Insuranceciooutlook

DeepFraud: The One-Stop Solution for Fraud Detection, Triage, and Investigation

Michael Saltzman, Co-Founder & COO and Tomas Vykruta, Co-Founder & CEO, DeepFraudMichael Saltzman, Co-Founder & COO and Tomas Vykruta, Co-Founder & CEO Headquartered in New York, DeepFraud’s mission is simple yet powerful: bring cutting-edge technology to the insurance world to increase fairness in claims processing and reduce costs for all participants, from carriers to customers. This mission takes heightened importance in today’s insurance landscape, where fraud detection is a top priority for CIOs. According to recent estimates, almost 10 percent of insurance claims are rooted in fraud, translating to over $40 billion in carrier losses each year. Moreover, fraudulent activities undermine the very purpose of insurance by hindering the timely processing of deserving claims. When executed correctly, fraud detection can make way for refined and fast insurance processes by narrowing down the pool of suspicious claims and boosting the productivity of the special investigation unit (SIU). Meanwhile, companies should make sure that fraud detection efforts don’t hamper their ability to deliver services to rightful claimants. DeepFraud helps carriers address these requirements with its novel digital forensics platform for claims fraud detection.

The Genesis

Detecting fraud has historically been a challenging task for insurers who are burdened with the time-consuming pre- and post-claims processing. A single carrier often processes many tens or hundreds of thousands of claims annually, each with hundreds of pages of additional supporting evidence. Often, the carrier’s inability to examine such a humongous amount of data gives fraudsters the edge.

However, on the bright side, claims data is a gold mine. No matter how unstructured and bulky it may be, it will facilitate a unique and in-depth look at the actors in a claim. “I saw that many insurance carriers had datasets that matched the richness of what we were using internally at Google,” says Tom Vykruta, DeepFraud’s co-founder and CEO, who has AI expertise that he honed at Google over a 9 year period. “I saw a chance to help carriers use it in a manner that is on a par with how Google handles data. There was an opportunity to make data the foundation for decisions, but it wasn’t technically simple,” he adds. This was when Tom joined hands with Mike Saltzman, a Stanford MBA who came from an algorithmic investing background at one of the most successful hedge funds in the world, Bridgewater Associates. Mike shared Tom’s vision, having seen firsthand how data-driven insights transformed Bridgewater.

Mike and Tom spent months talking to insurance executives, claim adjusters, SIUs, and managers. They realized that the folks in the insurance ecosystem were in dire need of data-driven products to carry out detection, triage, and investigation of suspicious claims. “They just kept echoing each other. Even with the tools we’ve gotten over the last few years, the anti-fraud effort was still largely a manual process,” says Mike, co-founder, and COO of DeepFraud. “It was clear that the objective of predictive tools is to make current employees more productive and not replace them,” he adds. With a clear understanding of these requirements, Tom and Mike set out to deliver Google-level technologies and techniques to the insurance industry through DeepFraud.

Solving the Fraud-and-Seek Paradox in Insurance

DeepFraud provides a purpose-built, cutting-edge, scalable, and easy-to-adopt data analytics solution that can augment the effectiveness of fraud detection, triage, and investigation. The company notes that carriers lack the right tools to improve the “hit rates” of identifying and investigating suspicious claims, while also reducing instances of “false-positive” claims. “Most carriers own a tremendous amount of data but their current fraud detection and prevention models—or even the solutions offered by other vendors—go about analyzing just two percent. Our product can analyze and learn from 95 percent of the data,” says Saltzman.

