CNA brings data into the cloud, uses AI for predictive analytics

Digital Insurance spoke with Gary Haase, executive vice president and chief operations officer at CNA. He joined CNA in fall 2021 after 13 years at Catalina, which included roles as group COO and as CEO of its U.S. and Bermuda units. At CNA, Haase’s greatest concern is better managing the wealth of data the company has from its long history in the insurance business, using all the technology now available, such as cloud computing.

When you joined CNA, what systems did the company have to handle the historical data, and what did it need to get the most out of that data?

Gary Haase, executive vice president and chief operations officer, CNA

As you can imagine, a 125-year old company goes through many permutations over the years with different systems and different lines of business. We have a lot of systems producing a lot of data, but like many insurers, we have begun a journey to the cloud. We see the cloud as the future. We are taking the data from all of these different systems and putting it in the cloud, into a major cloud partner, through a migration strategy. Once in the cloud, you get tons of benefits in security, scalability and computing power. The cloud's really been our secret weapon to harvest all this old data and make it consumable across the enterprise.

How did CNA go about migrating data to the cloud? Did the company build its own technologies or work with intermediaries?

We collaborate with some of the major consulting firms, but we really focus on an assembly line approach. We had a three-year plan to migrate all of CNA’s data into the cloud, working with different collaborations. We automated our process to get to the cloud, increasing our velocity of data migration by 10 times over the last 12 months. We've essentially gotten 90 percent of the way there in half the time. Using these automated approaches and leveraging our cloud partners, we can use Google BigQuery, essentially a giant data lake in Google’s cloud where we migrate our data.

It's been a very coordinated approach. We've been intentional in terms of the order with which we've migrated certain data streams. Once data is in Google BigQuery, we build a common model over the top of it, where we can take different data from many different claims systems, stitch it together and line it up in a way that it's all in a common format and can be consumed across the enterprise. We've done it again with this assembly line approach with a lot of thought about the order and the velocity.

Getting the data into the cloud is the early part of the battle. We're very proud of that accomplishment, but really what we want to do is use that data in the cloud to apply advanced concepts like artificial intelligence and machine learning – and really build predictive analytics and predictive insights that might not have been available in our old environment. The goal here is to gain new insights into the business. There’s a lot of potential in claims. With tons of data in claims, you can build predictive models to do things like triage certain claims to detect fraud. That’s just scratching the surface.

What do customers want to see in terms of how advanced an insurer is with customer service technology?

It's really changed and it's a multifaceted approach. Obviously, you can interact with humans one-on-one. Optimizing that call center experience for our customers, for our insured when they're getting a new policy or getting service on a policy, is really important – as is how they interact with us digitally through various portals. Chatbots help us automate the experience and generate additional efficiencies.

What issues do you find in claims reporting and how are these being addressed with technology improvements?

The buzzword in the industry right now is social inflation. We're seeing very large jury verdicts in certain cases that were not there in the past. The rate of inflation for these liability-type lawsuits is much higher than typical inflation. We’re trying to get ahead of that, by understanding what drives these big claims that have the potential to become very large verdicts, and by building models that give us insights into the data, to target those early and be creative at settling them.

How is CNA applying AI to service claims more easily and accurately?

AI has huge potential to automate a lot of the process. Once we input our data in the Google cloud, we now leverage Google Vertex AI. We use that tool to build a model factory. Continuing with our assembly line approach and mentality, one issue many insurance companies face with legacy technology is the actual implementation of models. It's one thing to build them, but then it can be very costly and time consuming to actually implement them into the production environment. The model factory lets us go from weeks or months to implement a new model, down to a matter of minutes. That's been a big win.

What is the potential for cloud computing in insurance? What can be achieved?

We get an additional layer of security. Google has access to huge companies all over the world, and has tremendous knowledge in security. There are more layers of security in the cloud, with the ability to scale. It's much easier in the cloud to scale up quickly and scale down quickly. So there's an additional layer of agility possible. This concept of computing power, the ability to use all of this data at our fingertips, use these tools and produce insights quicker with more granularity is a huge advantage at this point.

We split efforts into evolutionary versus revolutionary. Evolutionary is those building blocks that move things along. Revolutionary is things that are going to truly shake up the industry. Being in the cloud positions us to capitalize on those revolutionary objectives as they present themselves, even if we might not know what they are immediately at this time.