What a difference a year makes in the evolution of artificial intelligence. 2025 started with many banks and credit unions building out their first use cases for generative AI. It ended with a few select institutions moving on to testing, piloting and implementing agentic AI solutions.
Financial institutions have started using gen AI — which can create new content like text, images and audio based on user prompts — to analyze and summarize data, write reports and provide other efficiencies that allow employees to focus on more important tasks. Of the bank CEOs, technology executives and board members who responded to Bank Director’s 2025 Technology Survey, 48% said their institution was using gen AI, with marketing, IT and administration representing the highest percentages of use cases.
Agentic AI takes things to another level. It is designed to allow AI agents to execute complex, multistep tasks autonomously or with little human involvement. The technology is so new that few fintechs were even talking to financial institutions about it one year ago. But that changed in the second half of 2025, as vendors started partnering with banks and credit unions to develop agentic AI solutions.
One of the industry’s earlier use cases has shown improved efficiencies in loan processing. Lake Michigan Credit Union (LMCU), a $15 billion institution based in Caledonia, Michigan, started using agentic AI in 2025 to grow its robust mortgage business, which originates $3.5 billion in home loans annually. Noel Watts, LMCU’s senior vice president of lending operations, found that about 60% of the tasks the credit union’s loan origination system presented to his disclosure team didn’t need to be performed on most loans. Greenlight Consulting offered LMCU a solution that allows an AI agent to review all of those tasks, complete the ones that are required and clear those that aren’t before the disclosure team ever sees the loan file.
“Rather than adding headcount, the same number of disclosure team members that we had on board can handle 60% more files than what we were able to in the past,” he says. Watts, who anticipates an uptick in mortgage refinance volume, believes the AI agent gives LMCU a major competitive advantage. “At some point, when rates fall, we want to be prepared.” He plans to expand the use of the agentic AI tool to grow efficiencies in LMCU’s home equity and consumer lending departments as well.
Refining Agentic AI Solutions
Other financial institutions are trying to get where LMCU is by testing and piloting agentic AI solutions from their vendors. In late 2025, nCino launched five agentic AI tools — which it calls “digital partners” — that banks and credit unions could pilot in different areas of their institution.
WaFd Bank, the subsidiary of $27 billion WaFd in Seattle, has been working with nCino to test and pilot the company’s Analyst Digital Partner, which assists with credit risk assessment and financial analysis.
Jillian Boyle, the head of WaFd Bank’s commercial product and treasury solutions group, likes being on the front end of new solutions. “I think it gives us the opportunity to control the risk,” she says. “And I think it gives us the opportunity to help our people through this change.” Boyle plans for the tool to automate many of WaFd’s loan underwriting and quality control processes within the next two months. The agent will start by reviewing and validating loan documents to ensure the terms match what is in the loan origination system and help resolve any outstanding issues. Boyle hopes the agent will eventually perform continuous credit monitoring as well.
WaFd and the other institutions piloting nCino’s agentic AI offerings are helping the company shape the final version of the tools. Michael Chung, general manager of commercial lending for nCino, says that customer feedback has been vital. “This ramp-up period was really about us collaborating with a select group of our customers to refine our solution,” he says. “Now, we feel like it’s at a point where it is ready for the masses.”
The potential of agentic AI is something even small financial institutions are starting to explore. United Financial Credit Union (UFCU), a $339 million institution based in Saginaw, Michigan, set a March 1 go-live date for its own agentic AI solution, provided by loan origination system provider Algebrik AI. The tool will be applied to indirect auto loan applications, says UFCU CEO Barbara Appold. “How quickly you jump on those could make or break the deal. It’s going to come directly to our loan department, and they’re going to be able to react to that instantly,” she explains. “Plus, just like any other rules-based engine, if we have the right rules set up, it could just approve it on the spot as well.”
Appold believes the tool can help UFCU level the playing field when competing for those loans against larger financial institutions and neobanks. She eventually wants to use it for direct lending and membership applications as well. “We don’t want somebody to think [that] just because we’re small, we can’t do that,” Appold says. “I have a young management team. They’re very aggressive, very willing to take on the next task.”
Enough Human Control?
While the pilots appear promising, not all financial institutions are ready to embrace agentic AI solutions. Erin Simpson, executive vice president and chief operations officer at Encore Bank, the $3.6 billion subsidiary of Encore Bancshares in Little Rock, Arkansas, admits to being a skeptic. She says the reward isn’t worth the risk at this early stage.
Encore Bank focuses primarily on commercial banking and has business verticals in title management, third-party escrow, property management and association banking — all of which are connected through a network of application programming interfaces. “Our information security officer and our chief information data officer were just not comfortable with having an agent sit on top as a user in those systems and performing tasks,” Simpson says. “Risk to a retail client is not as high from a dollar standpoint as it is on these big commercial [transactions].”
Chung says nCino’s solutions are designed so a human can review the work of the AI agent at critical points throughout the process. WaFd Bank executives asked nCino to ensure there are system stops that allow for human oversight in the same places they have quality control in place for manual reviews. “Because we know our examiners are comfortable with that approach,” Boyle says.
And those bank and credit union IT examiners are important to consider when setting up any type of AI use. “They’re saying, ‘What are you thinking right now? Do you have an acceptable use policy? How much governance do you have around this? How involved is your management and board in making these decisions? Are you using consultants to help you?’ But they’re not saying, ‘No, you can’t do this. Yes, you can do that,’” Simpson says.
They aren’t saying that last part because there are currently no rules for AI use in banking at the federal level. But all of the executives interviewed for this story feel AI regulations will eventually need to be put in place. ”I think as people and as business owners and as corporations, our role in this digital evolution is to also make sure our banks are protecting us,” Boyle says. “And I think that federal regulation is going to help us know that there are some consistent guardrails.”
Previous