“The key approach we adopt is the same as what Google uses: never make ‘rules.’ Instead, build a system that uses the entire data to determine the cause and effect patterns. It is a huge differentiator for us that we train models on essentially all of the data while the competition trains on just two percent,” explains Vykruta. This helps DeepFraud deliver powerful predictive insights to the decision-makers over the life cycle of a claim, derived from every possible document contained in the application—from FNOL documents to doctor’s notes, and everything in between. “We’ve invested heavily in building proprietary language models that can understand claim-related documents in their own context so the platform can accurately ‘read’ them like a human,” says Vykruta. The result is a system that can find inconsistencies and suspicious patterns instantly across thousands of documents. Unlike other fraud detection offerings in the market that only go as far as identifying anomalous patterns within claims, DeepFraud’s platform is capable of “explaining itself,” by making inferences backed by reason. “All of that work and investment started by understanding the workflow of an SIU. They first start by reading parts of the claim file to boost their context of the claim,” Vykruta adds.


Most carriers own a tremendous amount of data but their current fraud detection and prevention models—or even the solutions offered by other vendors—go about analyzing just two percent. Our product can analyze and learn from 95 percent of the data

In addition to providing evidence and explanation, DeepFraud helps claims processors and managers triage incoming claims for investigation or quick settlement by fraud likelihood, complexity, and other factors, including the probability of attorney involvement. It allows SIUs to identify fraudulent claims and carry out countermeasures that can deliver the highest ROI for investigations.
It also allows managers to view how much fraud goes unresolved, which can help increase the productivity of SIUs as a whole. Managers get access to high-level dashboards and custom reports that provide a birds-eye view of the fraud detection and management workflow. “The platform offers tools to contextualize claims with one-click views, by carrying out proactive analytics from unstructured data, which helps visualize complex data abstractions—much like a financial tool,” Saltzman mentions.

"We’ve invested heavily in building proprietary language models that can understand claim-related documents in their own context so the platform can accurately ‘read’ them like a human"

In a nutshell, by embracing scalable data analysis, simplified visualization, and seamless process automation, DeepFraud enhances the fraud detection teams’ prowess in quickly and accurately detecting and handling fraudulent claims. Achieving a high hit rate on suspicious claims encourages carriers to invest more in expanding their fraud detection teams. In doing so, DeepFraud gives the assurance that the carrier’s efforts are not wasted on a vast pool of falsely flagged claims based on mere surface-level pattern anomalies. It instead relies on “deep” excavation of the narrow yet highly fraud-likely claims, which is necessary to discourage fraudsters and undermine their “hiding under data heaps” tactics. “One carrier that we interacted with went so far as to say that they would increase the size of the SIU staff 10 times if their hit rate went from 20 to 50 percent,” mentions Saltzman.

DeepFraud’s solution deployment model is a key differentiator for the company. The platform’s integration does not require any changes to a client’s workflow tools. Instead, claims data is redirected into a secure private cloud or on-premise data center where DeepFraud performs the analysis. As a result, carriers can go up and running with DeepFraud in a matter of weeks, as opposed to months. Also, all of DeepFraud’s technical staffs are experts in their field, comprising leaders from Google’s AI and speech recognition teams, along with natural language researchers from top universities across the world.

Striding Toward Success

With the right technology stack and strategy, DeepFraud has positioned itself in tune with the favorable startup ecosystem in today's insurance sector. The company recently received a large equity financing from arguably the most successful early-stage venture team in the world, First Round Capital to build a sophisticated engineering team, which includes a number of senior AI engineers and scientists from the world’s leading AI organizations.

With an innovative product and a powerful team to support it, the future looks bright for DeepFraud. “We’re going into Q4 2019 with a phenomenal product that is capable of delivering insights that haven’t been available before. Looking ahead to 2020, we’ll be deploying across the P&C industry within more business lines,” concludes Vykruta.

- Pamela Morgan
    September 18, 2019
Share this Article:
DeepFraud

Company
DeepFraud

Headquarters
New York, NY

Management
Michael Saltzman, Co-Founder & COO and Tomas Vykruta, Co-Founder & CEO

Description
Offers a purpose-built, cutting edge, scalable, and easy to adopt data analytics solution that can augment the effectiveness of fraud detection, triage, and investigation