TL;DR
The next generation of consumer AI agents will do more than recommend financial products. They will act on their humans’ behalf. Open accounts. Move funds. Refinance mortgages. Switch institutions when rates move. This changes a fundamental assumption that community financial services has been built on for a century, which is that the customer is always a person on the other end of the interaction. This post is about the end of that assumption, why it matters, and what it means for how you run your institution over the next three years.
Let me tell you about a conversation that has stuck with me for months.
A credit union CEO I respect, one of those people who has forgotten more about community banking than most of us will ever learn, looked at me across his desk and asked, “What happens when the member isn’t a member anymore. What happens when the member is a bot.” He was half joking and half genuinely confused. I told him he had just asked the right question about two years before his competitors would.
Here is the honest answer. We do not fully know yet. But we know enough to start preparing.
The slow collapse of a comfortable assumption
Every interaction your institution is designed for assumes a human is at the other end. The UX of your mobile app. The pacing of your phone tree. The friction in your account opening flow. The tone of your collections calls. The urgency of your fraud alerts.
All of it assumes a human is clicking, reading, listening, and deciding.
That assumption is about to quietly break. Not in a headline-grabbing way, but in a cumulative one. Your members will start delegating more decisions to AI agents. At first it will be small things. “Find me a better savings rate.” Then it will be bigger things. “Move my checking to whichever institution is offering the best bonus.” Then it will be the things that used to require a meeting. “Refinance my mortgage if the rate drops below six percent and notify me when it is done.”
Each of those small delegations is a human stepping back from the interaction and letting a representative act on their behalf. By the time you notice the shift is happening, it will already have shifted.
What this actually means for your institution
This is not an abstract future. Think through a few of the immediate implications.
Your onboarding flow was designed to reassure a nervous human. It has warm welcome messaging, explanatory tooltips, and a gentle progress bar. None of that matters to an agent. The agent wants an API endpoint, a machine-readable disclosure, and a clean yes or no.
Your collections outreach was designed to land in the hands of a member who might feel embarrassed, stressed, or defensive. That same outreach, if it lands in the hands of an agent, either gets ignored, rerouted, or translated into a one-line notification that strips all the context your team carefully worked into the message.
Your fraud alerts were designed to catch a human’s attention fast. When an agent is the one managing the account, your fraud alerts need to be parseable, categorizable, and machine-verifiable, not just emotionally urgent.
Every design decision you have made over the last twenty years of digital experience is about to be re-examined by a new kind of user. And the new user does not read your marketing copy.
The identity question nobody has answered yet
Here is the question that regulators, institutions, and consumers have not yet collectively answered. When an agent acts on a human’s behalf, who is responsible for the decision.
If an agent opens a checking account in your member’s name, using credentials the human authorized, and the agent makes a choice the human later disputes, whose problem is that. If an agent moves money out of your institution based on a rate shop, and the rate information was wrong because your website had stale data, who is liable. If a member authorizes an agent to manage their finances, and the agent executes a dozen transactions a day, each fully compliant on its own but collectively evidence of money movement patterns that look suspicious, what is your obligation.
These are not edge cases. These are going to be, within twenty-four months, the dominant policy questions in financial services.
The institutions that start grappling with them now will have an easier time shaping the answer. The ones that wait will be adapting to rules their regulators wrote without their input.
Why this is not all bad news
Reading this back, I realize it sounds a little grim. Let me balance it out.
Agentic members are going to be the most profitable members you have ever had. They will self-serve at a rate humans cannot match. They will react to product changes quickly and transparently. They will surface dissatisfaction early rather than silently churning. They will make your service economics better, not worse, if you build for them intentionally.
The institutions that embrace this shift will look back on this period the way good institutions looked back on the rise of mobile banking. There was a lot of handwringing at the time. There is a lot of handwringing now. In both cases, the institutions that leaned in ended up stronger, not weaker.
Agentic members are also, and this matters, going to be more loyal than we assume. Agents do not switch institutions for fun. They switch when the data says the current institution is underperforming on a metric the human cares about. If you are performing on the metrics your members care about, your agentic members will stay. The bar is transparent. Clear it.
What comes next
In the next twelve to twenty-four months, expect three visible changes.
The first is that the definition of a “member” will start to bifurcate. You will have humans, and you will have agents acting on behalf of humans. The smart institutions will start tracking both as separate cohorts, because their behavior, economics, and product needs are different.
The second is that mobile banking apps will get a programmatic counterpart. Not a redesigned app. A parallel surface, exposed through APIs or agentic interfaces, that is built for non-human use. The ones that do this well will look like they built two products. The ones that do not will look like their competitors built one.
The third is the beginning of agent-first product design. Products designed not for a human to read about and then decide on, but for an agent to evaluate against a set of criteria and select on a human’s behalf. The best-performing financial products in 2028 will be the ones that agents systematically choose, not the ones with the best magazine ads.
The bottom line
The member of the future is not always going to be a person. Sometimes it will be an agent acting on a person’s behalf. The institutions that start designing for that reality now will serve both humans and agents well. The ones that ignore it will serve humans well and confuse agents. Confused agents route around you.
This is not a threat. It is a transition. The financial institutions that have navigated transitions well before, from branch to digital, from desktop to mobile, from product-centric to member-centric, have the exact muscle they need to navigate this one. Same job, new member type.
Kevin Farley writes about the future of member experience, agentic AI, and growth strategy for financial services. Read more on the blog.