Hesper AI
BlogVendor Comparison
Vendor ComparisonMay 13, 2026·13 min read·Nitish Badu, COO

Top 10 insurance fraud detection platforms in 2026

A neutral 2026 buyer's roundup of the 10 insurance fraud detection platforms most commonly on US P&C carrier shortlists, with where each wins and where each falls short.

NB
Nitish Badu · COO, Hesper AI
May 13, 2026·13 min read
$308B
Annual US insurance fraud loss
Coalition Against Insurance Fraud
~10%
Of P&C claims involve fraud
Industry baseline
60-85%
Rules-based false positive rate
Pre-ML detection systems
1.8B
Records in ISO ClaimSearch
~95% of US P&C market

The 2026 detection category looks different from the one carriers were buying into in 2023. Three forces have reshaped it: a maturing cross-carrier data utility layer that Verisk continues to compound, the arrival of handler-assist agentic AI from Shift Technology, and a mid-market scoring tier where FRISS and a handful of newer entrants compete on deployment speed. Underneath all three is the same loss-cost math. US insurers lose roughly $308 billion to fraud each year per the Coalition Against Insurance Fraud, and roughly 10% of P&C claims contain a fraud element.

This is a buyer's-shortlist roundup, not a benchmark. We synthesized public product pages, named carrier deployments, and regulator filings to set the order. We did not run head-to-head precision-recall tests on flagged claims; no public dataset supports that. Hesper AI is not in the top 10. Hesper sits at the investigation layer downstream of detection - a different category from every vendor below - and we cover it separately at the end of the post.

If you want the full category map across prevention, detection, and investigation, read the full Pillar 5 comparison of AI fraud platforms. This post focuses on the detection layer specifically: who is on most US P&C shortlists in 2026, what each one is actually good at, and where each one stops.

The 2026 fraud detection landscape

Detection has matured into a layered category. The cross-carrier data utility - what a single carrier cannot see on its own - is dominated by Verisk's ISO ClaimSearch, which holds 1.8B records and serves around 95% of the US P&C market per its product page, with 175K daily ingest and 200K daily users. Above that data utility sits a scoring tier: Verisk ClaimDirector with its 0-999 methodology, FRISS, Shift Technology, BAE Systems NetReveal, and SAS Fraud Framework. The frontier in 2026 is agentic AI for handlers - Shift Claims pushed the category here in 2025, and the rest of the field is reacting.

The cost of stillness in this category is high. Rules-only detection systems run 60-85% false positive rates, which jams the SIU queue with claims that should never have been flagged. The carriers winning here are not the ones with the most rules. They are the ones with the deepest contributory data and the cleanest hand-off into the case workflow downstream. That is why Verisk's Q4 and FY2024 earnings show $2,882M in full-year revenue with the claims sub-segment growing 13% year over year in Q4. The data network compounds; the moat widens.

What changed since 2024

Three shifts define the 2026 evaluation: Shift Technology released Shift Claims (handler-assist agentic AI for adjusters), Verisk continued compounding its data-network advantage with FY2024 revenue of $2,882M, and a new procurement category opened up at the investigation layer downstream of the flag. Carriers who last refreshed in 2023 will see a meaningfully different shortlist today.

How we ranked the 10

Six criteria, equally weighted. We did not collapse them into a single score because the procurement-relevant trade-offs are between criteria, not within them. A vendor that wins on data depth but loses on deployment speed is a different decision from one that wins on deployment speed but loses on data depth.

  1. Detection accuracy: precision and recall on real flagged claims, as reported by the vendor and triangulated against named deployments.
  2. Data network depth: how much cross-carrier contributory data the platform draws on. Verisk's ISO ClaimSearch is the bar at 1.8B records.
  3. Integration footprint: native connectors to Guidewire ClaimCenter, Duck Creek Claims, Snapsheet, and custom claims systems.
  4. Regulatory posture: alignment with NAIC Model Act 680 antifraud-plan filings (adopted in 48 states) and state DOI audit defensibility.
  5. Deployment cycle: weeks for cloud-native scoring vendors, quarters or longer for enterprise platforms.
  6. Pricing transparency: most are subscription with custom quotes; we flag the few that publish, and the few that bundle into broader claims-tech contracts.

