In one week, the biggest banks in the country slammed the door on AI agents.
Between October 13 and 16, JPMorgan Chase, Bank of America, Citi, Schwab, and Wells Fargo all moved to restrict third-party AI agents from customer accounts. Meta’s Muse grabbed most of the headlines, but it’s just one of many. Grok, OpenAI’s Dots, OpenClaw, and other personal agents are all part of the same wave, and nobody has confirmed which ones are actually causing the problems. Schwab’s CEO says the firm blocks big agents “in the tens and twenties” every week. And Axios reported last week that banks are spotting unvetted agents inside customer accounts, some taking actions that don’t appear to come from the user at all. How banks are identifying these agents, let alone blocking them, is still largely unknown.
Oh, and the company behind Muse? Last week, a New Mexico judge ordered Meta to pay $25 billion after a jury found it misled consumers about the privacy of their data.
Cue the panicked Slack threads at every fintech company in America. Do we need a stance? Should our CEO be on CNBC by Friday? Are we missing the moment?
We asked fintech CMO Kevin Green. His answer: you’re probably worried about the wrong thing.
Most Fintechs Are Sitting This One Out (On Purpose)
Here’s the unresolved question at the center of all this: if an agent moves money without permission, who’s held liable? Banks can face consumer redress demands even when they didn’t build the agent. JPMorgan’s CFO is already pushing that risk back onto consumers, warning that customers who hand an agent their login could be on the hook for what it does. Meanwhile, Bank of America’s CEO wants a human in the loop before money leaves an account, and Jamie Dimon is warning about “rogue agents” numbering in the thousands.
Most fintechs sit in the middle of that stack. They can’t absorb the risk of agentic money movement, and Green says the consultancies and associations advising them know it. The guidance right now: stay away from letting AI write to critical systems until verification catches up.
So if your instinct is to jump into the agents vs. banks news cycle, pump the brakes. Unless you’re critical financial infrastructure, your commentary gets buried under the people who are. Reporters covering this story want to hear commentary from someone like Jamie Dimon.
The Real Risk Is Being Invisible to the Agent
Here’s the part fintech marketers can’t sit out. An agent doesn’t need to move a dime to decide where the money goes.
Muse was built to compare insurance, cancel forgotten subscriptions, and hunt down savings, then act on what it finds. Banks blocking the login doesn’t stop the agent from reading the web, sizing up options, and telling someone what to do next. The banks themselves are building around it: Schwab and JPMorgan are pursuing read-only connections that let AI tools view balances and do financial research without moving funds.
Read-only still means a spot in the discussion.
One of Green’s favorite examples is home equity. Agents already know a lot about people’s money, but they’re blind to most people’s biggest asset: the house they live in. According to ICE, 47.5 million mortgage holders are sitting on $11.7 trillion in tappable equity, about $212,000 each. If the agent can’t see that option, it will never recommend it, even when it’s the smartest move on the table.
“It’s a blind spot for humans, so it is absolutely a blind spot for AI.” Kevin Green
Swap home equity for whatever your category is, and you likely have the same problem.
Meet the Buyer in the Chat
Green’s advice boils down to one move: stop making the agent come to your website. Bring your product to the agent, minus the risk.
- Make your category legible. If an LLM can’t explain what you do and who it’s for in one sentence, it won’t recommend you. Plain-language explainers, clear FAQs, and answer-shaped content do the heavy lifting here.
- Earn the citations. Agents lean on sources they trust, and that means third-party validation. In our research with Pavilion, teams investing in PR were 107% more likely to report leads from AI search. Earned media is how you get into the answer set.
- Get in the room with your product team. Agent-facing tools like MCP servers decide whether an agent can act on what it finds about you. Marketing should have a voice in that roadmap, because it’s a distribution decision as much as a technical one.
Trust Is the Opening. Bring Receipts.
Consumers are being asked to hand their finances to agents, including one built by a company a jury recently found misled them about their data. That’s a trust gap, and fintechs can step into it, carefully.
The skeptics are getting louder, too. Gary Marcus called out the bank incidents last week as part of a growing pattern of agent-related security problems. Green puts it more simply: AI being capable of something doesn’t mean the infrastructure behind it is ready. And with consumers and banks pointing at each other over liability, people need someone to explain the rules before they hand over the keys.
The smart play is consumer education about what agents can and can’t see. Skip the public verdict on whether AI agents are good or bad. That keeps you useful to reporters without locking you into a position you’ll regret in six months.
And if you want reporters to actually call? Hot takes don’t cut it; data does. Original research on what people trust an agent to do with their money, where they draw the line, and which financial options they’d never think to ask about gives journalists something they can’t get from the agent makers’ press teams. Green calls it a huge research opportunity, and we agree.
The Bottom Line
Let the banks fight over who gets to move the money. Fintech marketers have a more urgent job: making sure that when an agent goes looking for the best option, yours is on the list.
The agents are already reading. Make sure they’re reading about you.
Building your fintech’s AI visibility strategy? Let’s talk.
