AI Fraud Detection Tools: Shift vs Friss vs SAS

Insurance fraud costs U.S. carriers an estimated $80 billion annually, according to the Coalition Against Insurance Fraud. That number has pushed nearly every mid-size and large carrier to look at AI-assisted fraud detection over the past five years. The three names that come up most consistently in claims operations conversations are Shift Technology, Friss, and SAS Fraud Management.

I spent three years in a special investigations unit at a regional P&C carrier before moving into a vendor evaluation role. We ran structured pilots on Shift and Friss, used SAS as the incumbent platform, and eventually had to justify a platform decision to the executive team. Here is what the experience actually taught me — not the pitch decks.

Comparison Criteria and Methodology

Evaluating fraud detection tools is harder than evaluating most software because the ground truth is delayed. A claim you flag today may not be confirmed as fraud for six to twelve months, after SIU investigation, potential litigation, and sometimes criminal referral. That lag makes A/B testing messy.

For our evaluation, we focused on five criteria:

1. Alert precision — What percentage of flagged claims were confirmed fraud or suspicious after SIU review? High volume of low-quality alerts burns investigator time and creates friction with legitimate claimants.
2. Recall on known schemes — Did the system catch fraud patterns we already knew about from our own case history?
3. Integration complexity — How long did it take to connect to our claims management system (Guidewire), and what ongoing data pipeline maintenance was required?
4. Explainability — Could the alert output justify an SIU referral? Could an investigator understand why a claim was flagged?
5. Total cost — Licensing, implementation, and the internal labor cost of managing the platform.

We did not evaluate detection of entirely novel fraud rings the system had never seen. That is a real capability difference between vendors, but it is nearly impossible to measure in a short pilot.

Feature-by-Feature Breakdown

Shift Technology

Shift is a purpose-built insurance fraud detection platform that started in France and expanded aggressively into North America. Its core product is FORCE, which handles fraud detection across claims, policy underwriting, and recoveries.

What it does well: Shift’s network analysis is genuinely useful for organized fraud rings. The system builds relationship graphs across claimants, attorneys, body shops, and medical providers, and it surfaces suspicious clustering that rule-based systems miss. In our pilot, it flagged a medical provider billing pattern that our existing rules had never caught. The provider turned out to be connected to a larger scheme across multiple carriers.

Alert quality: In our pilot on property claims, Shift’s precision rate on referred cases was roughly 38% confirmed fraud or suspicious activity after SIU review. That sounds low, but our baseline rule-based system was running at around 22%. Better, but not transformative.

Explainability: Shift produces score explanations that list contributing factors in plain language. Investigators found these usable. The explanation reads something like “claimant has prior claims at three carriers in 18 months” rather than “anomaly score 0.87.”

Integration: Shift uses REST APIs and has pre-built connectors for Guidewire and Duck Creek. Our integration took about 14 weeks including data mapping and testing. Ongoing maintenance required a part-time technical resource.

Pricing: Shift does not publish pricing. Based on our negotiation, mid-size carriers (500,000–2 million claims annually) should expect licensing in the range of $400,000–$900,000 per year depending on claim volume and modules licensed. Implementation services were quoted separately at $150,000–$300,000.

Friss

Friss is a Dutch company that focuses specifically on insurance fraud detection, risk assessment, and compliance. Unlike Shift, which started in claims fraud, Friss has historically been stronger on the underwriting fraud and policy-issuance side, though it has expanded its claims capabilities.

What it does well: Friss is strong on data enrichment at the point of underwriting and FNOL. It queries external databases — public records, credit bureau data, prior claim history — and compiles a risk score before an adjuster ever touches the file. For carriers with significant policy fraud exposure (applications with misrepresented risk), Friss adds value earlier in the lifecycle than Shift.

Alert quality: On claims fraud specifically, Friss’s detection quality was closer to our baseline in our evaluation. The precision rate was roughly 26% confirmed fraud or suspicious — better than nothing, but not a significant improvement over existing rules. The stronger signal came from the underwriting integration, where Friss caught several high-risk policy applications we would have written.

Explainability: Friss provides a fraud score with factor breakdowns. The interface is simpler than Shift’s but adequate for referral documentation.

