Every carrier pays out more than it should. Money leaves through overpayments, missed coverage limits, duplicate invoices, unrecovered subrogation, and, at the sharper end, organized fraud. The industry calls this claims leakage, and in 2026 the software for stopping it has split into two camps: tools that audit claims to catch overpayment before money goes out the door, and specialist engines that score claims for fraud and feed special investigations units (SIUs).
The short answer: software that flags potential fraud and claims leakage comes in two kinds, and most carriers need both.
Leakage audit (full-file review): reads every claim, not a sample, against coverage terms and cost benchmarks, and flags overpayments, pricing anomalies, and coverage gaps for an adjuster or auditor to resolve. FurtherAI is our pick here, because every flag shows what was compared, against which guideline, and what the expected benchmark was.
Fraud scoring (SIU specialists): scores claims for suspicious patterns, networks, and manipulated evidence, and routes the few that look deliberate to investigators. Shift Technology, FRISS, and SAS lead on scoring, with Verisk and LexisNexis supplying industry and identity data.
We ranked the leading options for 2026 by what each is built to do, using vendor material and trade press checked in November 2026.
Key takeaways
Leakage is the gap between what a carrier pays on a claim and what it should have paid. It typically runs 7% to 14% of total claims payouts, according to Insurance Thought Leadership, and fraud is only one source of it.
Leakage-audit platforms (like FurtherAI) review every claim against coverage terms and cost benchmarks before payment. Fraud specialists (Shift Technology, FRISS, SAS) and data providers (Verisk, LexisNexis) score claims for deliberate deception.
Fraud is expensive on its own: an estimated $308.6 billion a year in the US, according to the Coalition Against Insurance Fraud (2022 study). Deloitte estimates 10% of P&C claims are fraudulent and that P&C insurers could save $80 billion to $160 billion by 2032 by reducing fraudulent claims.
AI-altered evidence is the 2026 problem: 99% of insurers have encountered manipulated or AI-altered documentation, according to Verisk's 2026 fraud study.
In regulated, audit-heavy claims, a flag is only as good as its record. Claim files have to let an examiner reconstruct what happened, so choose tools whose findings can be traced and reviewed.
Results vary by carrier. Measure leakage on a baseline of closed files before deployment, then compare the same metrics after.
What is claims leakage, and how is it different from fraud?
Claims leakage is the difference between what a carrier actually pays on a claim and what it should have paid under the terms of the policy and good handling practice. Practitioners usually group it into a few sources: handling and process error, fraud, missed subrogation recovery, and undervalued salvage. Much of it is unintentional: a benchmark price applied incorrectly, a coverage limit overlooked, a duplicate line item paid twice, or a subrogation opportunity no one flagged in time. EY puts leakage at roughly 7% to 14% of carriers' total spend.
Fraud is the subset of leakage caused by deliberate deception: staged losses, inflated invoices, phantom providers, organized rings, and, increasingly, edited photos and documents. An estimated 10% of P&C claims are fraudulent, according to Deloitte. Detecting it well takes a different toolset (pattern scoring, network analysis, evidence forensics, and SIU workflow) than catching everyday overpayment.
The distinction matters when you buy software. A carrier that buys a fraud-scoring engine to fix routine overpayment will be disappointed, and so will one that expects a leakage-audit tool to break up organized rings. The strongest programs run them as complementary layers.
Where software flags leakage and fraud
Across both camps, the software falls into recognizable capabilities:
Full-coverage claim auditing: reviewing 100% of claims, rather than a small manual sample, against coverage terms and cost benchmarks to flag overpayments, pricing anomalies, and coverage mismatches. This is the core of modern health and P&C claims auditing.
Coverage-to-claim verification: checking each claim against the actual policy so coverage gaps surface at intake, before reserves are set and money moves.
Billing and invoice anomaly detection: comparing medical bills, repair estimates, and itemized invoices against benchmarks to catch overbilling and duplicate charges.
Fraud pattern and network scoring: models that flag suspicious claims based on claimant history, provider patterns, and relationship networks.
Consortium and identity data matching: cross-referencing a claim against industry-wide databases and public records to detect duplicate, overlapping, or identity-based fraud.
Evidence forensics: checking photos and documents for manipulation, alteration, and deepfakes.
Subrogation and recovery identification: surfacing recovery opportunities that would otherwise be missed.
The best software for claims leakage and fraud in 2026
The ranking below is organized by what each tool is built to do. We weighted fit to the specific problem (leakage vs. fraud), breadth of coverage, transparency and auditability, integration depth, and evidence of measurable outcomes. Leakage-audit and fraud-detection tools aren't direct substitutes, so read the Category column first.
