Key Points
- Rising claims severity and slowing premium growth are putting greater pressure on carriers to reduce leakage, accelerate cycle times and improve litigation triage within existing operations.
- Claims AI must provide explainable decisions, complete audit trails, bias testing and vendor oversight to meet expanding regulatory expectations and move successfully from pilot to production.
- Effective claims transformation automates rule-based work, preserves adjuster capacity for complex cases and orchestrates workflows across existing core platforms without requiring a costly system replacement.
US P&C carriers spent the last two years being sold “AI-powered claims transformation” as a productivity story: copilots, summarizers and point tools bolted onto the same stop-start adjuster workflow. Faster typing, same waiting. In 2026, that framing has run out of runway. Premium growth is decelerating while severity keeps climbing, capacity cannot be hired at any price, and claims AI itself has become something regulators actively examine. For the Chief Claims Officer, COO and Chief Data & AI Officer facing a board this year, the question is no longer whether AI belongs in claims. It is whether the claims automation already in place can survive an audit.
Why Is Claims Now Where the P&C Margin Problem Lives?
US P&C direct premium written continues to grow, but the pace is slowing even as casualty severity is not. Nuclear verdicts, jury awards of $10 million or more, rose 52% in 2024 to a record of roughly 135 cases, and casualty lines booked $15.8 billion of adverse prior-year development as a result. Aon’s own market commentary now calls this “litigation abuse” rather than social inflation, a distinction that frames the pressure as structural rather than cyclical (Aon, 2026 US P&C market outlook). A cyclical problem eases when the market turns. A structural one has to be managed out of the operation itself, and claims is where that operation lives.
That reframes the mandate for claims leaders in 2026. Growth will not fund its way out of severity pressure, which puts cycle time, leakage and litigation triage squarely on the same dashboard as loss ratio.
Where Does Claims Leakage Actually Come From?
Leakage is the one number in this equation carriers can actually control, because it is operational rather than actuarial. EY’s claims quality assessments put leakage at 7 to 14% of total claims spend, driven by inconsistent adjudication, weak documentation discipline, late litigation triage and poor vendor and expense control, not underwriting or pricing decisions made years earlier (EY, 2026 Claims Quality Assessment). That distinction matters for a board conversation: rate and appetite take quarters to move a loss ratio, while fixing adjudication consistency and litigation triage can move it inside a single accident year.
Most of that leakage accumulates in the dead time between steps, the wait for a document, a coverage check or a handoff, rather than in any single decision being wrong. That makes claims leakage less a data problem than a process design problem.
Why Can’t Carriers Just Hire Their Way Out of the Capacity Gap?
Talent attraction and retention is now carriers’ top global business challenge, with acute shortages specifically in claims management, specialized case management and legal and compliance roles, according to a 2026 survey of 250 US, UK and Australian carrier leaders (Gallagher Bassett, Carrier Perspective 2026). This is not a temporary staffing cycle. It is a structural gap between arriving claims volume, including catastrophe surge, and the experienced adjusters available to handle it, and every carrier is competing for the same shrinking pool of specialists.
That reality changes the question from “how do we hire more adjusters” to “how do we make sure the adjusters we have are only working the claims that actually need their judgment.” Volume growth and CAT surge now have to be absorbed by AI paired with right-shored expertise, not headcount expansion alone.
Why Has Claims AI Become a Governance Question in 2026?
This is the shift that changes procurement conversations more than any other. Twenty-nine US jurisdictions now impose AI governance expectations on insurers: 25 states plus Washington DC have adopted the NAIC AI Model Bulletin, while California, Colorado, New York and Texas run their own frameworks (National Association of Insurance Commissioners, AI Model Bulletin adoption tracking, 2026). The NAIC’s AI Systems Evaluation Tool pilot, running across 12 states through September 2026, is already feeding into market conduct exams, and carriers are explicitly accountable for third-party AI with audit rights extending to their vendors (NAIC, AI Systems Evaluation Tool pilot, 2026).
That last point is the one worth sitting with. If an examiner asks why a specific claim was denied, “the model decided” is not an answer a Chief Claims Officer can give, and under the current regulatory framework, the exposure belongs to the carrier, not the technology vendor that built the model. Explainability, audit trail, bias testing and vendor oversight have moved from future differentiators to procurement gates carriers apply today, before a contract is signed.
Is AI-Generated Fraud the Next Claims Risk?