Order in the list below tracks usage depth in US P&C, not alphabetical order and not our preference. Verisk is first because its data utility sits underneath almost every other vendor on the list.

The 10 platforms

1. Verisk (ISO ClaimSearch + ClaimDirector + Xactware)

Verisk Analytics (NASDAQ: VRSK, parent founded 1971) operates the industry's deepest claims data utility. ISO ClaimSearch holds 1.8B claim records covering ~95% of the US P&C market, ingesting 175K records daily across 200K users. ClaimDirector layers a 0-999 fraud-scoring methodology on top of that data, and Xactimate is used by 22 of the top 25 US property insurers for estimation. The NICB has operated through ISO databases since 1998. Verisk reported $2,882M in FY2024 revenue with a +13% Q4 claims sub-segment.

Where they're weaker: ClaimDirector flags suspicious claims; it does not run per-claim investigation. The SIU still owns the 14+ day manual workflow downstream of the score. Procurement is also the slower path - public-company procurement cycles and custom-quoted subscriptions take quarters, not weeks. For a deeper head-to-head, see our Verisk comparison.

2. FRISS

FRISS (founded 2006, Netherlands) is the mid-market detection vendor most US carriers shortlist alongside Verisk. The product covers FNOL scoring and claim-lifecycle scoring with a workflow-for-humans interface that SIU directors consistently rate as easier to live with than enterprise platforms. Auto is the strongest line. The 2024 Capgemini investment moved FRISS further into the European mid-market, and its US footprint has grown through Guidewire ClaimCenter integrations.

Where they're weaker: the cross-carrier data network is smaller than Verisk's, so the contributory-data signal is thinner. And like every detection vendor on this list, FRISS flags and hands off; the per-claim investigation is still manual. Pricing is subscription, custom-quoted, with mid-market deals typically in the $300K to $1.5M annual range based on volume.

3. Shift Technology

Shift Technology (founded 2013, Paris) is the most active detection vendor in 2026. Their AXA Switzerland deployment processed over 1M claims and stopped more than €12M in fraud, with Samuel Klaus on record as the program lead. Shift Claims (launched late 2025) is the handler-assist agentic AI tier; per their announcement, deployments cite 3% loss reduction, 30% faster handling, 60% automation, and 99% accuracy on assisted decisions. The Shift IDN cross-carrier network is the cleanest non-Verisk contributory-data play in the field.

Using Shift Claims Fraud Detection, we are able to consistently identify suspicious activities at FNOL and assign the claim to the appropriate expert for investigation.

Samuel Klaus, Head of Fraud at AXA Switzerland (<a href="https://www.shift-technology.com/resources/case-studies/axa-switzerland-insurance-fraud-detection-success" target="_blank" rel="noopener noreferrer" style="color:#3C3B31;text-decoration:underline">Shift Technology case study</a>)

Where they're weaker: Shift Claims is handler-assist, not autonomous end-to-end investigation. The 99% accuracy figure refers to assisted decisions surfaced to a human, not autonomous resolution of a flagged claim. US footprint is thinner than the European one; the strongest named deployments are AXA Switzerland and a handful of European mutuals.

4. BAE Systems NetReveal

BAE Systems NetReveal (lineage from 1999) carries financial-crime DNA into insurance. Its strength is network analysis and ring detection - the platform was built for AML before it was tuned for insurance, and that lineage shows up in the graph models. For organized rings in workers comp medical billing or staged-collision auto fraud, NetReveal's ring detection is the strongest match in this roundup, drawing on its AML-graph lineage.

Where they're weaker: enterprise sales cycle, slower deployment, and an interface that SIU directors usually describe as built for analysts rather than investigators. The platform is contracted as an enterprise license, not subscription, so the cost curve looks different from the rest of the field. And the same per-claim investigation gap applies: NetReveal surfaces ring connections; the investigator still does the case work.

5. SAS Fraud Framework

SAS Fraud Framework (SAS parent founded 1976) is the rules-plus-ML scoring tier inside the SAS Viya analytics platform. It wins inside carriers that already run SAS Viya as the corporate analytics backbone - the integration cost is near zero, and the rules engine is among the most customizable in the field. Strong fit for technical-buyer audiences with in-house data-science capacity.

