Citizens is working with software provider UiPath, which at its October annual product conference announced new agentic AI tools as part of its automation technology suite. The new capability combines AI agents, robots, people and models to expand the scope and impact of automation efforts, according to the company.

Matt Lavoie, senior vice president of enterprise automation development at Citizens, said the bank plans to integrate some of the AI models developed by Citizens' internal experts with automation software from UiPath.

"Today, we are triaging the work," Lavoie said in an interview at UiPath's annual conference in Las Vegas in October. "Agentic AI would determine what to do, much like the people are [doing] today."

Citizens is an early mover among firms tapping AI agents to unlock business value. A nascent trend, adoption across enterprise platforms is set to take off quickly, Gartner projects. By 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024, allowing 15% of day-to-day work decisions to be made autonomously, according to the research firm.

Citizens already uses automation without AI. For example, the bank uses automation to sort emails and distribute them to teams to execute. Agentic AI will help determine the right action to take, and then potentially hand that decision over to automation software, which would handle that action. 

Bots typically automate repetitive and rule-based tasks, but agents can adapt to changes, make decisions along the way and handle more complex processes. 

"We're advancing enterprise automation with [AI] agents, allowing customers to automate entire end-to-end processes and orchestrate workflows seamlessly. The result is more substantial business outcomes, greater productivity, and more customer-facing direct benefits from automation," said Daniel Dines, UiPath founder and CEO, in a statement.

Agentic AI and automation

Despite the potential of AI agents, Citizens isn't about to let AI agents run wild: Humans will ensure agents perform tasks effectively as it develops use cases.

"Our plan is to keep humans in the loop until we can really prove it's 100% effective," said Lavoie. The focus initially will be on internal, non-customer facing use cases.

Citizens began its automation journey seven years ago. It currently has more than 550 automations running in its environment, performing about 700 employees' worth of work every day, said Lavoie. The bank focused its automation efforts on operations, including back-office activities. 

"We have a lot of use cases that do mass updates to systems," he said. Examples include loan servicing, and an automation that enables customers to take advantage of promotional rates rapidly.

Citizens also uses automation to manage the mailbox for its syndicated loan portfolio. Using the technology, the bank went from having seven full-time colleagues managing a mailbox seven years ago, down to one person spending half their time managing the mailbox now, Lavoie said.

Opportunities and risks of AI agents

AI agents have the potential to help banks and other companies garner efficiency and cost savings from their investments in large language models, American Banker reported last month. Like Citizens, banks rolling out agentic capabilities typically have a human in the loop to review the model's work and catch any hallucinations, errors or bias in its output.

But this is bound to evolve over time. Data from Capgemini clients that have implemented agentic AI shows "we are getting to a point where they feel comfortable enough to let an AI agent make certain decisions without a human in the middle," Kartik Ramakrishnan, deputy CEO at financial services at Capgemini, said in a recent American Banker podcast.

Banks are early in their efforts to deploy agentic AI in production. Amin Kermali, vice president of enterprise digitization and automation at TD Bank, another customer of UiPath, recently told SiliconANGLE that automation and AI can drive "tremendous business value."

"We see that bringing these capabilities closer together puts us on the path to unlocking exponential potential, transforming the way we work and really future-proofing the bank," he said, noting that the bank is working to identify use cases that would benefit from the combined power of automation and AI. 

"I see us quickly getting to a spot where we're going to have a unified automation and AI operating model and ecosystem. I think that path towards agentic [is] where we're able to really start unlocking the full power," he said.

Tampa-based Suncoast Credit Union also said it's exploring what agentic AI capabilities could unlock, including instances where an agent can make decisions humans typically have to handle, said Michael Parks, senior vice president and CIO.

Agentic AI capabilities could include "more actionable items, and where it can start processing and thinking on its own and make decisions," allowing the agent to act autonomously, for instance in processing mortgages, he said.