The fraud conversation has also shifted. Roughly 67% of North American carriers now use generative AI for fraud detection, up 16 points year over year, yet close to half report seeing more suspicious claims that themselves contain AI-generated documentation, forged invoices, synthetic estimates and fabricated medical records (industry fraud-detection benchmarking cited in ISG, 2026 Provider Lens – P&C BPO). In effect, one AI system is now screening claims that another AI system may have helped fabricate.
Rules-based fraud scoring, built around behavioral and historical patterns, was not designed to catch a synthetic document that looks structurally correct. Detection has to move upstream into document forensics, provenance checks and biometric validation at intake, rather than relying solely on how a claim behaves after it is filed.
What Does Real Claims Transformation Look Like in 2026?
Two of the industry’s most closely watched analyst assessments this year point in the same direction. ISG’s 2026 Provider Lens for P&C BPO finds carriers shifting from process automation to decision-centric models, using agentic AI for triage and early-stage claims while modernizing through orchestration layers over legacy cores rather than core replacement (ISG, 2026 Provider Lens – P&C BPO). Everest Group’s 2026 assessment of the top P&C technology providers confirms that live production AI is concentrated specifically in underwriting and claims, not spread thinly across every function (Everest Group, Top 50 P&C Tech Providers 2026).
A few practical principles follow from that, regardless of which vendor or platform a carrier is evaluating.
Separate the volume from the judgment. Most claims operations already know, roughly, what share of intake is genuinely rule-based and straight-through versus what requires a licensed adjuster’s judgment. The highest-value automation target is almost always the rule-based share, freeing experienced adjusters for the complex liability, SIU and litigation files where their judgment actually changes the outcome.
Treat explainability as a design requirement, not an add-on. A decisioning process that cannot generate a rationale as it works, rather than after the fact, will not clear compliance review in a jurisdiction operating under the NAIC bulletin. This is the most common reason claims automation initiatives stall between pilot and production.
Look for orchestration over replacement. A full core claims system replacement is a multi-year, high-risk program most carriers cannot justify while premium growth is decelerating. Automation that sits as an orchestration layer over Guidewire, Duck Creek, EIS or FINEOS can deliver measurable cycle-time and leakage improvements without displacing the core.
Price the outcome, not the headcount. Cost-per-FTE benchmarking rewards whichever vendor supplies the most people, not the one that resolves claims fastest and most consistently. Cost per claim decision at a stated quality and defensibility standard is a more accurate unit for an AI-first operating model.
The Real 2026 Question
Growth is not going to outrun severity this year, and hiring is not going to close the capacity gap. The carriers that move their loss ratio in 2026 will be the ones that treat claims automation as a governance and operating-model decision first and a technology decision second, choosing a defined friction point, whether that is FNOL intake, document fraud screening or litigation triage, and moving it into a production process that a compliance officer, a board and a regulator can all sign off on.
That is a narrower, more disciplined bar than “adopt AI.” It is also the one that actually shows up in the loss ratio.
Frequently Asked Questions
What is claims AI governance and why does it matter now?
It is the set of controls, explainability, audit trail, bias testing and vendor oversight, that let a carrier demonstrate to a regulator why a specific claim decision was made. It matters now because 29 US jurisdictions have adopted formal AI governance expectations, and the NAIC’s evaluation tool is already feeding into market conduct exams (NAIC, 2026).
Is claims outsourcing still relevant now that agentic AI is available?
Yes, and increasingly the two are the same conversation. Carriers evaluating claims AI are asking who is accountable for the operating outcome, not just who supplies the software, which is pulling outsourced claims operations and AI adoption into a single decision rather than two separate ones.
How is AI changing claims CX for policyholders?
The clearest gains show up in cycle time and consistency. A claim that advances itself through intake, triage and adjudication without waiting for a human handoff reaches a decision faster and more predictably, which is what policyholders actually experience as good claims CX at the most emotionally charged moment in the relationship.
Why do so many claims automation pilots stall before reaching production?
Sources
- Aon, 2026 US P&C Market Outlook — premium growth, casualty severity and nuclear verdict data
- EY, 2026 Claims Quality Assessment — claims leakage benchmarking
- Gallagher Bassett, Carrier Perspective 2026 (n=250 US/UK/AU carrier leaders) — talent and capacity data
- National Association of Insurance Commissioners (NAIC) — AI Model Bulletin adoption tracking and AI Systems Evaluation Tool pilot, 2026
- ISG, 2026 Provider Lens – P&C BPO (July 2026) — decision-centric claims automation trends and fraud-detection benchmarking
- Everest Group, Top 50 P&C Tech Providers 2026 — production AI concentration in underwriting and claims
