes interview that they automated roughly 70 percent of back-office processes with AI, but the third-party administrators handling claims remained a bottleneck. "Many companies are born out of different reasons. Ours was born from frustration, borderline anger," Gertsberg told Forbes. He added: "Ask anyone in insurance where the weakest link is, and most will point to claims TPAs. We saw the claims process being painfully slow. Mistakes made by TPAs that should've been caught early ended up eating our entire profit margin."

So they shut Fluffy down and built ClaimSorted, describing the platform as "Claims TPA 2.0." The platform operates as an AI-native third-party administrator, not a SaaS tool. It automates fraud checks, compliance, and claim decision-making while pairing AI agents with a claims team drawn from Hiscox, Lemonade, AXA, Hartford, and Liberty Mutual. Gertsberg, who previously built transaction systems at Deutsche Bank processing over $100 million, told Beinsure the approach is a "next-gen service layer that directly replaces sluggish TPAs" rather than another software pitch.

Atomico, which backed Stripe and Klarna, frames the bet around a category it calls "Service-as-a-Software," referring to AI-native businesses that own the full stack from automation to service delivery. The global TPA market was valued at $390 billion in 2024 and projected to reach $812 billion by 2032, per Atomico's October 13, 2025 investment memo, yet core processes remain dominated by manual coordination across fragmented systems. Atomico's memo adds that up to half of the insurance workforce is set to retire by 2037, one in three policyholders reports dissatisfaction with claims experiences, and while 80 percent of claims executives believe AI can add value, fewer than half rate their organizations as advanced in applying it.

The fresh capital targets expansion of a roughly 20-person team toward 50, weighted toward engineering and claims operations across New York, London, Dallas, and remote hubs in Ireland and the U.K. Gertsberg's stated goal: 100 million policyholders worldwide. "We don't celebrate a fundraise once we close," he told Forbes. "We didn't have any company drinks, didn't do anything. It's not the goal. The goal is the customer and what we deliver."

Where the Speed Comes From

The numbers that moved Atomico's investment committee were cycle-time compression of more than 50 percent, paired with a 10-point Net Promoter Score uplift at major carrier partners. Those two figures, drawn from Atomico's October 13, 2025 investment memo, are the cleanest window into what ClaimSorted actually sells: claims that close in a fraction of the time a legacy TPA takes, processed so cleanly that policyholders notice. Atomico put it in plain words: clients describe the platform as "faster, more transparent, and more adaptable" than incumbent providers.

Forbes reported in October 2025 that ClaimSorted settles claims three times faster than rivals, at lower cost per claim and with higher customer satisfaction. Beinsure, writing the same week, quoted the founders making the same claim with sharper language: a five-star policyholder experience while saving insurers millions. The earliest published data point, from Fintech Global's October 2024 coverage of the $3 million pre-seed, described a hybrid human-and-AI workflow that ran claims "up to ten times faster than traditional methods" while cutting payout errors by more than 20 percent.

The mechanics sit inside a larger shift Atomico's analysts called "Service-as-a-Software." Large language models and multi-agent systems now handle the unstructured, communication-heavy tasks that traditional robotic process automation could not: reading claim emails, populating forms, coordinating between adjuster, body shop, and policyholder. That frees ClaimSorted's human adjusters to focus on indemnity accuracy and complex cases rather than data entry. Atomico wrote that because ClaimSorted owns the full service lifecycle (pricing, user experience, and outcomes), every claim processed strengthens its dataset, optimizes its models, and improves future performance. In a market where 80 percent of claims executives believe AI can add value but fewer than half rate their organizations as advanced in applying it, per Atomico's data, that compounding loop is the moat.

The leakage story is also a service-quality story. Atomico cited industry data showing one in three policyholders reports dissatisfaction with claims experiences, mostly due to slow resolution times. A 10-point NPS uplift, layered on top of a 50-percent-plus cycle-time cut, is what happens when a bottleneck gets removed at the workflow level rather than the marketing level.

