AI in Insurance 2026: Underwriting, Claims, and Fraud
AI - Artificial Intelligence
June, 2026
The AI insurance spending market was valued at $8.63 billion in 2025 and is anticipated to reach $59.5 billion by 2033. Industry spending on AI is expected to grow by more than 25% in 2026 alone.
The question for insurance carriers is no longer whether to deploy AI in insurance. It is whether their operating model is ready to capture its value. It must also manage regulatory and fraud risks from poorly governed AI.
This guide delivers an in-depth look at the evolution of these reimagined business processes. Learn what actual operations look like behind the headlines. In addition, carriers at different stages of maturity can use this guide to identify the priorities. Thus, they will know what to tackle first in pursuing their AI initiatives.
Who is This For?
This guide is intended specifically for CDOs, CTOs, heads of underwriting, claims leaders, and AI strategy teams. They may belong to property & casualty, life, and health insurance carriers. So, this guide will assist them in evaluating how best to allocate their AI investment resources for 2026.
Quick Answer: AI in insurance services is transforming three core functions in 2026. Underwriting is moving from static annual assessments to continuous real-time risk evaluation. Claims automation is delivering 75% faster resolution with 30-40% cost reductions. AI fraud detection in insurance is evolving into an arms race. The same tools enable both better detection and more sophisticated fraud.
Why 2026 is Different: Three Forces Converging
Generative AI Has Moved From Pilot to Production
According to Gartner, generative AI is expected to be the #1 investment priority for insurers at the start of 2026. This transition has not been based on hope. 86% of insurers anticipate increased machine learning spending in 2026. Besides, generative and agentic AI workflows are their top two priority investment areas. The era of pilot programs is over. Carriers that are still running proof-of-concept projects in 2026 are no longer evaluating AI; they are lagging behind.
Regulatory Accountability Went Live
The NAIC Model Bulletin regarding the use of AI in insurance has been adopted by 23 states and Washington, D.C., as of Q1 2026. Automated decision-making will not be obscured any further. Insurers will be required to document how AI-driven decisions are made, maintain audit trails, and provide explainability.
Black-box models can no longer be defended in a regulated environment. Additionally, the EU AI Act will place further regulatory pressure on carriers operating in the European Union, classifying insurance decision-making for credit, health, and life insurance as high-risk AI applications.
The Insurtech Gap is Already a Reality
Insurtech companies like Lemonade and Clearcover have been operating AI-native claims operations for years. Incumbents do not have to worry about potential future competition from insurtechs; instead, they are now behind, and the competitive gap continues to widen as insurtechs’ operating costs decrease while carriers’ operating costs remain relatively stable and high.
Read more: AI-Powered Fraud Detection in Banking: Guide
Underwriting Rebuilt: From Static to Continuous
Traditional underwriting was originally based upon a single point in time. Underwriting involves completing a form, having an underwriter make a risk decision based on the current risk snapshot, and then revisiting it at renewal each year. This model was built for an era when data was collected slowly, and human review was the only means of processing it.
Continuous underwriting now replaces the traditional underwriting model and performs real-time risk assessments using streaming data from IoT sensors, telematics applications, satellite imagery, credit signals, and behavioral measurements.
The achievement of continuous underwriting has had a significant impact on insurer operational productivity, with many carriers seeing reduction rates in underwriting production cycle time from three days down to three minutes, while for carriers that have implemented an AI-powered risk assessment pipeline, their automated straight-through underwriting production rate has increased from a range of 10% – 15% to 70% – 90%, respectively.
AIG’s generative AI-assisted underwriting assistant, created with Anthropic and Palantir, is the most cited production deployment of 2026. It ingests and prioritizes each carrier’s submission for excess and surplus, resulting in superior value risk selection within the underwriter team, while automating the routine decisions that previously consumed the most time for underwriters during a routine review process.
The technology for continuous underwriting exists. The real blockers are organisational: clean data infrastructure, model governance that satisfies NAIC documentation requirements, and underwriter workflows redesigned around AI-assisted decision-making rather than AI replacement.
Claims Automation: Where AI Delivers Immediate ROI
If continuous underwriting creates a long-term, sustainable competitive advantage for an insurance carrier, then claims automation provides the carrier with a measurable return on investment immediately. Thus, claims automation is the starting point for most carriers’ AI applications.