Where they're weaker: heavy lift for carriers not already on SAS Viya. Deployment requires SAS-skilled headcount and a multi-quarter implementation cycle. Investigation depth is shallow - same flag-and-hand-off pattern as the rest of the detection tier.

6. CCC Intelligent Solutions

CCC Intelligent Solutions (founded 1980, IPO 2021) anchors the auto-claims data network in the US, with deep integrations into OEMs and the repair-shop economy. Detection is part of a broader claims-tech bundle that includes estimating, photo AI, and subrogation. For carriers whose auto book drives the loss ratio, CCC's data network is the most relevant non-Verisk option.

Where they're weaker: CCC is auto-only in any meaningful sense. Property, workers comp, and liability carriers will not find a fit here. And the bundle pricing makes apples-to-apples comparison against pure detection vendors hard - the detection capability is rarely procured stand-alone.

7. Snapsheet

Snapsheet (founded 2011, Chicago) is mobile-first FNOL and digital triage; the platform has 170+ customers, runs in 16 of the top 20 US P&C carriers in some capacity, and has processed roughly $6B in appraisals. Detection signals are generated as a byproduct of the digital intake workflow - the platform catches inconsistencies in photo evidence, geolocation, and timeline at FNOL before a claim ever reaches an adjuster.

Where they're weaker: Snapsheet is not a standalone fraud detection platform. It is a claims workflow vendor whose digital intake surfaces some fraud signals as a side effect. Carriers shortlisting Snapsheet for detection alone are evaluating the wrong product; the right framing is FNOL workflow plus a downstream scoring vendor.

8. Tractable

Tractable (founded 2014, London) applies computer vision to auto damage assessment and property estimation. Tokio Marine's deployment was reported by Insurance Journal in April 2020, and the platform now serves auto carriers across Europe, Japan, and the US. Cycle time on visual damage assessment drops from days to minutes for the claim types Tractable's vision models handle.

Where they're weaker: Tractable is not strictly a fraud detection platform. Their computer vision processes legitimate damage assessment; fraud detection is not the layer they sit at. Including Tractable here is a buyer's reality check - it shows up on shortlists because carriers conflate "AI for claims" with "AI for fraud," and the two are different problems with different vendors.

9. Akur8

Akur8 (founded 2018, Paris) builds AI for pricing and underwriting, not detection. The platform automates rate-making and risk segmentation at underwriting, which has the effect of preventing some classes of fraud before a claim is filed. Carriers ask about Akur8 in fraud RFPs because the pricing AI conversation has spilled into the detection conversation; the two share "AI" in the headline but solve different problems.

Where they're weaker: Akur8 does not score or investigate claims. If the procurement question is detection, Akur8 is not the answer. We include it because it appears on shortlists, not because it competes head-to-head with the rest of this list.

10. Lemonade (illustrative, not a vendor)

Lemonade (founded 2015, public 2020) is a carrier with carrier-built AI, not a vendor selling to other carriers. We include it as a reference point for what carrier-internal fraud AI looks like at scale in 2026. Lemonade's claims-handling AI flags suspicious claims as part of the integrated underwriting-to-claims stack, and the company has published partial details on detection methodology in shareholder letters.

Where they're weaker: not sold to peers. Carriers cannot buy Lemonade's stack. We include it so the shortlist reflects the actual procurement reality: a few large carriers are building, and they are not in the market to buy. The carriers who do buy are evaluating the nine vendors above.

Side-by-side comparison

The matrix below collapses the profiles to five fields most procurement teams care about: parent-company founding year, tier orientation, whether the platform draws on cross-carrier contributory data, and pricing transparency. Tier is for orientation only - it is not a ranking.

