It's been about a year since Teresa Heitsenrether's team started rolling out LLM Suite, a proprietary abstraction layer through which large language models like OpenAI's GPT-4 are swapped in and out, across the bank's entire employee base.
"We now have about 232,000 people who are entitled to the tool, and more than half of those people actively use it," Heitsenrether told American Banker. "It's become very much a part of the mainstream of how people do their work every day."
As
LLM Suite won American Banker's Innovation of the Year award for 2025. The AI research firm Evident has ranked
Recently, the bank has been adding knowledge bases to
Business value generated from the use of AI at
In the next wave of agentic AI, in which the bank will start to deploy reasoning models that can plan, reason and act autonomously, she expects to see an impact on the balance sheet.
"We're in the early stages of that," she said.
There's value to figuring out where you can insert an agent into an existing process to make it work better, she said.
Deploying agentic AI will require workforce education and technology infrastructure, including for access control for the agents, she said. "If you have machine-to-machine agents able to make judgments and take actions, you have to permission them and make sure you have the right guardrails around what they can and can't do, and how many autonomous decisions they can make. All of that is different, and doesn't exist yet."
Heitsenrether has also been championing data quality throughout the bank, which holds an exabyte of data and moves up to $15 trillion of cash each day.
Policies have to be translated such that an AI agent can present user-friendly answers to customers. "When it's an agent, they don't work well with ambiguity," Heitsenrether said. "So it really does need to be very clear, this is what I'm looking at, and it's this version of the contract, not the three other previous versions that are stored in the database."