The FNOL Moment
The FNOL of an insurance claim is the single most-leveraged point in the claims life cycle. The quality of information captured at FNOL drives the estimated level of the payable reserve, the speed of settlement, and the overall effectiveness of fraud monitoring. When insurers deploy AI at FNOL for routing, triage, preliminary damage assessment, and fraud scoring, they can reduce a formerly multi-day process to minutes.
As a result, carriers utilizing AI automation in their claims resolution processes are resolving claims 75% faster, with an average cost reduction of 30% – 40%. While specific to US-based P&C carriers, straight-through processing rates for the highest-volume motor vehicle claims at a defined level of severity have exceeded 60% at leading carriers.
The Agentic Shift
The 2026-2027 period is the transition from AI-assisted claims workflows, where an adjuster uses AI tools, to AI-orchestrated workflows, where AI manages the claim end-to-end, and the adjuster reviews outcomes. The compound AI model architecture is the approach gaining the most traction: a primary orchestrating model that simultaneously directs specialized sub-models for document classification, damage assessment, fraud detection, and reserve calculation.
The STP ceiling is no longer technical. AI no longer limits the straight-through processing rate; organizational readiness to hand control to a system limits it. Carriers that have reached 60-70% STP consistently report the remaining barrier is governance and workflow redesign, not model capability.
The AI Fraud Arms Race: Both Sides Are Using the Same Weapons
When the topic is AI and insurance fraud, discussions often focus exclusively on detection. Rarely do they address the other half of that story: AI is also facilitating the commission of insurance fraud at an unprecedented rate, while simultaneously decreasing its detection. The asymmetry created by AI capabilities will be the predominant fraud-related challenge that carrier organizations face in 2026.
The insurance fraud problem in the U.S. is a significant economic issue that costs consumers more than $300 billion per year, or roughly $900 per policyholder, due to higher premiums. Approximately 10% of all property and casualty losses are attributable to claims fraud. These numbers do not represent a new occurrence. However, AI continues to change both the detection and commission of insurance fraud at a speed no one has ever seen before.
Fraud Detection
On the fraud detection side, AI has made real advancements. Using graph databases allows insurers to identify fraud-ring behavior that might never surface in a traditional review of an individual claim. Behavioral signal analytics can identify anomalies at FNOL, flagging them to a human adjuster before that person has an opportunity to review a single document in the claim. Carriers that have deployed these systems report over 30% improvement in fraud detection rates compared to pre-AI baselines.
The offense side has a much graver outlook. According to the Verisk State of Insurance Fraud Study published in March 2026, 98% of insurance companies reported that AI-based photo-editing tools drive digital fraud. Only 32% of the insurers who completed the survey expressed being very confident in their ability to detect deepfakes used in correspondence with claims submissions.
Ghost broker scams now use realistic, AI-generated websites, counterfeit branding, and AI-driven consumer engagement to market fraudulent policies, with victims only becoming aware of their status when they attempt to submit a claim. Most importantly, 55% of all Gen Z respondents indicated they would consider using AI technology to edit a claim photograph or document.
What to Do
Carriers must recognize the implications of these findings on their organizational structure. Fraud detection can no longer be a post-submission, claims function. Instead, it must be initiated at the underwriting stage by using AI-based vehicle image recognition systems to scan photographs submitted to underwriters and identify any pre-existing damage, which is then tagged so it can be easily identified if and when the insured later submits a claim for that damage.
An important principle is that the companies that will win the war against insurance fraud are not the ones with the best fraud detection models; they are the ones that have fully integrated all data and information across their underwriting and claims management systems, enabling the kind of real-time feedback that makes AI fraud detection in insurance genuinely effective rather than reactive.
The Regulatory Reality: What NAIC and EU AI Act Mean for AI Deployments
Regulatory accountability for AI in insurance is not a future restriction; it is an imminent, accelerating reality.
As of early 2026, 23 U.S. territories (states and Washington, D.C.) have formally adopted the NAIC Model Bulletin on the Use of AI in Insurance.
NAIC
The NAIC Model Bulletin requires insurers to provide documentation to support how AI impacts both underwriting and claims management decisions, that there is a thorough process in place to ensure human oversight in any dependent decisions based on a consequential outcome, and that each carrier must establish model governance that demonstrates that all automated decisions are free of any unfair discrimination.
A carrier that does not have properly documented model governance for its AI-based underwriting and/or claims systems will likely not comply with the NAIC Model Bulletin in many of the 50 U.S. territories.
