Hesper AI
BlogVendor Comparison
Vendor ComparisonMay 8, 2026·11 min read·Nitish Badu, COO

Hesper AI vs. Verisk: how the investigation-vs-detection difference plays out

Verisk dominates the detection-data layer; Hesper sits at the investigation layer. Here is how the two pieces fit together in a 2026 P&C SIU stack.

1.8B+
ClaimSearch records
Verisk contributory database
0-999
ClaimDirector score range
Verisk fraud-scoring output
+13.0%
Verisk claims sub-segment Q4 2024
Anti-fraud + property estimating
2-4 hrs
Hesper investigation time
vs. 14+ days manual SIU

Verisk and Hesper get compared because both have the word fraud in their marketing, but they live at different layers of the special investigations stack. Verisk sits at the data and detection layer: ClaimSearch is a contributory database of more than 1.8 billion claim records covering roughly 95% of the U.S. property and casualty market, and ClaimDirector turns those records into a 0-999 fraud score. Hesper sits at the investigation layer: a flagged claim goes in, an audit-ready report comes out in 2-4 hours.

The framing matters because the buying question is rarely Verisk or Hesper. It is whether the investigation step that follows a Verisk hit or a ClaimDirector score is run by people at 14+ days per case or by an AI agent at 2-4 hours. This post walks through what Verisk actually ships, what Hesper ships, the layer-cake of an SIU stack, and the deployment pattern most carriers end up with.

For the broader vendor map, the 2026 AI fraud platforms compared pillar places Verisk in the data tier and Hesper in the investigation tier. This piece is the close-up on the boundary between those two tiers.

Detection and investigation are different layers

Detection produces a signal. Investigation produces evidence and a recommendation. Those are different jobs and they sit on different rows of the SIU stack. A signal can be a ClaimSearch match, a ClaimDirector score above a carrier-set threshold, a NICB Questionable Claim referral, or an internal rules-engine alert. None of those signals decide a claim. They identify which claims need work.

The volume problem is on the investigation side. About 10% of P&C claims involve some form of fraud or material misrepresentation. A typical Top-25 P&C carrier processes around 3 million claims per year, which means roughly 300,000 suspect claims annually. Manual SIU work runs at 200+ cases per investigator open caseload and clears roughly 10 investigations per investigator per month at 14+ days per case. The arithmetic does not close: only about 25% of flagged claims get a real investigation. The other 75% gets a desk decision.

Detection vendors like Verisk are not built to close that 75% gap, and they do not claim to be. Their product page tagline for ClaimSearch is "Fast-track claims and detect fraud," and ClaimDirector's tagline is "Claims Scoring detects fraud quickly and accurately." The verb is detect. The investigation step is left to the SIU team, by design. We covered the full layer-cake in the 2026 SIU technology stack post; this post is the row-level walkthrough of where Verisk and Hesper actually meet.

A score is not an investigation

NAIC Insurance Fraud Prevention Model Act 680 requires every admitted carrier to "maintain effective procedures and resources to deter and investigate fraudulent insurance acts, including a unit that will investigate suspected fraudulent insurance acts." A 0-999 number does not satisfy that obligation. An investigative process does.

What Verisk actually does

Verisk runs three product families that touch P&C claims: ClaimSearch (the contributory claims database), ClaimDirector (the fraud scoring engine that runs on top of ClaimSearch), and the property estimating stack formerly branded Xactware, which includes Xactimate. All three sit in the data, detection, and estimation layers. None of them are positioned as investigation engines.

ClaimSearch (data layer)

ClaimSearch is a contributory database with more than 1.8 billion claim records, covering roughly 95% of the U.S. P&C market, with about 175,000 new claim records added per day from 2,800+ contributing carriers and roughly 200,000 daily users. Verisk has been operating it for 50+ years. Verisk's own positioning piece calls it "the backbone of the P&C claims ecosystem". Through the ISO and NICB partnership renewed publicly in 2024, ISO has operated NICB-acquired auto claims databases since 1998, and Questionable Claim referrals flow through the ClaimSearch pipeline.

ClaimDirector (detection / scoring layer)

ClaimDirector runs analytics and rules on top of ClaimSearch records and produces a 0-999 fraud score. The Verisk Claim Scoring product page describes it as a tool to "detect fraud quickly and accurately." Like every rules and statistical scoring engine, ClaimDirector inherits a 60-85% false positive rate at typical sensitivity thresholds, which is the structural reason carriers staff SIUs in the first place. We unpacked that scoring problem in legacy rules versus autonomous AI fraud detection.

Property estimating (Xactimate, ESL, related)

Verisk's property estimating stack is used by 22 of the top 25 U.S. property insurers and 10 of the top 10 Canadian insurers. Xactimate is a line-item estimating tool, not a fraud product, but it shows up in the same workflows because inflated estimates are one of the most common P&C fraud vectors. The output is a structured estimate; the investigation step that asks "is this estimate consistent with the loss" is not what Xactimate does.

Verisk reported $2,882 million in full-year 2024 revenue with the Insurance segment growing 7.5% and the claims sub-segment up 13.0% in Q4 2024, attributed in the Q4 and FY2024 earnings release to "property estimating solutions and anti-fraud solutions." Carriers are buying more detection, not less. That is the upstream pressure that is making the downstream investigation bottleneck worse in 2026.

What Hesper does at the investigation layer

Hesper takes a flagged claim and runs 15+ investigation phases in parallel: policy and coverage analysis, claim history correlation against carrier and external records, document forensics, behavioral pattern analysis, public-records cross-referencing, and evidence assembly. The output is an audit-ready report with a recommendation and a citation trail back to every signal it used. Time per case: 2-4 hours. Cost per case: roughly $150, against an industry baseline of about $2,500 per manual investigation.

The ingest side is deliberately permissive. ClaimSearch hits, ClaimDirector scores, NICB referrals, FRISS or Shift signals, and a carrier's own rules-engine alerts all flow into Hesper as starting signals. Hesper does not require the carrier to switch detection vendors. It consumes whatever detection layer is already in place and finishes the job. From fraud detection to fraud resolution.

The throughput shift matters more than any single phase. Manual SIU runs at roughly 10 investigations per investigator per month. With Hesper handling the procedural work, an investigator reviews and signs off on 800+ cases per month. Coverage of flagged claims moves from 25% to 100%. The investigator's role shifts from execution to decision-making.

Verisk dominates the detection-data layer; Hesper sits at the investigation layer; carriers increasingly run both.

- Hesper AI product research

Side-by-side: where each fits in the SIU stack

The clearest way to map the two vendors is to put them in the same row-by-row table. ClaimSearch, ClaimDirector, and Xactimate sit in three adjacent rows of the data and detection layer. Hesper sits in a different row entirely.

ProductLayerCapabilityOutputTimeCoverage
Verisk ClaimSearchDataCross-carrier claims database with 1.8B+ records, ~95% U.S. P&C coverageMatch hits with supporting contextSeconds (query)All claims submitted to the contributory pool
Verisk ClaimDirectorDetection / scoringRules + analytics fraud scoring on ClaimSearch data0-999 fraud scoreSeconds (score)All scored claims
Verisk XactimateEstimationLine-item property loss estimating used by 22 of top 25 U.S. property carriersStructured property estimateHours (human-driven)Property claims only
Hesper AIInvestigationAutonomous AI agent runs 15+ investigation phases against carrier + external dataAudit-ready investigation report with evidence and recommendation2-4 hours per case100% of flagged claims

Time per case in this table is not a marketing comparison; it is a reflection of what each tool is built to do. ClaimSearch returns in seconds because it is a database query. ClaimDirector returns in seconds because it is a scoring function. Hesper takes 2-4 hours because it is doing the procedural investigation work that a human investigator would otherwise spend 14+ days on.

Time per case by stack layer

ClaimSearch query (data lookup)seconds
ClaimDirector score (detection)seconds
Hesper investigation (15+ phases)2-4 hrs
Manual SIU investigation14+ days

The carrier deployment pattern: most run both

A typical mid-market or Top-25 P&C carrier already pays Verisk for ClaimSearch contributions and ClaimDirector scoring, and almost certainly pays for Xactimate seats on the property side. Adding Hesper does not displace any of those contracts. Hesper consumes ClaimSearch hits and ClaimDirector scores as inputs and produces investigation outputs the existing detection stack cannot produce.