What the Incumbents Are Doing

The seed round did more than fund a startup; it put the incumbent TPA complex on notice. Sedgwick, Crawford & Company, and the broker-TPA hybrids that dominate the market have spent the past year layering AI onto legacy stacks, restructuring leadership, and acquiring niche capabilities to blunt the speed and leakage advantages an AI-native model promises.

Sedgwick, the largest U.S. TPA by volume, has pursued the most visible two-track response. Its "Darwin" claims management system, a cloud-based platform that incorporates AI and robotics to automate registration, triage, and reporting, was already in market before ClaimSorted's seed round. In February 2026, in a TechTarget feature on agentic AI readiness, Sedgwick chief transformation officer Vishy Padmanabhan described a "Quad model" aligning Business, Operations, Technology, and Go-to-Market teams to sequence AI deployment across multistep workflows. "We are working closely with our claims examiner front line to redesign the workflow," he said, adding that some clients are now discussing initiatives to reimagine aspects of claims handling. On the talent side, Sedgwick global head of talent acquisition David Reed told HR Executive in December 2025 that colleagues who stay and grow internally fill nearly a quarter of open positions, referrals account for another 9 percent, and returning alumni represent 6 percent of hires. The firm's Leadership Academy includes AI-assisted conversation simulators for adjuster training.

Crawford & Company has taken a more defensive posture. The firm has continued to emphasize operational resilience and capital returns to shareholders; it scheduled its second-quarter 2026 earnings call for late July. Public commentary has emphasized disciplined balance-sheet management over platform overhaul, suggesting Crawford is betting that its scale and global footprint will retain carrier contracts while it incrementally modernizes.

Aon, which operates as both broker and TPA, launched Aon Claims Copilot in November 2025, per Insurance Business, debuting first in Germany before a global rollout through 2026-27. The platform combines the firm's Broker Copilot and risk-analysis tools for claims resolution and analytics, targeting its 1,800 claims professionals across more than 50 countries and over 20 product lines. Aon's play leans on its broker-side data advantage (loss history, policy wording, market placement) to train models that sit upstream of the TPA handoff.

Gallagher Bassett, the TPA arm of Arthur J. Gallagher, chose acquisition over build. In 2025 it acquired Reck & Co., a specialist in complex liability and captive claims, adding niche expertise that AI-first entrants lack. The deal doubles as a talent grab: Reck's adjusters bring domain depth that pure-play tech firms struggle to replicate, and Gallagher can layer its own automation tools on top.

The pattern across incumbents is clear: add AI as a feature layer, protect existing carrier relationships with leadership and talent moves, and acquire specialized capabilities that an AI-native startup would take years to build. What none have done yet is rebuild the core claims operating system from the data layer up. That architectural gap is the opening ClaimSorted and its peers are exploiting.

Carriers and MGAs Are Signing Up

ClaimSorted's commercial traction is widening faster than its headcount. The startup says it already works with more than 20 insurance carriers across the U.S., U.K., and EU and handles tens of thousands of policyholders through its AI-native claims platform. Forbes reported the 20-plus-carrier figure on October 13, 2025, and Beinsure confirmed the same scope nine days later. The claim is anchored by a list of insurers the company has hired from, not sold to, including Hiscox, Lemonade, AXA, Hartford, and Liberty Mutual, which suggests the relationships sit on both sides of the table: carriers handing ClaimSorted claims volume, and alumni of those same carriers seeding the team.

That mix is doing the selling for ClaimSorted. Co-founders who built, then shut down, a pet-insurance startup after watching third-party administrators "slow everything down, botch consistency, and in the end gut profitability" now lead a roughly 50-person operation with claims veterans from the same firms they're courting as customers. The Forbes profile notes ClaimSorted is headquartered in both New York and London, "a reflection of how crucial the U.S. market is to its growth, even as the U.K. remains home for its cofounders." That dual footprint lets the company staff U.S. claims desks in Eastern Time while keeping product and engineering in London, a structural advantage when carriers decide which TPA can handle their highest-volume books.

Multiple secondary reports put ClaimSorted's premium under management above $200 million, with the fresh capital earmarked to push toward a $1 billion target. For context, the company employs only around 20 people as of August 2026 per third-party tracking, having grown from roughly 50 at the time of the seed round. This unusually high revenue-per-employee ratio signals how much of the workflow is now handled by automated agents rather than adjusters.