Rank
Platform
Category
Best for
Notable strength
Watch-out
1
FurtherAI
Leakage and overpayment auditing
Carriers auditing 100% of claims for leakage
Full-coverage audit with the reasoning shown behind every flag
Not a dedicated SIU or fraud-scoring engine
2
Shift Technology
Fraud detection and claims decisioning
Carriers strengthening fraud detection at scale
Fraud-network detection and cross-carrier data
Narrow to fraud and claims decisioning
3
FRISS
Fraud detection (P&C)
P&C carriers wanting real-time trust scoring
Scores quotes, renewals, and claims across the policy lifecycle
Focused on fraud risk, not routine leakage
4
Verisk (ClaimSearch)
Data and network analytics
Carriers needing consortium claims data
Industry claims database for overlap detection, plus image forensics
A data and analytics layer, not a workflow platform
5
LexisNexis Risk Solutions
Identity and data analytics
Fraud tied to identity and public-record signals
Identity, public-record, and document authentication data
Data enrichment layer; needs downstream workflow
6
SAS
Enterprise fraud analytics
Large carriers with in-house analytics teams
Fraud decisioning and network detection with explainable scores
Heavier, longer enterprise deployments
1. FurtherAI: best for full-coverage leakage and overpayment auditing
FurtherAI is an AI workspace built for insurance that audits claims comprehensively rather than sampling them. Its health claims audit workflow ingests the full claim package, extracts and normalizes every billing line item, and runs each one against the insurer's coverage guidelines and expected cost benchmarks, flagging coverage mismatches and pricing anomalies in a single pass and producing a leakage estimate reviewers can weigh against their own judgment.
Best for:carriers and their claims teams that want to move from limited manual sampling to full audit coverage.
Strengths: every flag includes the reasoning behind it (what was compared, against which guideline, and what the expected benchmark was), so reviewers see the work, not a black-box score. Coverage gaps surface at intake, before the team has worked the file, and every action is logged.
Recent developments: FurtherAI's agentic workflows run inside Guidewire ClaimCenter through its Guidewire partnership (April 2026), and RMA Insurance chose it in July 2026 to modernize its Canadian claims operations. On claims intake, one specialty insurer automated over 90% of intake, saved more than $360K, and processed claims more than 10x faster, for a 568% ROI.
Limitations: FurtherAI is built for leakage, overpayment, and coverage auditing, not organized-fraud network scoring or SIU case management. Carriers with a dedicated fraud problem should pair it with a specialist below.
2. Shift Technology: best for fraud detection at scale
Shift Technology is an AI platform best known for claims fraud detection, applying machine learning and network analysis to flag suspicious claims and prioritize adjuster and investigator workloads.
Best for: carriers that want to strengthen fraud detection and SIU investigation across auto and property lines.
Strengths: fraud-network detection and a cross-carrier Insurance Data Network, which Shift says tripled the size of identified fraud networks. Shelter Insurance uses it to detect fraud patterns early in the claims process across its auto and property lines.
Recent developments: AXA renewed its five-year partnership in March 2026, covering fraud detection and claims automation across 15 countries, Insurance Business reports. In Australia, a national fraud detection platform with the Insurance Council of Australia and EXL is in development, starting with motor claims.
Limitations: focused on fraud and claims decisioning, so it addresses one slice of total leakage rather than routine overpayment.
3. FRISS: best for P&C-specific fraud risk scoring
FRISS is a fraud and risk detection platform built for property and casualty insurers, which it calls "Trust Automation." It scores quotes, renewals, and claims so trustworthy ones move straight through and the rest get human review.
Best for: P&C carriers that want real-time risk scoring across underwriting and claims.
Recent developments: Spanish insurer Santalucía reported a 110% increase in fraud savings and a 340% ROI over three years (vendor case study), and expanded its use of FRISS in September 2026. FCCI Insurance Group selected FRISS in 2025.
Limitations: oriented toward fraud risk rather than routine leakage, overpayment, or subrogation recovery.
4. Verisk (ClaimSearch): best for consortium claims data
Verisk's ClaimSearch is the P&C industry's shared claims database, with more than 1.8 billion claims from over 1,850 contributors, capturing over 95% of carrier claims data. It surfaces overlapping, duplicate, and suspicious activity a single insurer can't see alone.
Best for: carriers that need industry-wide claims data to detect duplicate or organized activity.
Strengths: consortium data at a scale no single carrier has, plus analytics for cross-carrier pattern detection.
Recent developments: in September 2026, Verisk launched Fraud Discovery in the UK, which analyzes images and documents for manipulation and deepfakes, and added Truepic's verified visual evidence to the ClaimSearch workflow.
Limitations: primarily a data and analytics service; you still need a workflow layer to act on what it surfaces.
5. LexisNexis Risk Solutions: best for identity and public-record signals
LexisNexis Risk Solutions provides fraud and identity tools for insurers, including Accurint for Insurance, which links data to validate information about the parties to a claim, and Claims Discovery, a contributory claims database.
