With billions of dollars in their care, local credit unions and banks are building different kinds of guardrails to keep AI from creating risks, even as they look for productivity boosts. iStock illustration

As Massachusetts’ banks and credit unions increase their usage of artificial intelligence, building guardrails will be essential to keeping customers safe, executives say.

First-quarter earnings calls highlighted just how much local lenders are leaning into the use of AI.

Eastern Bank executives said they will use the technology to better serve customers and increase employee efficiency. Rockland Trust leaders touted a new Office of Digital Innovation along with a “governance framework” for the bank’s AI usage. That framework includes a steering committee.

“This will allow us to make AI investments in those areas that have a meaningful payback and avoid the proverbial boiling the ocean,” Rockland Trust President and CEO Jeffrey Tengel told stock analysts April 17. “I expect us to start with some relatively easy use cases as we build muscle memory over time. This should enable us to gain confidence and our ability to execute and take on bigger, more impactful applications.”

Why AI Is Hot, and Getting Hotter

Talk of increasing AI uses is likely to grow, leaders at Massachusetts’ banks and credit unions say, as institutions find more use cases.

“As this technology gets smarter, as people become more competent at using it, you’re just going to continue to see these applications expand,” Metro Credit Union COO and Chief Strategy Officer Traci Michel said. “I think there’s going to be more conversation about [AI], rather than less, even though it feels like it’s everywhere already.”

Part of the reason for increased chatter surrounding AI at local financial institutions could be their digital technology providers. These firms, which provide software for everything from managing loan processes to payments, are increasingly including AI in the products that banks are already using.

“What we’re seeing is vendors are desperately trying to put in AI components,” Hanscom Federal Credit Union President and CEO Peter Rice said “That is like the Trojan horse, that’s the danger. They aren’t systems. They aren’t totally AI-designed. They’re just AI components in already existing processes.”

And financial institutions who took a more patient approach in the early years of corporate America’s AI enthusiasm are now ready to dive in.

“We kind of consider ourselves like a fast follower,” Beacon Bank Chief Systems Officer Ryan Melle said. “We’re not behind the ball, but we’re just making sure that as we’re implementing things, we’re doing it right. We’re making sure there’s proper governance, there’s a reason being why we’re doing this stuff.”

How Do You Manage Adoption?

Hanscom FCU is taking a somewhat unusual step of hiring a chief AI officer, Lenai Camacho, to manage its use and implementation of the technology. Camacho, a policy and compliance expert, will also oversee new guardrails regarding its usage, Rice said.

“That’s where the greatest understanding of how to harness AI is for us at Hanscom,” he said. “Understanding the guardrails, privacy, the firewalls that are needed to make sure that our members information remains safe and unexposed.”

Metro Credit Union is taking a more traditional approach and treating AI like any other technology. It’s using the same risk management framework it would use for any other technological tool or provider, Michel said.

“It’s definitely an emerging technology, but it fits pretty neatly into the risk management frameworks that we’ve been using for years to manage the business,” she said. “As we looked at being an AI-forward organization, we approached it from a policy perspective, a procedural expectation from our staff on a day-to-day basis. We do risk assessments like we do on any other technology that we might participate in.”

Who’s Responsible for AI Actions?

As Metro employees continue to experiment with AI, the credit union is making sure that it is auditing and monitoring AI usage. That kind of governance is a priority for Beacon Bank, too, said Melle, due to the rapid nature of AI development. That includes establishing acceptable uses and standards for employees.

Metro’s Michel noted that while credit union leaders hope AI can aid employees, ultimately the lender is implementing the software with human oversight. Employees need to ensure that guidance or outputs put out by AI are correct, she said.

“We also have drawn a really clear line that we can use AI to assist our business, but humans are accountable for decisions, and humans are accountable for output,” Michel said. “You won’t find anyone at Metro pointing a finger at the computer and saying, ‘Oh, the computer got it wrong.’ Our people are responsible for the work that we do.”

But with technological providers implementing more AI tools, setting up internal guardrails for what AI can and can’t do will be even more important, executives interviewed for this story said. Now, financial institutions not only need to evaluate the technologies that they use, but the AI inserted into the back ends of those products.

“When you put so many band aids on things, how do you govern it?” Hanscom FCU’s Rice said. “How do you control your information if you’ve got so many different pieces of AI operating within your system, and how do you know that that vendor has firewalled their AI?”

Sam Lattof

Vendor Oversight a Key Step

Vendor oversight is important for financial institutions right now, Melle said.

“As your vendors start to leverage AI, you make sure that you have proper vendor due diligence in place,” he said. “Everyone’s documenting where those AI use cases are internally. So it’s really just making sure that you have the proper oversight of where AI is being used, and having some type of committee to have that oversight when a use case wants to come in.”

For Metro, this meant updating the questionnaires that staff send out to technology providers to make sure products measure up to its own internal standards. Beyond looking at the product, Metro staff look at the underlying model to evaluate its risk and how it will be using its data.

But credit union leaders want to ensure they don’t stifle innovation while being able to maintain the trust of Metro’s customers, Michel said.

“We have to maintain the trust that we have with our members and with our employees,” she said. “So, we’re thinking very hard about which guardrails we put in place. The flip side to that is we don’t want to stifle experimentation. We don’t want to be sitting in the back seat waiting for this to happen to us. We want to be a part of the process and driving how it happens.”