How Meta's Muse Changes the Banking Relationship

Jim Marous (0:00)

Meta wants to be your customer's new AI agent. Within about 10 days of launch, its agent, Muse, was the most downloaded free app on the iPhone, ahead of ChatGPT and Instagram. It can connect to your customer's financial accounts through Plaid, and Meta's chief AI officer has challenged people to see whether it can find them $1,000 in savings. Think about what happens after his finds a forgotten charge or helps lower a bill. The next time that customer has a money question, they may ask Muse before they ask you. Your bank or credit union could still hold their deposits while Muse becomes the first place they turn to for financial advice. In an earlier Banking Insights video called The Agentic Bank Run, I asked what happens if an AI agent decides where your customers keep their money. Many of the people weighing in on Muse, including the chief economist at Apollo, the investment firm, are now raising the same question. I still think it's a real risk, but what I think deserves more attention is what could happen long before any automated mass transfer when your customer starts relying on an agent to decide what to do with their money. With a customer's permission, Muse can use Plaid to reconnect to financial accounts, including balances and transactions, and depending on the account, investments and certain loan details. Nothing Meta or Plaid has announced lets it move money between bank accounts, and Meta says Muse will check with users before sensitive actions like purchase even take place. So for now, the exposure is influenced over your customer's decisions well before any transaction. Mark Zuckerberg has said Meta expects to make money from taking a small fee from transactions.

You know, that raises a question your customers are going to keep wanting to have answered. Why did the agent recommend a particular product or action? A good comparison should explain the costs and trade-offs, including whether a commercial relationship with Meta affects who gets recommended. And Meta can put Muse in front of people through services they already have and use every day, including Facebook and WhatsApp. That's a powerful starting point, although chatting in a messaging app or on social media is a very different commitment from letting your agent manage your finances. Many would say that as long as an agent can't move deposits, we have time. I think that reading could really cost us. In my Banking Insight video on primacy, I warned that we could become little more than a customer's vault. Let's take Beth. Her paycheck still lands in your checking account every two weeks, and your dashboard shows a funded account with regular activity.

Meanwhile, Beth is asking an agent which card to use and where to borrow for her next big purchase, and you may never get a chance to compete for it. And there's one more consequence on the other side of our balance sheet that almost no one is covering. I'll come back to that in a minute. Under the hashtag, Muse Money Challenge, people are posting what the agent found for them. One founder said Muse found car insurance that cost $3,500 less a year. Another user got money back on a book subscription that had been forgotten about for a year. These are individual claims, and nobody has independently verified the savings reference. And many of the posts come from Meta's own employees. Still, this is how a habit starts.

Someone begins with a forgotten subscription, then comes back for help comparing a credit card. Each good experience could make them more comfortable bringing the agent into a bigger decision. Now think about what we can see.

Your transaction data can show a recurring payment or an increase in the bill. It won't tell you whether an insurance policy has the right coverage or whether the customer still needs a subscription. But it gives you a reason to start a conversation about financial wellness. So how often are we turning these signals into something the customer can actually use? That brings us back to the other side of the balance sheet, your loan book. Plaid's liability data covers credit cards, mortgages and student loans, with the details available depending on the account. Where a card balance and interest rate are available, an agent has a starting point for asking whether another option could be better or cost less. McKinsey has already flagged this, pointing to deposits and credit card lending in particular, as areas AI agents could significantly disrupt by cutting through customers' inertia.

Muse isn't refinancing anyone's balances today, but Beth could keep her checking account with you and still take her next borrowing decision somewhere else. As I've discussed, AI is already influencing which institutions get considered. A research firm called Ivo tested more than 2,000 small business banking conversations with AI agents. In its Chat GPT and Gemini tests, one fintech, Muircury, received more outright recommendations than the 30 banks it tested combined. That study didn't test Muse or measure actual switching, and banks did much better on credit decisions.