EU AI Act
In addition to NAIC requirements, the EU AI Act also classifies insurance as a high-risk AI application due to the use of AI in creating insurance-related decisions that involve credit, health or life; therefore, AI systems that support insurance decisions related to credit, health or life, will also require compliance assessments, documentation regarding the technical aspects of how the AI model operates, human oversight mechanisms (e.g., processes) that will be in place and continuous monitoring of the AI model once deployed.
Any insurers that have conducted business within the European Union or have reinsurance relationships are subject to the aforementioned components of the EU AI Act.
The Increased Demand for AI Explainability
The most underrated challenge for many carriers is explainability. A better underwriting decision has no value unless the carrier can explain it satisfactorily to an auditor, regulator, or claimant. Carriers investing in model performance without equally investing in explainability infrastructure are building a compliance liability alongside the capability.
The Carrier Maturity Framework: What to Prioritize Based on Where You Are
| Maturity Tier | Where They Are | Priority Action | Why |
| Early | Manual workflows, limited data infrastructure | Claims automation at FNOL | Fastest ROI, lowest risk, builds AI organisational muscle |
| Developing | Some automation, limited AI governance | Model governance and data infrastructure | Scaling without governance creates regulatory exposure |
| Advanced | AI in production, scaling challenges | Continuous underwriting and fraud arms race defence | Long-term competitive advantage and fraud cost reduction |
Most industry professionals believe they have a greater opportunity to develop a sustainable competitive advantage by implementing AI and should focus on underwriting and developing AI underwriting models first.
In reality, claims automation first delivers faster ROI, builds the organisational trust and workflow redesign capabilities that underwriting AI requires, and creates the connected data environment that continuous underwriting needs.
Underwriting AI built on fragmented claims data produces risk decisions that are worse than those of the manual process it replaces.
Therefore, we recommend starting with claims automation to build a data infrastructure that connects underwriting to claims, and then deploying continuous underwriting on a foundation built with that data.
How SG Analytics Supports Insurance Carriers Navigating AI Transformation
SG Analytics is a global data and analytics services firm helping enterprises turn data into decisions. For insurance carriers navigating AI transformation, our research and insights practice tracks regulatory developments, technology trends, and operational frameworks to help leaders ask better questions before committing to platforms and deployments.
Contact us today to streamline insurance operations, optimize risk calculations, and demonstrate solid compliance.
FAQs
AI in underwriting is shifting carriers from static, annual risk assessments to continuous underwriting. It is a real-time risk evaluation drawn from IoT sensors, telematics, satellite imagery, and behavioral data. Leading carriers are also reducing underwriting timelines from three days to three minutes. Thus, straight-through processing rates reach 70-90% for standard risk profiles.
Continuous underwriting is a model in which insurers assess risk in real time using streaming data rather than at a single point in time, such as during application or annual renewal. It replaces the traditional snapshot model with a continuously updated risk position, enabling more accurate pricing, faster decision-making, and earlier identification of material changes in risk.
AI fraud detection systems combine graph database analysis to identify fraud rings. They leverage behavioral signal monitoring to flag anomalies at FNOL. Furthermore, AI-enabled computer vision can detect manipulated claim photographs. Similarly, predictive scoring lets you assess the probability of fraud before a human adjuster reviews a submission. The most effective deployments connect underwriting and claims data. Thus, fraud flags set at policy inception automatically carry forward when someone files a claim.
Straight-through processing measures the percentage of claims that the system assesses, validates, and settles without any intervention from a human adjuster. Leading US P&C carriers are achieving 60% or more STP for motor claims under a defined severity threshold. The primary barrier to higher STP rates in 2026 is not technical. It is organizational readiness to operate under AI-orchestrated workflows.
The EU AI Act classifies insurance decisions touching credit, health, and life as high-risk AI applications. These require conformity assessments, technical documentation, human oversight mechanisms, and ongoing monitoring. Carriers with European operations must meet these requirements before deploying AI for covered decision types.
In AI-assisted claims workflows, adjusters use AI tools to support their decisions. In AI-orchestrated workflows, AI manages the claim end-to-end, and the adjuster reviews outcomes rather than driving them. The 2026-2027 period is the transition point between these two models for most US P&C carriers.
Deloitte projects that P&C insurers could save between $80 billion and $160 billion in fraudulent claims by 2032 through AI-driven detection. At the individual carrier level, carriers that have deployed connected underwriting and claims fraud systems report fraud detection rate improvements of more than 30% compared to pre-AI baselines.
Related Tags
Author
SGA Knowledge Team