Graph Adoption in Property and Casualty Insurance
- Blog >
- Graph Adoption in Property and Casualty Insurance

This is an abbreviated version of a presentation by Sai Burra, Abhay Solutions, during the Graph + AI Summit 2021 Conference.
Many insurance companies have multiple lines of business: this can include auto, residential, commercial, casualty insurance, and more. It’s not unusual for there to be 30 to 35 lines of business in large insurance companies, and even mid-tier companies can have six to 18 lines of business. Insurance companies also interact with third-party organizations, such as credit bureaus, the DMV (Department of Motor Vehicles), repair shops, parts vendors as so on. So we’re talking about 50 to 60 integration points for an insurance company.
Insurance companies use a multitude of applications across the organization, including applications in policy centers, claim centers, customer service, along with agent portals and more. Each of these applications is collecting data. So all this has to pass through in multiple applications through multiple sources, which makes it very challenging to collect the data for analytics.
Plus, analytics are often run in silos: there can be analytics for claims, analytics for underwriters, analytics for agents, analytics for policy setters, etcetera—and what has proven very challenging for insurance companies is bringing all of this data together to provide a unified view. In one context, this could be a unified view of the customer, across multiple silos; in another context this would be a unified view of claims, across different lines of business. Insurance companies want to unify these large sets of data.
The following diagram shows a variety of applications and analytics capabilities interplay in a hypothetical insurance company:
Achieving a unified view of its data is a key goal for an insurance company, but the amount of data, the myriad applications, the differing needs of various departments, the wide array of analysis, all present obstacles.
Now, let’s pivot and examine how graph and AI and ML can help insurance companies overcome these challenges and accomplish their goals.
One of the first things to talk about is relationships: a graph database stores data, and the relationships between data. Graph analyses these relationships and uncovers insights. Artificial intelligence, with machine learning, once adapted to run on graph data, will help insurance companies by yielding more insights. And TigerGraph, in particular, has the ability to integrate data from multiple silos, to scale without impact to performance, and provide more and better insights.
Take the example of claims fraud. With a graph database, investigators can see the entire ecosystem surrounding a claim: every person, every vehicle, every incident report, every policy. A graph database enables an investigator to deeply explore relationships. There could, for example, be a legacy claim, which is fraudulent—and a connection from the person associated with that claim to a new claimant—the relationship might be twice or thrice removed, but a graph can uncover this insight. So, even if the new claim doesn’t seem to be fraudulent, this connection could warrant closer scrutiny of the new claim.
Graph algorithms speed the ability of investigators to identify connections. Centrality algorithms, for example, can tie information in a database back to a single person, while Community Detection algorithms can determine the data associated with a person, claims.
Artificial intelligence can enable this type of analysis at scale, and machine learning can help the analysis become more accurate and more precise, and more effective over time—and ability to identify patterns, and features associated with different types of fraud, can be used in training the machine.
Moreover, Explainable AI is very important, especially as business decisions, policy changes, customer communication, and so on, are implemented by humans—if the output from an artificial intelligence cannot be understood, that output cannot be used. And if legal action is needed, Explainable AI can be crucial in prosecuting cases successfully.
I can’t think of any better technology than graph as of today, to get this larger view and a better view and a deeper view.
You can listen to the full presentation here.