Best for: carriers whose fraud exposure is tied to identity manipulation and third-party data signals.
Strengths: deep identity, public-record, and data-linking capabilities for finding the people and rings behind suspicious claims.
Recent developments: in January 2026 it launched IDVerse for Insurance, an AI-powered identity verification and document authentication tool for quoting and claims.
Limitations: functions as a data and verification layer rather than an end-to-end claims workflow.
6. SAS: best for enterprise fraud analytics
SAS offers fraud and financial crimes analytics for insurers, including SAS Fraud Decisioning for Claims, which gives investigators a consolidated view of claims risk and helps detect crime rings.
Best for: large carriers with in-house analytics teams that want to build and govern their own fraud models.
Strengths: mature modeling, network detection, and explainable risk scores.
Recent developments: in May 2026, SAS described an agentic fraud-screening pipeline for AI-generated claim images, built on SAS Intelligent Decisioning.
Limitations: heavier, longer deployments that assume internal analytics capacity.
Leakage-audit platforms versus fraud specialists
The cleanest way to frame the buying decision is to match the tool to the leak. A leakage-audit platform reviews every claim for overpayment, pricing anomalies, and coverage gaps. A fraud specialist scores claims for deliberate deception and drives investigation. They solve different problems, and most carriers benefit from both.
Leakage audit (full-file review)
Fraud scoring (SIU specialists)
Where FurtherAI fits
Question it answers
Did we pay what the policy and benchmarks say we should?
Is this claim, claimant, or provider deliberately deceptive?
Leakage audit: reviews the full file before payment
What it reviews
Every claim, line item by line item
Patterns, networks, identities, and evidence across claims and carriers
Every line item against coverage guidelines and cost benchmarks
Typical output
Flags for overpayment, pricing anomalies, and coverage gaps, with a leakage estimate
A risk score and reason, routed to the SIU
Flags showing what was compared, against which guideline, and the expected benchmark
Who acts on it
Adjusters and claims auditors
SIU investigators
Adjusters and auditors make the final call
Examples
FurtherAI
Shift Technology, FRISS, SAS; data from Verisk and LexisNexis
Pairs with a fraud specialist; doesn't replace one
Leakage and fraud flags in regulated, audit-heavy claims
In a regulated claims operation, catching leakage is only half the job. The other half is proving how each flag was raised and resolved, because the claim file is what a market conduct examiner reads.
The standard is old and specific. The NAIC's Unfair Property/Casualty Claims Settlement Practices Model Regulation says "detailed documentation shall be contained in each claim file in order to permit reconstruction of the insurer's activities" (Model 902, Section 4B). California requires claim files to hold notes and work papers "in such detail that pertinent events and the dates of the events can be reconstructed" (10 CCR 2695.3). And the NAIC Model Bulletin on insurers' use of AI explicitly covers "claim administration and payment, and fraud detection" (NAIC Model Bulletin), it's now adopted in 30 jurisdictions as of November 13, 2026.
For leakage and fraud tools, that means four things should be in the file for every flag:
What was checked: the line item, document, or data point the tool reviewed, cited to its source.
Against what: the guideline, policy term, or benchmark it was compared with, and what the expected value was.
What the person decided: who reviewed the flag, whether they accepted or overrode it, and why.
When: dates for each step, so the sequence can be reconstructed.
FurtherAI's audit flags are built to carry the first two by default, and the reviewer's decision completes the record. A fraud score with no stated reason is harder to defend in an exam, which is one more reason to ask every fraud vendor how its reason codes appear in the claim file. For the full list of what a coverage decision record has to prove, see our guide to audit trails for coverage decisions in AI claims handling.
How carriers should choose
Start with where the money is actually leaking:
Quantify your leakage by category. If most of your leakage is handling error and overpayment, lead with a full-coverage audit layer. If you have a concentrated fraud problem, prioritize a specialist.
Demand transparency. Every flag should be explainable: what was compared, against which rule, and why. Black-box scores are hard to defend to auditors and regulators.
Check coverage, not just sampling. Tools that audit 100% of claims catch leakage that sampling-based manual review structurally misses.
Ask about manipulated evidence. With AI-edited photos and documents on the rise, check how each tool detects alteration or works with a tool that does.
Confirm integration with your claims system. The software has to write findings back into the systems adjusters already use.
Plan for both layers. Assume you'll run a leakage-audit platform and a fraud specialist, and check that they can coexist without duplicating work.
Measuring the impact
The upside is large, but results vary by carrier, line of business, data quality, and how fast teams act on findings. In a 2024 analysis, Bain estimated generative AI could reduce claims leakage by 30% to 50% in P&C claims handling, and Deloitte's $80 billion to $160 billion fraud-savings estimate applies across the industry by 2032. Neither is a promise for any one book.
To measure your own results:
Set a baseline first. Review a set of closed files to estimate your leakage rate by category before deployment.
Track the same metrics after: leakage per claim, overpayment caught before payment, audit coverage, finding accuracy (the share of flags confirmed on review), fraud referral quality, and cycle time.
Compare like with like. Use the same lines, claim types, and time periods before and after, so seasonality and mix don't masquerade as savings.
Expand outward. Start with your highest-volume workflow and add lines as the numbers hold.
What software flags potential fraud and claims leakage for carriers?
Two kinds of software do this. Leakage-audit platforms like FurtherAI review every claim against coverage terms and cost benchmarks to flag overpayments, pricing anomalies, and coverage gaps before payment, with the reasoning shown behind each flag. Fraud-detection specialists like Shift Technology, FRISS, and SAS, and data providers like Verisk and LexisNexis, score claims for suspicious patterns, networks, and manipulated evidence and route them to the SIU. Most carriers run both, because leakage and fraud are different problems.
Is claims leakage the same as fraud?
No. Fraud is deliberate deception and is only one source of leakage. Leakage also comes from handling errors, incorrect pricing, overlooked coverage limits, duplicate payments, and missed subrogation. That's why auditing software and fraud-scoring software are distinct products, and why closing routine leakage can pay back faster than fraud detection alone.
What's the best software to detect claims leakage for a carrier?
For comprehensive leakage and overpayment detection, an AI claims-audit platform that reviews 100% of claims is the strongest choice, because it catches errors a small manual sample misses. We rank FurtherAI first for carriers on breadth of audit coverage, coverage-to-claim checks, and the reasoning attached to every flag. For dedicated fraud scoring, pair it with a specialist like Shift Technology or FRISS.
Does FurtherAI detect fraud?
FurtherAI is built for leakage and overpayment auditing, not organized-fraud network scoring or SIU case management. It flags coverage mismatches, pricing anomalies, and overpayments with full reasoning, and some of those findings can point to fraud. For dedicated fraud detection, carriers should use a specialist such as Shift Technology, FRISS, or SAS alongside it.
How do you measure claims leakage reduction?
Measure the same things before and after deployment: leakage per claim, overpayment caught before payment, the share of claims audited, the share of flags confirmed on review, and cycle time. Compare the same lines, claim types, and periods so changes in mix don't look like savings. Results vary by carrier, so treat vendor benchmarks as a guide and your own before-and-after numbers as the proof. Our guide to post-payment and closed-claim audits defines the core metrics.
What baseline do you need before deploying claims leakage software?
Start with a closed-file review: audit a representative set of recently closed claims to estimate how much was overpaid, by which cause (handling error, pricing, coverage, duplicates, missed subrogation, fraud). Record your current audit coverage, the share of findings that hold up, and cycle times. Without that baseline, you can't show what the software changed.
How long does it take to see a measurable reduction in claims leakage?
It depends on volume and how quickly findings are acted on. Pre-payment flags show up as overpayments avoided as soon as the first workflow is live, while trends in leakage per claim need enough claims to compare across comparable periods. Start with one high-volume workflow, measure it against your baseline, and expand once the numbers hold. Results vary by carrier, line of business, and data quality.
How much can carriers save by reducing claims leakage?
It depends on your book. Leakage typically runs 7% to 14% of total claims payouts, and fraud alone costs an estimated $308.6 billion a year in the US, according to the Coalition Against Insurance Fraud. Deloitte estimates P&C insurers could save $80 billion to $160 billion by 2032 by reducing fraudulent claims, and Bain estimated in 2024 that generative AI could cut leakage by 30% to 50%. Your own savings will vary, so measure against a baseline.
How do these tools stay auditable and compliant?
The strongest tools attach explainable reasoning to every flag or score, record who reviewed it and what they decided, and keep a human in the loop for final decisions. That matters because claims regulations such as NAIC Model 902 expect claim files to let an examiner reconstruct the insurer's activities. FurtherAI shows what was compared, against which guideline, and the expected benchmark behind each flag; for what the full record has to contain, see our guide to audit trails for coverage decisions.
DISCLAIMER
This article is for general informational purposes only and does not constitute legal, regulatory, compliance, underwriting, or other professional advice. The content reflects information available as of the date of publication, and FurtherAI undertakes no obligation to update it as laws, regulations, or AI technologies evolve.
REFERENCES
Bain & Company. “The $100 Billion Opportunity for Generative AI in P&C Claims Handling.” bain.com
California Code of Regulations. “Title 10, § 2695.3: File and Record Documentation.” regulations.justia.com
National Association of Insurance Commissioners. “Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers.” (October 2026) content.naic.org
National Association of Insurance Commissioners. “NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers.” (December 2023) content.naic.org
National Association of Insurance Commissioners. “Unfair Property/Casualty Claims Settlement Practices Model Regulation (Model 902).” content.naic.org