Those that have operated in the credit union space since the early 2000s have become quite familiar with margin compression – rising deposit costs, lower loan spreads, increasing operational expenses, regulatory pressure on non-interest income, increased competition, increased compliance costs, and of course, challenges in finding efficiency while continuing to expand and invest in new channel delivery tools.
This paradigm is due to reset and it will largely be fueled by agentic AI using the operator model (while still leveraging existing API).
Agentic AI is making inroads in areas society once thought so entrenched that change might never reach the depths of their work, including finance and medicine. Researchers in medicine, for example, have tested the large language models to perform patient interviews and patient diagnosis.
Research published in the Journal of Nature on AMIE (Articulate Medical Intelligence Explorer) demonstrated that this AI system was able to interpret patient symptoms in controlled environments to arrive at more complete and accurate diagnostic results better than the physician cohort. Stated another way, AI has limited ability to do comprehensive patient intake to make highly detailed diagnostic suggestions today at levels meeting or exceeding those of practicing clinicians. Not to replace the clinician, but for the clinician to partner with AMIE to improve patient care, diagnostics, and remove some of the arduous paperwork to allow more time to focus on the humanity of the patient encounter. (Source: Journal of Nature, April 9, 2025. Towards accurate differential diagnosis with large language models.)
Agentic AI in medicine in many ways closely parallels the work being done in finance.
Both disciplines are largely based on fulfilling an individual’s needs (health or finance), with relationships deeply based on trust, accuracy and regulatory oversight. In banking, similar AI breakthroughs are driving improvements in efficiency. Examples include:
- Financials and their vendors are leveraging AI to perform large pattern recognition to identify fraud and reduce losses both internally and with card use.
- New tools allow 24/7 account monitoring that can “freeze” account activity if the AI detects a probable fraud event. Previously, these cases would have been in suspense until the next business day, serving more value in autopsying the loss than in preventing the loss. With today’s AI intervention, those losses can be curtailed in real time despite staff not working 24/7.
- AI large language model BOTS can often handle up to 60%-80% of incoming calls. As agentic AI continues to evolve and more human-like experiences are delivered, we can assume member acceptance will also continue to increase.
- Loan underwriting is being done utilizing AI and being reviewed by internal staff providing final approval on consumer loans and preapproval on mortgage loans.
- On the investment side, AI is monitoring specific investment portfolios to alert portfolio managers of changes when triggers are reached.
But of all the AI agentic operator models, I believe the single most important tool will be the eventual opportunity to automate back-office work to gain efficiency.
This could represent the single largest area of opportunity for credit unions in our lifetime. AI automation can reduce costs, automate processes that members never see and provide staff more time to spend with members on meaningful and deeper financial needs.
In other words, AI can drive incredible efficiency in the back office so teams can spend more time with members on the front side of the credit union.
Leveraging AI and agentic operator models is not only the present and the future, but it is likely the magic key that credit unions will need to leverage for survival over the next 20 years. Here’s just a small list of ways credit unions are and will be leveraging AI:
- Automated GL balancing;
- Preparing documentation;
- Preparing, sending, double checking and ultimately filing documents into a document storage system;
- Creating a lending summary for commercial accounting reports;
- Removing repetitive and manual task work;
- Drafting emails;
- Summarizing documents;
- Monitoring exception reports and alerting managers to exceptions daily; and
- Marketing resource management and finding the next best action for members.
What is limiting the use of AI in back-office processing?
Many early adopters that jumped into AI feet first could not generate results they had hoped for, and the process workflows continued to break and required constant tweaking that eroded reliability and cost efficiency of the systems. Another reason for slow AI adoption is many AI automation firms in the market today have danced closely around the financial services “fire” but have not entered in directly. In other words, AI providers need to come forward offering guarantees of performance and billing tied to automation achieved (and to date at least one firm has). While these are not complete showstoppers, they are things that need to be addressed in contractual format.
The next five years will determine the future of the next 20 for efficiency leveraging automation. The large financial institutions are investing considerably in this technology, and it is up to credit unions to leverage our collective power to find vendors that can support the unique efficiencies we all seek using fairly standard processes.
Credit unions that partner with very respected and proven AI providers to automate back-office tasks gain the ability to save significant dollars will certainly be winners of the AI race and will be positioned with long-term growth and success.
As an industry, this is our time. Our time to leverage technology, our time to automate, and our time to drive down operating costs while not eroding the value we place on member human interaction.
This represents the profound value of agentic AI. I firmly believe agentic AI will become one of the most transformative forces in our industry (if not the single most) in our lifetimes. We need to leverage it and leverage it now. And like medicine, it should not be used solely to replace humans but to compliment and partner with the human employee to deliver better solutions and member experiences for our members.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.