Why One New York Life Group CIO Is Choosing An AI Platform Over Point Solutions

When New York Life's CEO challenged every business unit to accelerate AI adoption in late 2023, Group Benefit Solutions was unusually prepared.

After spending the past decade eliminating legacy systems and rebuilding its technology stack, New York Life Group Benefit Solutions (GBS) CIO Matt Marze and team already had the foundation needed to scale AI rather than simply experiment with it. Now, a common AI operating system is quickly bringing the technology into the core of how the business runs.

“We’re not looking at point solutions,” Matt told CIOnews. “We’re looking at creating more of a persona-based interaction model with our users to drive better experiences, guided solutions that are bringing in signals in real time and helping to curate the work and the prioritization of the work.”

A nearly 40-year veteran of the IT industry, Matt spoke to CIOnews about why data is more important than tools, and how his team built an AI operating system that’s helping GBS reconcile payments faster and respond to RFPs quicker.

Cost center to value driver

The strategy of integrating AI instead of latching it on to existing processes reflects the new mandate for CIOs. For much of Matt’s early career, the IT organization wasn’t looked at to create value. Even just a few years ago, it wasn’t uncommon to work under CFOs or COOs.

Now, CIOs are expected to work at the frontlines of where the business makes money. At New York Life, for example, Matt reports into the head of GBS and has a line into the company’s Technology, Data, AI, Ventures (TDAV) organization. And investments are guided by the business value of the technology, not the novelty of the solution.

“IT used to be a cost center,” said Matt. Now, “everything happens locally. You have to have enterprise standards, scale, security and efficiency. But you’ve got to have solutions that make sense, that are fit for purpose for your business or industry,” he added.

For Matt, those are ones that help improve how customers engage with GBS. Insurance is moving from a reactive model where providers simply pay out claims, to one where companies can engage with customers to help prevent claims and drive deeper loyalty. Whether serving brokers, employers, or employees, the expectation is the same: a digital, seamless experience. And data and AI are key to Matt’s plan to differentiate from the rest of the industry.

“We’ve been able to accelerate the journey because of that modernization. Now, we can really start to talk about the next level of good,” Matt said.

Data over tools

Over the last decade, Matt’s team has consolidated and upgraded the IT portfolio.

They shut down hundreds of legacy systems and are down to just four core platforms, including SAP and Salesforce. GBS also relies heavily on AWS. Its SAP system runs through AWS, for example. However, Matt acknowledged the need to be multi-cloud in the AI era. And they ended third-party administrator contracts and brought services back in-house.

The new simplified, integrated environment has helped Matt’s team bring products to market faster than any other time in the company’s history. And they’re able to respond more quickly to regulatory changes.

“Some organizations get too enamored with tools,” said Matt. “If you haven’t figured out how to do strategic data management, spending money on an expensive tool isn’t going to solve your problems. It may mask them in the short term, but you're going to run out of runway,” he added.

Rather than deploying separate AI tools for individual business functions, GBS built a common AI operating layer that different teams can reuse across workflows. Built around over 50 AWS services and open source tools, it functions akin to an integration layer, connecting GBS’ various systems and data.

This includes GBS Data Hub, a unified layer combining analytic and operational data. On top are execution, reusable intelligence, and domain experience layers that include tools for agent-to-agent interaction, security and access controls, memory sharing, chain-of-thought reasoning, and observability.  

For Matt, the lesson is straightforward: AI success depends less on choosing the latest model than on building the data foundation beneath it.

"There's no quick fix for data," he said. "You've got to do the hard work.”