MGA partnerships are the second growth vector. Managing general agents, which bind and service policies on behalf of carrier partners without large in-house claims teams, are an obvious fit for an outsourced AI-first TPA, and ClaimSorted has been positioning the platform as end-to-end infrastructure rather than a point tool. Beinsure quotes the founders saying carriers "don't want another ambiguous 'AI claims platform'. They want an end-to-end service that removes the headache, serves policyholders at their most critical moment, and protects margins." That pitch is landing with MGAs in particular because they lack the legacy claims stacks of the top 20 carriers and can onboard faster.

The competitive dynamic is sharpening. Legacy TPAs such as Sedgwick and Crawford & Company are losing both volume and margin on the small-to-mid commercial and specialty books ClaimSorted targets, while AI-first competitors like Shift Technology have moved to defend the same ground with their own agentic claims products. Carriers running parallel pilots with ClaimSorted and incumbents are quietly re-bidding TPA contracts, and several are splitting claims volume between a legacy administrator for complex liability and a new AI-native TPA for high-volume, low-severity property and pet lines. If current adoption curves hold, the $1 billion managed-premium milestone becomes a question of which carriers move first, not whether the model works.

The Hiring Surge

ClaimSorted came out of Y Combinator with roughly 20 people, hit $3 million in 2024 revenue, and within twelve months of its founding had grown to a 50-person team working across New York and London. That jump, roughly two-and-a-half times headcount in a single year, is the clearest signal that the seed money is going straight into payroll. Forbes, reporting on the October 2025 fundraise, said the fresh capital funded expansion of the team, "particularly in its engineering and claims operations." For a startup that started by automating document verification, anomaly flagging, and fraud detection, that emphasis maps directly to two hiring profiles: machine-learning engineers who can ship production models, and seasoned claims operators who can supervise them.

The open roles confirm where the money is flowing. ClaimSorted's job board lists on-site positions with the following salary bands, per Ashby postings:

Those bands sit comfortably above the median US claims-adjuster wage and overlap with senior software-engineering compensation at larger tech employers, a deliberate signal that ClaimSorted is pricing claims operations as a technical discipline, not an administrative one. The dual-currency listings also reflect the transatlantic hiring model: London hires handle UK and EU carriers, while New York and Dallas staff cover the US book.

The pressure runs the other way too. As legacy TPAs such as Sedgwick and Crawford & Company feel the squeeze from AI-native competitors, they are being forced to compete for the same scarce talent: engineers who understand both insurance workflows and production machine learning. Shift Technology's recent hire of a former Ocrolus executive to lead global engineering, reported at the time of Shift Claims' launch, shows the same dynamic playing out on the platform side: established AI-claims vendors are paying top-of-market to keep their technical edge. For now ClaimSorted's scale lets it pay like a tech company while selling to insurers at TPA margins, a combination that is hard for incumbents to match without restructuring their own cost bases. Whether that advantage holds past the 100-person mark is the open question every AI-native claims shop is about to answer.

Who Else Is Moving Into the Same Lane

Shift Technology didn't wait for ClaimSorted to prove out the AI-first TPA thesis before pivoting its own product line. In September 2025, the Paris-based firm launched Shift Claims, a suite built on agentic AI that handles four jobs: assess complexity, classify and prioritize cases, advise handlers, and automate the rest of the lifecycle. That's a tighter claim-by-claim scope than Shift's prior fraud-detection roots, and a tell that the company now sees itself competing for the same operations budget ClaimSorted is going after, not just selling point tools into legacy stacks.

The product push comes with hard numbers for early adopters. Shift says customers running Shift Claims have cut claims losses by 3 percent, sped handling by 30 percent, hit a 60 percent automation rate, and crossed 99 percent accuracy in assessment. AXA Switzerland is the named early adopter. The shift matters strategically: rather than positioning as a fraud module that sits next to a TPA, Shift is now selling the automation of the claim itself, directly adjacent to what ClaimSorted calls its end-to-end TPA. Shift's chief scientist and chief product officer, Eric Sibony, drew the line himself: rules-based systems, he said, "struggle to handle the nuance and variability of most real-world cases."

