Editor’s Note — This article is sponsored by Method. As with all sponsored content in Fintech Takes, this article was written, edited, and published by me, Alex Johnson. I hope you enjoy it!

We’ve been talking about digital wallets for a very long time.

PayPal has been around since 1999, originally born from security software for handheld devices, back when Palm Pilots were a thing. Apple Pay and Google Wallet have been kicking for over a decade, and more recently, crypto has offered up yet another species of digital wallet. The concept is old enough to have survived generations of technology that were supposed to reinvent it.

Physical wallets and digital wallets aren’t fundamentally different experiences.¹ But that’s changing. And what’s changing isn’t the wallet; it’s who’s reaching into it.

Now that we have AI agents like Muse and Instinct that can search, compare prices, and transact on our behalf, the digital wallet may have finally met its match.²

For decades, we digitized the wallet without fundamentally changing the concept. Now the digital wallet has a new user, one that can tap into all the information sitting behind those payment credentials.

An AI agent needs controls. It needs visibility and permissioning. And it needs information about your cards that, until now, has organically lived within you. Which suggests something that’s slightly strange; to innovate, digital wallets may have needed AI agents all along.

AI agents have arrived to unlock the digital wallet.

But what’s missing is a wallet actually designed for them, which is what I hope to dig into with this piece.

First off, what does a digital wallet designed for agents need to know?

So glad you asked. Let’s start with first principles.

A wallet is multi-product by default. A wallet isn’t a wallet if there’s only one card in it. It only becomes one if it holds multiple payment credentials that you can switch between as you please. I’m not going to use my Capital One card at the gas station or my Shell card at the grocery store, but you do you. And if you're anything like the average American, you likely have 3.7 active cards, according to Experian, and each one has its own quirks. Maybe your Chase Sapphire pays more at restaurants, or your Amazon card is edging toward its limit. You know all of this without thinking; an AI agent doesn’t.

Here are the five things an AI agent needs to know about every card in your wallet:

• Identity. What exact card is this? Not merely Visa or Mastercard, but the specific product, issuer, network, fees and other attributes that distinguish it from every other card in your wallet.
• Preference. How does this person actually use it? Which ones see most of their spending? Your habits are information too, even if an agent ends up overriding them.
• Economics. For a particular purchase, what are the rewards categories, earn rates, APR, or fees? Is there a welcome bonus or spending threshold that makes this purchase more valuable on one card than another? Which brings us to the next question.
• Capacity. Can it support the purchase right now? If the card with the best rewards is $20 away from its limit, it’s not the best card for a $500 transaction.
• Authority. Is the agent even authorized to transact with this card? Once the cardholder and the entity that makes the payment are no longer one and the same, permissioning becomes part of the wallet itself.

These five pieces of information transform the concept of the wallet into fabric that an agent can reason across, and once it can reason across your wallet, an old assumption in credit cards begins to fray: Top of wallet.

Issuers have fought for the top of wallet position because humans have habits. A card can offer better economics for a purchase and still lose if the customer never thinks to reach for it.

Suppose my usual card earns 2x on travel, but another card in my wallet earns 4x. The second card looks better until the agent sees that it sits $20 from its credit limit. Suddenly my default card wins again, as it should. The agent needs to be able to weigh preference, economics and capacity simultaneously, just like we do unconsciously.

But this intelligence layer won’t unlock if an agent only has access to one card; it needs access to the entire digital wallet, plus enough information about each credential to make a decision.

And this is where our first principles exercise starts to look a whole lot less theoretical.

Method, for example, can identify eligible cards across more than 4,000 financial institutions using a name, phone number and one-time passcode, instead of asking someone to enter every card by hand. From there, APIs can return a specific card product and its rewards structure, while a separate layer adds behavioral context on how the person uses their cards. Payment-readiness signals can indicate whether a card appears able to support the transaction, and once a card is chosen, tokenized credentials can move it into a merchant's existing payment flow without the raw number ever showing up.³

The point isn’t that any one provider has solved agentic payments. It’s that the pieces are coalescing into something concrete: Access to the full wallet, enough context to compare the cards inside it, and a secure way to act on the choice.

Bilt is a good example of what happens when just this first piece falls into place. After Bilt partnered with Method to design a card-enrollment flow as part of Bilt's onboarding, Bilt reported a 148% increase in card enrollment and an 83% increase in neighborhood merchant spend per member.

Now of course, that doesn't show us what agentic shopping will do, but it does show us that connecting more of a customer’s wallet gives a platform more opportunities to recognize, reward, and engage their spending, even before an agent comes near a transaction.

Imagine what happens, though, when an agent does?

A wallet has always held your card, but it’s never held the judgment that accompanied them: Which one to reach for, whether it was the right card for the right purchase. That stayed in your head because no one ever asked, but someone, or rather some thing, is asking now.

The first generation of digital wallets digitized your cards.

The next one has to digitize the context required to choose among them.