Andrew Dresner has spent over 40 years in payments. He was a partner at Oliver Wyman, head of payment strategy at JP Morgan (where he co-founded Zelle), and a partner at McKinsey's Payments Practice. He now writes Payments at Full, one of the sharpest newsletters in the space. In this conversation with Eshita and Khushi, he covers why payments innovation is slow, what's really happening with stablecoins, and why he's deeply skeptical of agentic commerce.
The 20-Year Adoption Curve
Every major payments innovation—online banking, PayPal, Apple Pay, automatic bill pay—follows the same S-curve. Early adopters spike, then growth stalls until generational turnover carries it to mass adoption.
"A lot of these things only marginally improve on what they replace. It appeals mostly to a younger generation whose payments habits are not ingrained yet. And then it's just a matter of waiting for those people to age into their high earning years and for retirees to pass on the scene."
Apple Pay's early growth came from in-app payments, not tap-to-pay. Paying in an app was genuinely better than typing in a card number. But tap-to-pay at a terminal? That was barely different from tapping a card. The real penetration came years later, through generational inertia.
PayPal's Shrinking Niche
PayPal's original value was safety—protecting your card credentials in the Wild West of early e-commerce. That fear is gone. Amazon (35–45% of all e-commerce) doesn't accept PayPal. Shopify aggregates small merchants with its own payment stack. The unsafe internet that made PayPal essential no longer exists.
Why PayPal Is Losing Ground
- Amazon (~40% of e-commerce traffic) doesn't accept it
- Shopify absorbs small merchants with ShopPay
- The average PayPal user is aging—they adopted around 2000 and stayed
- Younger users default to Apple Pay or card-on-file
- PYUSD can't pay yield under the GENIUS Act
On PYUSD: it has merchant acceptance (PayPal's real asset), but without yield incentives and with limited cross-border share, it's unlikely to restore PayPal's position.
How Zelle Actually Happened
Dresner was one of six bank representatives who came together at The Clearing House to build what became Zelle. The impetus: banks were worried that Venmo and Cash App were capturing young users' digital engagement, and those users would eventually become the high-earning customers banks wanted.
The two founding design principles:
1. You already have it
Zelle lives inside your bank app. No separate download. Mobile banking adoption was in its hockey-stick phase—everyone had their bank app on their phone already.
2. Direct settlement to checking
Unlike Venmo and Cash App (which settle into a stored-value account), Zelle puts money directly in your checking account. Free. Instant. No 1.75% fee to "cash out."
"I actually went to my son's friends and said, 'At your age, aren't you Venmo fans?' They said, 'If our counterparty is a Venmo user, we use Venmo. Our preference is Zelle.' I said, 'Why?' They said, 'It's already in my bank app. And it settles directly into my checking account.' We got it right from the beginning. We just had to wait for the audience to catch on."
Stablecoins: Cross-Border Yes, Domestic No
Dresner is bullish on stablecoins for cross-border payments. Every major cross-border fintech is moving to stablecoins internally. The correspondent banking system is expensive, slow, and fragile. Stablecoins are better in every dimension.
Domestically? He doesn't see it.
"I would be shocked if 1% of payments are stablecoins in 10 years. Domestically, it's basically another version of a debit card. No chargebacks, no credit float, no Reg Z compliance, no rewards. And we just spent 10 years building real-time payment systems—FedNow and RTP. They're cheap, they work, they're instant."
What Conventional Rails Can't Do
There is one area where Dresner sees blockchain winning domestically: micropayments. Conventional systems can be cheap (ACH) but not real-time, or real-time (RTP/FedNow) but not cheap enough. Blockchain can be both—and infinitely divisible.
Agent-to-agent payments of a penny or a tenth of a penny are impossible on existing rails. This is where stablecoins or tokenized deposits have a genuine edge.
Tokenized Deposits
JP Morgan and Citi both have tokenized deposit products for corporate cross-border transactions. The problem: they're not interoperable. Just like pre-Zelle P2P systems (Chase QuickPay, BofA's version, Wells Fargo's version), they work great on-us but can't talk to each other.
The industry knows the answer—come together as a club, exactly like Zelle. Those discussions are happening now.
Agentic Commerce: Deep Skepticism
Dresner is perhaps most pointed on agentic commerce, which he compares to the IoT payments hype of the mid-2010s ("your refrigerator will buy your milk").
Why It's Harder Than It Looks
- Data access: Amazon, Walmart, and major retailers block scraping. Without their data, agents can't guarantee best price/selection.
- Prompt problem: For non-commodity goods (clothing, shoes, furniture), consumers don't know enough to write a good prompt. You can't ask for the right snowblower if you don't know the difference between a snowblower and a snow thrower.
- Taste and fit: No agent can tell you if a shoe will rub against your ankle. Size 8 isn't always size 8.
- Walled gardens: Amazon has its own agent. Walmart outsources to ChatGPT but keeps data locked to on-Walmart commerce. The big players are building proprietary, not open.
"ChatGPT just changed the way it does agentic commerce. It does the search for you, but drops you into the retailer's checkout. It's agentic search. That's great—I'd use it every day—but it's not the original vision of agentic commerce."
The one concept he finds promising: conversational commerce (Walmart's term), where the AI leads you through a Socratic dialogue to help you figure out what you actually want before searching. But he hasn't seen it in the wild beyond a press release.
How Payment Networks React
Visa and Mastercard invest in every trend because they're "rightly paranoid" about being displaced. Sometimes it pays off (tokenization for Apple Pay became central to all card processing). Sometimes it's wasted (IoT payments, provisioning tokens into Fitbits that nobody used).
"If agentic commerce ends up being agentic search—where ChatGPT drops you into checkout but the merchant processes the payment the same way they always have—then nothing's different. Checkout is identical. The investment may look, in hindsight, like wasted effort."
Advice for Builders
"In the battle between every startup and every incumbent, the question is: will the startup get distribution before the incumbent gets innovation?"
Dresner's core advice: think about distribution costs before you build. B2B2C models—where you provide technology to a bank (or Amazon, or Apple) and they aggregate the consumers—are often smarter than going direct.
The Chime Example
Chime is innovative and successful. But they spend more on marketing than PNC Bank does with $500 billion in assets. Why? PNC has branches. 95% of Americans open bank accounts in a branch. Direct-to-consumer requires massive CAC spend that B2B2C models avoid entirely.
Analysis: What This Means for Machine Payments
Dresner—a mainstream payments insider with no crypto agenda—independently validates a thesis that's already being built out in practice:
The Micropayment Gap Is Real
He explicitly says conventional rails cannot do real-time + cheap enough for sub-penny transactions. Agent-to-agent payments "will eventually emerge" on blockchain. But "eventually" is already here. BSV-based micropayment agents are processing 1–100 sat API calls today. This transcript was transcribed by one such agent for 83,751 sats (~$0.05).
Consumer Skepticism Doesn't Apply to Machines
Every objection Dresner raises about agentic commerce—taste, fit, prompt quality, data access—applies to consumer shopping. Machine-to-machine service calls (transcribe audio, generate an image, look up data) have none of these problems. The agent knows exactly what it wants. The 20-year adoption curve doesn't apply when there's no human behavior to change.
Distribution Through Invisibility
His B2B2C advice maps perfectly to how micropayment protocols can scale: don't ask consumers to adopt BSV. Embed the payment invisibly. A developer builds an agent, a user interacts through a familiar interface, and the blockchain is just plumbing. The user never needs to know.
The gap between "this should exist" and "this already exists" is the opportunity. The mainstream payments world sees the micropayment need. The infrastructure is already live.