That repositioning pressures every other AI-adjacent claims vendor into a choice. Tractable, which has long focused on visual damage assessment for auto insurers and recently settled its five-year patent fight with CCC, has its own straight-through processing partnerships. GEICO is already using Tractable to speed accident recovery. The strategic question for Tractable, and for CCC, Mitchell, and the rest of the visual-AI cohort, is whether to keep selling best-of-breed assessment modules into carriers and TPAs, or to follow Shift's lead and try to own more of the claim.

Underwriting the whole shift is an enterprise IT market that has stopped debating whether to buy agentic AI and started debating how fast. A Deloitte survey of 100 US health care technology executives found 61 percent of organizations already building or budgeting for agentic AI initiatives and 85 percent planning to increase investment over the next two to three years; 98 percent of executives expect at least 10 percent cost savings, with 37 percent expecting savings above 20 percent. Gartner reports that organizations with successful AI initiatives invest up to four times more in data and analytics foundations than peers, and that the highest-maturity organizations see up to 65 percent greater business outcomes as a result.

The result is a market where incumbents, point-tool vendors, and AI-native TPAs are all converging on the same proposition: end-to-end claim automation, sold on cycle-time and leakage metrics, priced against adjusters' fully loaded cost. Shift's product launch, Tractable's funding, and ClaimSorted's seed round are three faces of that convergence.

What $1 Billion in Premium Actually Requires

Hitting a billion dollars in premium under management would put ClaimSorted in a category well above where any AI-native TPA currently operates. Crossing that threshold would require onboarding carriers and managing general agents that today still direct most of their claims volume to Sedgwick, Crawford & Company, or in-house desks.

The arithmetic points to a multi-year build. ClaimSorted disclosed more than $200 million in premium under management in connection with its seed round. Reaching $1 billion would mean roughly fivefold growth in premium flow, a trajectory that depends on three milestones the public record hints at but does not yet confirm.

The first milestone is geographic. The Ashby postings show a US-UK split, with dollar-denominated roles in New York and Dallas alongside sterling roles in the UK. Closing the geographic gap to a majority-US book is the clearest path to $1 billion.

The second milestone is line-of-business expansion. The Y Combinator profile describes an "AI-first TPA" that automates those same checks and decisions across multiple lines rather than a single peril. Expanding from low-frequency, high-severity lines into high-frequency personal auto and homeowners is where premium volumes compound quickly. Shift Technology offers a template: it launched Shift Claims and has continued to extend its collaboration with AXA Switzerland as an early adopter. If ClaimSorted is to hit $1 billion before Shift absorbs the same carrier relationships, the line-of-business footprint has to widen before competitors lock in renewals.

The third milestone is operational scale. Crawford & Company announced a board-approved dividend increase in a recent earnings cycle. Sedgwick appointed Vishy Padmanabhan as chief transformation officer to lead its agentic AI agenda. Those moves signal that legacy TPAs are raising their game precisely because new entrants are eating into their renewal base. ClaimSorted's $1 billion target effectively becomes a race against incumbents who can absorb fixed costs across far larger books.

The story also has explicit limits worth naming so readers know where to stop. The public record contains nothing on the architecture of ClaimSorted's underlying AI model, no descriptions of training data, parameter counts, or vendor stack. There is also no material on regulatory change: no mention of state-by-state licensing progress, no NAIC filings, no Department of Insurance approvals. Any claims about which states the company is licensed in, how its model handles biased claims decisions, or whether AI-driven claim denials are subject to specific consumer-protection rules would be unsourced speculation.

What the record does support is a directional read: the $13.3 million seed, the $200 million-plus premium under management, the senior hiring at the published bands, and the public proof points together describe a company scaling up, not one that has arrived. Whether the bottleneck they set out to replace proves more fragile than the claims stack it replaces is the bet every insurer re-bidding a TPA contract this year is quietly making.

Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.