VendorFoundedTierCross-carrier dataPricing transparency
Verisk (ClaimSearch + ClaimDirector)1971Tier 1 incumbentYes (industry utility)Subscription, custom-quoted
FRISS2006Tier 1 detectionLimited (vendor-held)Subscription, custom-quoted
Shift Technology2013Tier 1 detectionYes (Shift IDN)Subscription, custom-quoted
BAE Systems NetReveal1999Tier 1 enterpriseLimitedEnterprise contract
SAS Fraud Framework1976Tier 1 platform-bundledNoEnterprise license
CCC Intelligent Solutions1980Specialized (auto)Yes (auto network)Bundled, custom-quoted
Snapsheet2011Digital-first FNOLNo (workflow-driven)Subscription
Tractable2014Adjacent (vision)NoSubscription
Akur82018Adjacent (pricing AI)NoSubscription
Lemonade2015Carrier-builtN/A (carrier)N/A

Coverage of flagged claims: detection vendors stop at the flag

Flagged by detection vendor100%
Reach human SIU investigator~95%
Fully investigated within case-cycle window~25%
Resolved with audit-ready documentation~18%

Every vendor in the top 10 contributes to the first bar. None of them contributes to the third or fourth. That is the gap the rest of this post is about.

Detection vs investigation - the layer no platform on this list occupies

Detection is upstream; investigation is downstream. Every vendor above ends its job at the flag. The per-claim investigation that follows - document forensics, OSINT, statement cross-reference, timeline reconstruction, financial pattern review - is the manual SIU workflow at 14+ days per case. Coverage sits at roughly 25% of flagged claims at the median US P&C carrier; the rest are paid, denied without full work, or queued indefinitely.

Hesper AI builds the investigation-layer agent that runs end-to-end on every flagged claim. The agent runs 15+ investigation phases in parallel - the parallelism is structural, not optional, because a software agent's per-case attention is not a constraint the way a human investigator's is. The cycle compresses from 14+ days to 2-4 hours, and coverage shifts from ~25% to 100% of flagged claims. The output is an audit-ready report a human SIU lead reviews, not a score that hands off to an investigator who then does the work.

Hesper is complementary to every vendor in the list above. Not a replacement. A carrier running FRISS, Shift, or Verisk continues to flag with that vendor; Hesper takes the flag and resolves it. Read the three-layer model for the full breakdown of where each layer ends, and the Verisk head-to-head for how this looks against the largest data utility specifically.

How to pick: a decision framework for Top 50 P&C carriers

Three tiers of carrier need; three different shortlists. The wrong move in this category is to copy a peer carrier's stack without checking whether your loss ratio is concentrated in the same lines theirs is.

Tier A - National Top-25, complex multi-line

Anchor on Verisk for cross-carrier data (the 1.8B-record utility is not optional at this scale), layer FRISS or Shift Technology for scoring tuned to your strongest line, and add an investigation-layer vendor downstream. The CIO conversation will center on Guidewire ClaimCenter or Duck Creek Claims integration; the Compliance Officer will want documented-decision posture per NAIC Model Act 680.

Tier B - Mid-market multi-state

FRISS or Shift Technology as the primary scoring vendor, with Verisk ClaimSearch as the cross-carrier data utility underneath. The investigation-layer decision is the new 2026 procurement question for this segment; the math works because flagged-claim coverage at mid-market carriers is typically the lowest, and the cost-per-case delta from manual to AI is the largest.

Tier C - Regional, digital-first

Snapsheet or carrier-built AI for FNOL detection signals, plus Verisk ClaimSearch for cross-carrier match. The investigation layer often stays manual at this scale, but that is changing - the per-case cost curve makes investigation-layer agents accessible to regional carriers as the volume threshold drops. For the full stack picture, see the 2026 SIU technology stack and the claims management systems comparison.

Key takeaways

  • The 2026 US P&C detection shortlist is dominated by Verisk, FRISS, and Shift Technology; the other seven platforms address narrower segments and should be evaluated against specific line-of-business needs.
  • Verisk's industry-utility data depth (1.8B records, ~95% of US P&C market, FY2024 revenue of $2,882M) is the largest moat in detection and the safest procurement anchor for any carrier.
  • Shift Technology's move into handler-assist agentic AI is the most consequential category shift since 2024, but Shift Claims is detection-and-assist, not autonomous investigation.
  • Every detection vendor in the list flags claims and hands them to a human investigator; the per-claim investigation downstream is the manual SIU workflow at 14+ days per case and roughly 25% coverage.
  • The 2026 procurement decision for most Top-50 P&C carriers is not which detection vendor to buy - it is what to add at the investigation layer downstream of the detection vendor they already run.

Frequently asked questions

Most US P&C carriers above 100K annual claims run more than one. The pattern is a cross-carrier data utility (Verisk ClaimSearch is the de facto standard at ~95% of US P&C market) layered with a scoring or agentic-AI vendor (FRISS or Shift Technology) tuned to the lines the carrier writes most. The data-utility layer surfaces history a single carrier cannot see on its own; the scoring layer rank-orders new claims against that history plus carrier-specific patterns. Running both is not redundant; they answer different questions. The cost question is whether the scoring vendor's incremental lift over Verisk's native scoring justifies the second contract. For most multi-state mid-market carriers, the answer is yes.

Claims management systems (Guidewire ClaimCenter, Duck Creek Claims, Majesco ClaimVantage, Snapsheet) are the workflow backbone. They hold the case, route it, and store the file of record. Detection platforms sit alongside the claims management system and feed it scores or flags. Guidewire's Celent 2024 Luminary status across all four regional reports reflects workflow dominance, not detection capability. A carrier running Guidewire still needs Verisk, FRISS, or Shift Technology for fraud scoring, and the flagged-claim file lives in ClaimCenter while the score travels in from outside. See the side-by-side in our claims management systems comparison.

Most flagged claims are handed to a human investigator inside the carrier's Special Investigations Unit. The SIU runs the manual workflow: document forensics, OSINT, statement comparison, timeline reconstruction, financial pattern review. That workflow takes 14+ days per case at the median US P&C carrier, and the team gets through roughly 25% of the flag queue before the rest are paid, denied without full work, or queued indefinitely. The detection platform's job ends at the flag. The investigation is a separate layer with separate vendors - or, much more commonly today, no vendor at all and just headcount. See the three-layer model for the full breakdown.

For a single line of business with strong in-house data-science capacity, yes. Open-source scoring models (gradient-boosted trees, graph-based ring detection) can deliver detection lift comparable to the lower tier of commercial platforms. The trade-off is regulatory posture. State DOI antifraud-plan filings under NAIC Model Act 680, adopted in 48 states, require documented anti-fraud methodology. Commercial vendors carry that documentation and audit defense as part of the contract. An in-house open-source stack has to build it. For carriers with a Compliance Officer reviewing the antifraud-plan filing, the documentation overhead usually pushes the decision back toward a commercial vendor.

A full re-evaluation every 24 to 36 months tracks how fast the category is moving. Three triggers should pull it forward: a state DOI antifraud-plan filing flagged for review; a loss-ratio deterioration on a specific line where fraud is concentrated (workers comp, auto BI, property); or a new CCO, CFO, or Claims VP reviewing fraud-tech as part of a 100-day plan. The detection category itself is shifting from rules and scoring toward agentic AI; Shift Technology's handler-assist agents and the emergence of investigation-layer vendors are the two largest 2026 changes. Carriers who last re-evaluated in 2023 will see a meaningfully different shortlist today.

There is no single answer. Verisk ClaimSearch and ClaimDirector cover all three with depth, anchored on the industry-utility data set, and Xactimate adds property-specific estimation across 22 of the top 25 US property insurers. FRISS has historically been strongest on auto scoring. Shift Technology covers auto and property with European-deployment depth, including the AXA Switzerland deployment that processed over 1M claims and stopped more than €12M in fraud. BAE NetReveal's ring-detection lineage fits organized workers-comp medical billing rings. CCC anchors auto specifically. The right answer depends on which line drives the largest share of your loss ratio.

Downstream of detection. Every platform in this roundup ends its job at the flag; the per-claim investigation that follows is the manual SIU workflow at 14+ days per case and roughly 25% coverage of the flag queue. An investigation-layer vendor - Hesper AI is purpose-built here - takes flagged claims from FRISS, Shift, or Verisk and runs the SIU playbook autonomously, compressing the cycle to 2-4 hours and lifting coverage to 100% of flagged claims. The model is not replacement of any detection vendor on this list. It is the layer none of them ship. See how this plays out against Verisk specifically.

← More articles on the Hesper AI blog

See Hesper AI on your documents

Request a demo and we'll run an analysis on your real document samples.