How AI uncovers fraud in high volume, low value claims
Though inexpensive and not cost effective to investigate, the fraud losses on these claims quickly add up.
Facing continued uncertainty and instability, insurers are greatly challenged to fight emerging fraud and its effects across insurance organizations. In addition to pandemic-related challenges, insurers must also juggle shifting customer needs, behaviors and expectations as customers continue to demand faster and better service from their providers.
With resources stretched thin, special investigations teams don’t have the capacity to review low-value claims (i.e., claims for minor windshield damage) and instead write them off. Though inexpensive and not cost effective to investigate, the fraud losses on these claims quickly add up – especially as fraudulent activity continues to increase.
Tackling fraud and automating claims processing with artificial intelligence, which processes and compares millions of claim details in real time, will alleviate the strain on insurers and proactively avoid fraud that may have otherwise gone undetected – resulting in significant time and financial savings.
Why low-value claims are under-investigated
Documents are submitted along with most insurance claims. A simple review validating names, addresses, and policy numbers for all of an organization’s claim documents translates to a full document management department — requiring the labor of dozens of people.
Analyzing claim documents to investigate for potential fraud adds even more labor — manipulated document metadata, previously used images from other claims, and images or documents that are inconsistent with the geographic location are all flags for fraud that need to be looked out for.
Due to the significant labor involved, as well as the availability of resources, it is simply not cost effective to spend human hours investigating the details of all low-value claims. Investigative efforts will instead be reserved for higher value, materiality and complex claims that are of greater priority — requiring more human touch.
However, AI can be used to not only detect fraud, but to support autopayment decisions — deciding whether a claim can be paid or should be denied due to suspected fraud. The result of which is minimizing investigative costs and recouping significant fraud losses.
How AI will eliminate fraud losses on low-value claims
Fraud is made up of human behavior that differs from what is typical within a group of similar people or similar claims. For example, a claim for windshield damage where the repair costs significantly exceed the typical costs for similar vehicles with similar damage may be indicative of fraud.
Since an individual or claim can differ from peers in varying degrees — i.e., one can be moderately different, slightly different, etc. — all the degrees of difference must be distinguished from every group of like individuals or claims to successfully avoid fraud.
This is where AI comes in. AI systems will autonomously choose the mathematical representations for the degree to which any claim or individual differs from their peers. The system will then perform billions of comparisons, assigning a score for the degree of difference. A score that depicts significant anomalous behavior compared to multiple peer groups makes the case for fraud and abuse clear.
Ultimately, utilizing AI technology means the low-value claims that could not be investigated previously, suspicious or otherwise, can now be taken on by the AI system — alleviating the strain on insurers and saving them from paying millions in fraudulent claims.
About AI-powered claims automation
AI supports straight-through processing by making autonomous claim payment decisions. The system will be leveraged to make recommendations in scenarios such as:
- Should a claim be autopaid or autodenied due to fraud?
- Does the claim require investigation for potential fraud, waste, or abuse?
- Does the claim qualify for accelerated human review or require full manual adjudication?
AI systems will autonomously analyze claim details — identifying outlying claims, people, physical and virtual places, and social networks. Based on the outcome of the automated document verification, the AI system will make a recommendation for whether a claim qualifies for straight-through processing or should be denied due to suspected fraud. The system can even recommend lower-cost human investigative or adjudicative review for claims requiring some element of human touch.
When we let AI software analyze claim details and automate claims management processes, we can increase straight-through processing rates. This will reduce the time it takes to reimburse customers — providing an elevated customer experience — and will minimize unnecessary labor.
With fraud continuing to rise, and opportunistic fraudsters taking advantage of the overburdened industry during these turbulent times, special investigations teams are left vulnerable to the fraud hidden among the high volumes of low-value claims.
However, when AI is leveraged in all aspects of claims management, insurers will significantly reduce fraud and recoup significant losses. The end result is increased profitability, as well as time saved and better spent on effectively servicing customers. Insurers who get started with AI today will be better positioned to compete in the challenging post-pandemic space.
Gary Saarenvirta is Daisy Intelligence’s founder and CEO and a preeminent authority on artificial intelligence. The former head of IBM Canada’s data mining and data warehousing practices, Gary is passionate about AI and its ability to transform how insurers grow their businesses and establish an edge in an increasingly challenging environment. Under Gary’s leadership, Daisy has established a track record of delivering verifiable financial outcomes for a rapidly growing list of clients.
Opinions expressed here are the author’s own.
Related: