NEW YORK — Building AI fluency among workers and managing AI costs go hand-in-hand when it comes to realizing AI’s value, executives said Thursday during a panel discussion at the Evident AI Symposium.
Monica Caldas, EVP and global CIO at Liberty Mutual, launched LibertyGPT to build AI fluency, or what she calls “intuition,” among employees. The multimodel interface lets employees interact with a variety of available models and build their own AI-enabled assistants.
LibertyGPT helped the insurance giant expand AI usage responsibly among workers, Caldas said. So far, employees have built more than 30,000 assistants to help with day-to-day operations, she said.
“The whole point was: How do you create fluency across the organization?” Caldas said during the panel. “My philosophy and the way we talk about it in our organization is this idea that you can’t learn how to swim by reading about water.”
Building AI while curbing costs
At Scotiabank, CIO and Group Head Tim Clark wrestled with how to provide the right set of AI capabilities to employees and encourage creativity, but then “bring them back to reality with respect to cost,” he said during the panel discussion.
Scotiabank launched Scotia Intelligence in April with the goal of balancing employee access to AI tools with cost management, Clark said. Scotia Intelligence — which unified the bank’s AI capabilities, platforms and governance — has been rolled out to more than 70,000 employees. The unified approach helped Clark identify areas of value, refine usage and expand offerings, he said.
Security served as the bedrock for Scotiabank’s approach, including outlining how the bank assesses what models are doing and understand model permissions. The bank also built a framework to monitor and increase visibility into AI’s value generation.
The increased visibility has helped the bank make better decisions on which models are best for the task requested of them. This approach can lead to a future where workers have full model selection and can be strategic on where simpler, open-weight models fit the bill.
Banks are adopting tools and methods to manage costs, including using gateways to route workloads to models best suited for completing certain tasks, Sameer Gupta, EY Americas’ financial services AI leader, said during the panel.
However, managing AI costs is a balance between technology and employee behavior, he said.
Liberty Mutual has adopted the forward deployed engineering model to help deploy FinOps capabilities and streamline employee education on using appropriate models, Caldas said. While she supports federated capabilities, Caldas said setting such capabilities up for success involves FinOps, security guidelines and educating the broader organization on responsible AI use.
Providing AI capabilities and training to employees at the outset also helps reduce shadow AI use, Clark said.
“That’s stopped us having lots of people buying random things, asking to procure different capabilities we didn’t think were of value,” he said.