Integration: Friss offers connectors for major platforms but its integration documentation was less mature than Shift’s at the time of our evaluation (2023). We spent about 18 weeks on integration, and we encountered data mapping issues with our policy system that required vendor support to resolve.

Pricing: Friss prices by module and claim volume. For a claims-focused deployment, mid-size carriers should expect $200,000–$500,000 annually. The lower price point is real, but the lower detection lift on claims fraud means the cost-per-caught-case may not be better.

SAS Fraud Management

SAS is the incumbent in enterprise fraud analytics. SAS Fraud Management is not an insurance-specific product — it is used across banking, healthcare, and government — but it has deep insurance deployments, particularly at large carriers.

What it does well: SAS is the most configurable of the three. If you have a data science team and the willingness to build custom models, SAS gives you the tooling to do it. The platform supports both real-time scoring (sub-second API calls) and batch processing, and its model management capabilities are mature. Carriers with existing SAS licenses for other analytics work can often expand into fraud detection without starting from scratch.

Alert quality: SAS’s out-of-the-box models for insurance fraud are not its strength. Carriers that get good results from SAS typically have spent 12–24 months building and tuning custom models on their own claims data. The platform is powerful but it does not deliver value off the shelf the way purpose-built vendors do.

Explainability: SAS supports model interpretability through its own tooling and integrates with external explainability frameworks. For regulatory and SIU use cases, this is workable but requires more setup than Shift’s built-in explanations.

Integration: SAS’s integration story depends heavily on your existing technology stack. Carriers already in the SAS ecosystem have a much shorter path. Greenfield deployments are complex. Plan for 20–30 weeks for a full claims integration if you are starting from scratch.

Pricing: SAS enterprise licensing is complex and highly negotiated. A mid-size carrier deploying SAS Fraud Management as a standalone product should expect $500,000–$1.5 million annually including platform access and support. Large enterprise deals are negotiated differently. Implementation services — whether SAS Professional Services or a partner — add significant cost.

Pricing Comparison Table

All figures are estimates based on published case studies, public discussions in industry forums, and direct vendor conversations. Actual pricing will vary significantly based on deal structure and claim volume.

Best For Recommendations

Shift Technology is best for: Mid-size to large carriers whose primary exposure is claims fraud, particularly organized rings and medical provider fraud. If you want a working system in under six months without a data science team, Shift is the fastest path to measurable detection lift. It is also the strongest choice if your SIU team is the primary end user — the investigator-facing interface and explainability are built for that workflow.

Friss is best for: Carriers with significant underwriting fraud exposure or carriers that want to address the full lifecycle from application through claims. If you are a smaller carrier with budget constraints, Friss’s lower price point and module-based structure let you start with underwriting risk and expand later. It is also a reasonable choice if your claims fraud volume is lower and you need the underwriting detection more than claims detection.

SAS Fraud Management is best for: Large carriers that already have SAS in their analytics stack and a data science team capable of building and maintaining custom models. If you are a greenfield SAS deployment for fraud detection only, the cost and timeline are hard to justify against purpose-built alternatives. SAS earns its place in organizations where fraud detection is one part of a larger analytics investment, not the primary use case.

Honest Verdict

None of these three tools is a set-it-and-forget-it solution. All of them require ongoing model governance, alert calibration, and investigator feedback loops. The biggest driver of detection quality is not which vendor you pick — it is whether your organization commits the internal resources to maintain and improve the system after go-live.

That said, the platform choice matters. Shift is the most complete purpose-built insurance fraud detection product as of early 2026. Its network analysis and out-of-box model quality are ahead of Friss for claims-focused deployments, and its implementation timeline is shorter than SAS. The premium pricing is real, but for carriers where claims fraud is a material loss driver, the return on investment is defensible.

Friss is a credible alternative if your fraud exposure is more weighted toward underwriting and policy fraud, or if budget constraints make Shift’s pricing untenable. The claims fraud detection lift is less dramatic, but the underwriting capabilities are legitimately strong.

SAS is the right answer for large carriers with existing SAS investments and data science capacity. It is the wrong answer for carriers that want results in the first year without a significant internal build.

The question I always ask carriers before they sign anything: what is your current confirmed fraud recovery rate, and what does a 5-percentage-point improvement in detection precision actually mean in dollars? Run that math before you commit to a platform decision. The number usually makes the vendor pricing conversation much simpler.