Run the back-of-envelope on a Top-25 carrier: 3 million claims per year, 10% suspect rate gives 300,000 flagged claims annually. ClaimSearch and ClaimDirector identify them. With manual SIU at 200+ cases per investigator and roughly 10 investigations per investigator per month, the carrier closes about 25% of the flag pile. The remaining 75% of flagged claims, somewhere on the order of 225,000 per year, never gets a full investigation. That is the gap Hesper closes without touching the detection contract.

The regulatory framing tightens this further. NAIC Model Act 680 requires every admitted insurer to maintain a unit that will investigate suspected fraudulent insurance acts. A 0-999 score does not satisfy that obligation; closing only 25% of flagged claims also does not. The regulator wants investigation. Detection plus a thinned-out SIU is a weaker compliance posture than detection plus an investigation layer that touches every flag.

On cost, the right comparison is not Verisk versus Hesper line-item but the all-in cost of a fully-investigated flagged claim. Detection is a fixed-ish subscription cost; the investigation step is variable. At ~$2,500 per manual investigation, fully-investigating 300,000 flagged claims is an arithmetic non-starter, which is why carriers settle for 25% coverage. At ~$150 per Hesper investigation the math closes, and 100% coverage of flagged claims becomes a procurement decision rather than a headcount problem. The procurement framework for the investigation-vendor side of that decision is in our 12-point evaluation checklist for AI fraud investigation vendors.

When carriers pick one and not the other

There are real edge cases on both sides of this decision and they are worth naming directly.

  • Small regional carriers with under 100,000 claims per year typically rely on ClaimSearch hits and adjuster judgment without funding a dedicated SIU. For them Hesper is overbuilt; the investigation volume does not justify it.
  • Carriers that have internalized the data layer through direct NICB feeds, an in-house data warehouse, and proprietary scoring models sometimes skip ClaimDirector but still drown in flagged claims. Hesper makes sense without expanding their Verisk footprint.
  • Property-only specialty carriers that mostly need Xactimate parity may not need either ClaimDirector or a dedicated investigation layer until their volume crosses the SIU staffing threshold.
  • Workers' compensation monolines with high-volume medical fraud exposure ($34 billion in annual U.S. losses per Coalition Against Insurance Fraud) typically need both detection breadth and investigation depth, because the fraud schemes are layered and procedural.

Per the Coalition's 2022 Impact of Insurance Fraud report, the $308 billion annual U.S. fraud loss decomposes into Life $74.7B, P&C $45B, Workers Comp $34B, Auto Theft $7.4B, and Medicare $68.7B. The line-of-business mix shapes which detection layer matters most, but the investigation gap exists at every line, because rules-based scoring carries the same 60-85% false positive structure regardless of LOB. The cross-vendor management-system context is in our claims management systems comparison.

The forward implication for 2026 procurement is uncomfortable for carriers running detection-only stacks: as detection vendors keep growing their anti-fraud lines (Verisk's claims sub-segment was +13.0% in Q4 2024), the flag-to-resolution gap widens every quarter unless an investigation layer absorbs the volume. The carriers that deal with this in 2026 will not pick between Verisk and Hesper. They will run both, and they will measure the SIU function on resolved claims rather than detected ones.

Key takeaways

  • Verisk and Hesper live at different layers of the SIU stack: detection-and-data versus investigation-and-evidence, so the choice is rarely either/or.
  • ClaimSearch's 1.8 billion records and roughly 95% U.S. P&C market coverage make Verisk the default data layer; Hesper consumes those signals rather than replacing them.
  • ClaimDirector outputs a 0-999 fraud score, but NAIC Model Act 680 requires an investigative unit that produces evidence, which is the Hesper-shaped piece of the stack.
  • Verisk's claims sub-segment grew 13.0% in Q4 2024 driven by anti-fraud and property estimating; more detection produces more flags, which makes the manual SIU bottleneck worse.
  • The right cost comparison is the all-in cost of a fully-investigated flagged claim: detection via Verisk plus investigation at ~$2,500 manual or ~$150 with Hesper.

