Investigation Copilot for Insurance Fraud

[#1]

Reducing manual processes. Improving investigator efficiency. Streamlining investigations.

Nick Feast, Solution Lead, SAS Fraud Decisioning for Claims

Investigation Efficiency: The Next Competitive Advantage

While advances in fraud detection technology have enabled insurers to identify suspicious activity in real time and at scale using advanced analytics and AI, fraud investigation remains a highly manual and resource-intensive process. Investigators often need to source data from multiple systems, look for patterns and connections across structured and unstructured data, analyse entities linked to an alert, explore networks to uncover hidden relationships, review interview transcripts, document findings and build a case file which confidently supports any decisions that may arise downstream of their investigation. These activities all take time, limiting the number of cases that can be reviewed by even the most efficient Special Investigation Units (SIUs).

Many investigators don’t have the luxury of a dedicated investigation platform to assist with these tasks. Instead, they rely heavily on coordinating their investigation across spreadsheets, emails and other office applications. This can introduce additional inefficiencies, meaning investigators spend more time on mundane tasks and information gathering than on actual investigation. The recent surge in AI-enabled fraud has added fuel to the fire, further hindering the investigation process, as insurers are now having to spend time scrutinizing images and documents to prove their authenticity.  As a result, insurance fraud investigations are becoming increasingly time-consuming and expensive, putting unwanted pressure on SIUs and impacting their ability to efficiently analyse and investigate a growing number of alerts.

It's therefore no longer enough to only effectively detect fraud, the real value comes from how quickly and efficiently an insurer can investigate it. Insurers that can shorten the time between fraud detection and investigative action will enable their SIUs to process more alerts, accelerate case progression and generate greater value from their fraud operations. It’s therefore no surprise that insurers are considering how AI can help to reduce the manual burden on investigators, allowing them to do more with less.

Transforming the Investigator Experience

SAS Fraud Decisioning for Claims includes SAS Viya Copilot, which delivers an investigation copilot and related AI capabilities designed to improve the day-to-day work of the investigator. An AI-assisted search allows investigators to define and execute searches using natural language rather than spending time manually constructing queries, whilst AI generated text summaries provide investigators with automated and detailed information about a particular record or object. No longer do investigators have to navigate multiple pages or trawl through all the elements of a claim or case to understand the details, regardless of its complexity, as SAS’ investigation copilot can pull this information together in seconds, with minimal effort from the investigator.

Figure 1: An AI-generated summary has been included within the Investigation Details page, providing the investigator with a detailed narrative about the investigation, along with suggested next steps.

The real power of the investigation copilot lies within the conversational layer it introduces into the investigation process. Embedded directly within SAS Visual Investigator, the copilot is not only able to interact with investigators through natural language, but also understands the context of the investigation itself, including the claims, entities, relationships and records currently being reviewed. By enabling users to analyse, summarise and document information through natural language interactions, the copilot helps shift effort away from routine and repeatable tasks to where it adds the most value: making investigative decisions. Imagine being able to ask, "What is suspicious about these claims?", "Show me the key connections in this network” or “What can you tell me about this supplier?” and receive an immediate, contextual response. That's the value of SAS’ investigation copilot.

Figure 2: An investigator has asked SAS Viya Copilot  to provide information on the supplier, Fast Cash Auto Repair, that features in the bottom right of the network. The copilot has returned a short summary, including reasons as to why this organisation may be a concern.

Beyond Productivity

As the fraud landscape continues to evolve and new modus operandi (MO) are being trialled by increasingly AI-savvy fraudsters, anything that insurers can do to empower their fraud investigators to work more efficiently is a win. AI assistants that can organise and summarise disparate information, analyse complex networks and produce well-structured consistent reports should therefore be a welcome addition to any SIU. However, the value of an investigation copilot extends beyond improvements in investigator productivity.

Much of an investigator's ability to identify fraud is shaped by their experience, expertise and understanding of known fraud patterns. This is undoubtedly a strength, helping investigators quickly recognise familiar red flags and focus their efforts on areas of greatest concern. However, experience can also naturally influence how data is interpreted, potentially leading investigators to view a suspicious claim, network or pattern through a particular lens. As fraud continues to evolve, this can make it more difficult to recognise emerging behaviours that do not fit established fraud typologies.

An investigation copilot can provide an additional perspective. It can analyse large volumes of data to identify unusual behaviours, relationships and indicators that may have otherwise gone unnoticed. By complementing investigator expertise, it can help to uncover new insights, provide additional context, support decision-making and potentially reveal the early indicators of new and emerging fraud MOs.

Where follow-up actions are required, the AI assistant can also help to support the investigation workflow by identifying intelligence gaps, recommending the necessary investigative steps needed to fill them and even initiating elements of the tasking process, again assisting the investigator to manage and progress the investigation as efficiently as possible.

Looking Ahead

Although the use of AI assistants within insurance fraud investigations is still relatively recent, the impact they are already having on investigator productivity is significant. As these capabilities continue to evolve, they will increasingly be viewed as trusted allies, helping investigators access information faster, analyse evidence more efficiently and document findings with greater consistency.

However, the future of fraud investigation is unlikely to be fully autonomous. Human judgement, experience and domain expertise remain critical components. The greatest value will therefore come from combining the strengths of human investigators and AI, allowing each to focus on what they do best.

About the author: Nick Feast is a Global Insurance Fraud Solution Lead at SAS Institute, with 20 years' experience spanning crime analysis, fraud analytics, fraud prevention and investigations. Nick supports insurers, fraud bureaux and industry consortiums across Property & Casualty, Life and Health markets, providing strategic guidance on fraud detection and investigation.