What’s driving the global common data capability at RGA

When the board says, “AI!” CIOs have the daunting task of educating it on the various flavors of this capability, and steering them to the most beneficial investments and strategies. Mark Brooks, who became CIO of Reinsurance Group of America in 2023, did just that, and restructured the technology organization to support the platform, redefined the program’s success metrics, and proved to the board that IT is a good steward of the dollar. Here, he walks through the journey and offers transformational CIOs some in-the-trenches advice.  

What role is data playing in RGA’s profitability and growth?

Data is a primary asset to RGA’s growth, and our ability to leverage it is critical to increase the speed and precision of our core business processes, such as underwriting and actuarial. We leverage data to increase the pace of evaluating the clinical information necessary to underwrite on behalf of our partners, and continuous research into new and novel approaches to applying data to our business processes. In the growth area, we also use data to accelerate the cycle time of our external-facing business processes. Our data capability finds global commonality across all our regional solutions.

It takes a lot of change to build a global common data capability. What was your approach to generating the mindset necessary to get this done?

When I joined RGA, there was already a recognition that we could grow the business by building an enterprise data strategy. For years we were using what I would now call an old-school use case to make processes more efficient, less costly, and of higher quality. We were already talking about data as a product with some early building blocks of an enterprise data product program. Since then, we’ve extended that conversation into areas with the greatest potential to generate value.

How did you educate your board about modern uses of data?

I first described the overall AI landscape and made sure they realized we’ve been doing AI for quite a while in the form of machine learning and other deterministic models. I then described what I think of as the three categories of generative AI. The first is FOMO gen AI, which happens when the board reads about AI pilots and says, “We need to do something!” An example is pointing Microsoft Copilot at SharePoint and calling it gen AI. This can cause risk without a clear business case. If Copilot finds a misclassified 2018 payroll file on SharePoint, then it can answer questions about people’s pay. This enforces the need for good data governance, as AI models will surface incorrect data more frequently, and most likely at a greater cost to the business.

Then there’s commercial gen AI, any of the pretrained models from the hyperscalers, which look to consume all the data in the world. These commercial gen AI tools allow people to be more productive in different ways by helping to create a bio or summarize a group of PDFs. While this commercial type of gen AI is a productivity lift, it isn’t enterprise gen AI transformational.

Enterprise gen AI is where the true value is. For example, at RGA, we can create a solution leveraging a fine-tuned large language model by infusing our client’s data with our own, and then upsell their customers with new insurance products reinsured by RGA. That’s gen AI driving revenue. We can bring a client’s data into our data platform and tune a model against it that generates an exceptional outcome.

What needs to be in place for gen AI to work at an enterprise level?

An organization needs the right engagement model between the business’ profit and loss leaders, and the technology groups that are solutioning on behalf of internal and external customers. That’s a critical piece. 

An Agile and product management mindset is also necessary to foster an experimentation approach, and to move away from the desire to control data. The focus should be on the intersection of internal and external datasets, and exposing the data in different channels through partnerships. Ultimately, the opportunity is in creating training datasets that represent the best data associated with a specific business problem. That’s the critical transom to cross.

After years of doing this, as simple as it sounds, the hardest message to sell internally is that documenting processes is critical to enterprise success with AI.

The trick is to use examples, tangible working software, to illustrate possible use cases. Have a capability within IT that has deep enough business knowledge to generate good ideas to pilot, build those pilots in relatively short cycle times, and illustrate the benefit they can produce. Then use that credibility, and the shock-and-awe moment that inevitably comes with real AI use cases, to step into deeper educational conversations. This is why Agile and product mindset matter.

What changes did you make to the technology organization to build and maintain the data utility?

We created four unique capabilities within the technology department. Business engagement, enterprise data, delivery centers, and enterprise architecture.

The business engagement team, organized both by region and business processes, started out by educating their business partners, but now they’ve shifted from education to creating high-value use cases.

The enterprise data function is tasked with providing the global data platform as well as implementing the appropriate data governance. They also have responsibility to build out the critical data products that are core to our business.

The delivery centers are concentrations of technology experts, who together bring a depth of knowledge in everything from legacy programming to the newest technologies, including how to implement gen AI in the solutions. They are RGA’s deepest technologists, but we have intentionally created a partnership network to augment that concentration.

The enterprise architecture function is a new group at RGA. We had architects embedded in each of the delivery teams, and we had a group of advisor architects. We centralized the architects and put more emphasis on common platforms to make sure we’re always being good stewards of the dollar.

One significant change we made was in our use of metrics to challenge my team. We now define two categories of metrics: progress and value. When transforming, it’s difficult to put hard value-improvement metrics in moving from Waterfall to Agile. It’s more reasonable to say, “Here are our progress metrics goals.” Progress metric number one, for example, can be that all teams will have a Scrum master. These metrics can be used to drive steps in the transformation when value metrics don’t work.

What advice do you have for CIOs driving this level of transformation?

One of the first obligations as CIO is to create transparency to garner trust, which always comes down to the money. Establish a financial framework that demonstrates that the global technology group has skin in the game and will be a good steward of the dollar.

Our starting financial framework was to self-fund our transformation with a list of initiatives, which was a progress metric. That doesn’t mean the organization won’t invest incrementally more in technology, but it means we can be trusted to get every bit of value we can from the technology budget.

Once the money hurdle is cleared, then the conversation about a strategic enterprise gen AI program is a completely different conversation.