Frequently asked questions

They sit at different layers, so most carriers run both. Verisk is the data and detection layer: ClaimSearch is the 1.8 billion claim contributory database covering roughly 95% of the U.S. P&C market, and ClaimDirector turns those records into a 0-999 fraud score. Hesper is the investigation layer; it takes a flagged claim and produces an audit-ready report in 2-4 hours. A score is not an investigation, and an investigation without cross-carrier data is harder to defend, so the two pieces are complementary. The only overlap is in marketing copy that uses the word fraud. The actual workflows touch different parts of the SIU process.

No. ClaimSearch's contributory database has 50+ years of accumulated cross-carrier claim records and roughly 200,000 daily users; that data layer is not something a new vendor displaces. Hesper consumes detection signals, including Verisk's, and turns them into investigation outputs. If a carrier uses ClaimSearch and ClaimDirector to identify the suspect 10% of P&C claims, Hesper investigates them. Replacing Verisk would mean rebuilding the industry's contributory data infrastructure; that is not the problem Hesper solves. The problem Hesper solves is that 75% of flagged claims never get fully investigated because manual SIU work runs at 14+ days per case.

Yes. Those are typical detection inputs. ClaimSearch hits, ClaimDirector scores, NICB Questionable Claim referrals, and the carrier's own rules-engine alerts all flow into Hesper as starting signals. Hesper then runs 15+ investigation phases in parallel against carrier policy data, claim files, public records, document forensics, and behavioral analysis, and produces an investigation report that cites every piece of evidence it used. Carriers with existing Verisk subscriptions do not rip anything out; they wire the detection feed into Hesper's intake and let the investigation run downstream. This is the deployment pattern across most P&C carriers we work with in 2026.

They are priced for different jobs. Verisk pricing is contributory plus subscription and usage tiers; Verisk reported $2,882 million in full-year 2024 revenue, with the Insurance segment growing 7.5% and the claims sub-segment up 13.0% in Q4 driven by anti-fraud and property estimating. Hesper is priced per investigation, around $150 per case, against an industry baseline of roughly $2,500 per case for manual SIU work. The right comparison is not Verisk versus Hesper line-item; it is the all-in cost of a fully-investigated claim, which is the detection subscription plus the per-case investigation cost combined. Detection alone does not close a claim.

NAIC Insurance Fraud Prevention Model Act 680 requires every admitted insurer to "maintain effective procedures and resources to deter and investigate fraudulent insurance acts, including a unit that will investigate suspected fraudulent insurance acts." A 0-999 fraud score does not satisfy that requirement on its own. An SIU that closes only 25% of flagged claims is also a thin compliance posture. Regulators expect investigation, not just detection. Hesper's investigation layer is built around producing the evidence and reasoning a state DOI or court would accept, which is what 680 actually asks for. Detection plus investigation, run together, is the compliance-aligned shape.

Better detection generates more flags, not more investigations. Verisk's Q4 2024 release explicitly attributes claims-segment growth to "property estimating solutions and anti-fraud solutions," which means carriers are buying more detection. Manual SIU throughput is structural: roughly 10 investigations per investigator per month, 200+ cases per open caseload, 14+ days per case. More detection without more investigation capacity widens the gap between flagged claims and resolved claims. That is the gap Hesper is built to close: 800+ cases per investigator per month at 2-4 hours per case, with 100% coverage of flagged claims.

Edge cases exist on both sides. A small regional carrier processing under 100,000 claims per year may rely on ClaimSearch and adjuster judgment without funding a dedicated SIU; for them Hesper is overbuilt. A carrier that has already internalized the data layer through NICB direct feeds and an in-house warehouse may skip ClaimDirector but still drown in flagged claims, in which case Hesper makes sense without expanding their Verisk footprint. The default for mid-market and Top-25 P&C carriers is both: Verisk for cross-carrier data and detection scoring, Hesper for the investigation step that turns scores into resolutions.

← 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.