Think the use of Generative AI and large language models in insurance is still off in the future? Think again.
Brands as big and well-known as Equitable are using GenAI now. The New York-based company provides life insurance, retirement planning, employee benefits and financial advice. It’s a big business; Equitable has assets under management worth $754 billion. And as Eric Colby, the company’s chief technology officer, recently explained, Equitable is also jumping into GenAI.
Colby made his remarks at the recent ITC Vegas event for the insurance industry. There, he was part of a lunchtime workshop presented by DXC Technology entitled “Understanding the World of AI in Insurance: Large Language Models, Foundational Models and Responsible AI.” Colby was interviewed by Eddie Jones, director of Insurance market strategy at DXC. The following has been edited for clarity and length.
A conversation with Eric Colby
Q: How is Equitable approaching GenAI today?
A: We’re very bullish on GenAI. Personally, I’m super-excited about this technology. Do you remember the first couple of prompts you put into ChatGPT and how it absolutely floored you? For me, it was a seminal moment, the closest thing I’ve seen to magic. It’s a game-changer.
So at Equitable, we’re doing a lot. For one, we’re working with partners including Microsoft and DXC to build out a number of different AI solutions. Our first goal was to democratize this capability. We wanted to get it to the masses, to as many people as possible. So we created a privatized version of ChatGPT for the organization. And we now have a lot of AI PoCs [proof-of-concept projects] too.
Beyond that, we’ve got 10 or so focused use cases. One is reinsurance treaties. These are contracts that can be hundreds of pages long, and sometimes, people haven’t looked at them in decades. Initially, we brought in an army of people from a high-priced consulting firm to read these documents, summarize them and extract the data. But then we suspected we could use Generative AI to do this. And the technology actually does a good job
Q: Are you using GenAI for knowledge management?
A: Yes. With traditional knowledge management, you would ask a question and then get back three documents that tried to answer a question. If you were lucky, it would show you the phrase in those documents that would answer the question. But we found that people were calling our service centers and getting different responses based on who they called and when they were using the knowledge-management tool. In part, that’s because our products are complex.
Now we find that GenAI can actually read the articles, understand the nuances, and say, for this product, here’s the answer. And for that product, here’s that answer. So, we’ve built a framework and replicated it across the organization.
Q: When it comes to adopting GenAI in insurance, what do you see as the biggest impediments?
A: We’ve heard a lot about AI bias, hallucinations and security issues. Those are certainly out there. But eventually, we’ll get over those issues by using various techniques.
For me, the massive issue is change management. There are two camps. One, those who have a defined process. They’ll essentially be forced to use AI because it will be built into their workflow. And two, those who do everyday tasks. That’s going to be the big one. Yes, it’s where a lot of the value is. But we’re asking people to change the way they work, practices they may have built up over decades.
I’ll use myself as an example. I recently got Microsoft Copilot [an AI “companion” that works with Microsoft products], and the first thing it makes you do is upgrade your version of Outlook. But it looks very different! Well, I didn’t last a day, even though I was one of the most enthused to use it. So that’s a big thing: How do we get over that initial change management, changing the way people work?
We almost have to trick people into using AI without them realizing they’re using it. And to do that, it’s got to be good on the first try. Because if they don’t get what they want, there’s going to be a very quick retraction. Instead, we want an experience like that initial query of ChatGPT I mentioned, which absolutely floored me. If that’s the experience people have with AI, they will adopt.
Q: How about GenAI use across the insurance industry? What are you seeing?
A: Initially, the large insurers were reluctant. It was like, “Hey, we’re in a regulated environment.” Also, they were concerned about AI bias and hallucinations. But about five months ago, it suddenly turned. A lot of that came from the top. You’ve got executives using ChatGPT, and they saw the value of it very quickly. That was a big shift.
Q: Are there any current AI use cases you’d consider transformative?
A: I haven’t seen any yet, but I can easily think of a lot that would. For example, AI could change the way we build IT systems. I’m not just talking about code generation. Although that’s fantastic, it’s only the tip of the iceberg. The hard parts are system integration, networking, security and the ecosystem your code lives in. They’ve always been the barriers. But that will be different with GenAI.
One example is how hard it’s been to organize and structure different types of data in a CMDB [change management database]. Now that an AI can read unstructured data and make sense of it, that’s going to revolutionize the way we build systems. You could have, say, 300 systems that today have all these user screens and reports, and with AI that could all become a prompt. All of sudden, there’s no more 300 systems with 10,000 screens, no more portal, and no more dashboard. It’s just one question: How do I perform this transaction? The AI will figure it out.
Q: How about within specific insurance applications? How is GenAI likely to be transformative there?
A: I believe this technology will transform the insurance distribution channel. Here at Equitable, every year we have the goal of increasing our internal salesforce. Yet every year, we hire and onboard 1,000 new advisers, and every year we also shed 1,000. We’re just treading water. The reason is, our products and services are complex, and that can be difficult for our new advisers.
This is an area where we could use one of GenAI’s superpowers, namely, reducing time to mastery. According to a recent study, when it comes to learning new tasks, GenAI is most helpful for the lowest performers. After all, they’re the ones with the biggest knowledge gap. We could apply that same concept to bring up the capabilities of our newer and lower-performing advisers. The AI could understand our products and services, and it could understand how to articulate a rational proposal. That could really help people elevate their capabilities, and that would be a game-changer.
Do more:
- Listen to Eric Colby’s ITC Vegas fireside chat discussion. [URL TK]
- Read an insurance AI discussion with Brian Bacsu, director of architecture and platform engineering at DXC.
- Explore DXC’s offerings for insurers.
Eric Colby is the Chief Technology Officer (CTO) for the Equitable. He is responsible for the technology strategy, enterprise/solution architecture, enterprise data management and analytics, enterprise automation/security/middleware services, and the execution of enterprise IT projects.
Eric's 25+ years of professional work experience include positions at AIG, Continental insurance, an internet start-up, and various consulting assignments.
Eric earned his bachelor’s degree in computer science from Rochester Institute of Technology, and holds a CISSP certification and Finra Series 99 registration.
