Agentic Commerce Reality Check at Money20/20

Stablecoin merchant acceptance is growing, but the reality on the ground is more nuanced than the headlines suggest.   Flagship Advisory Partners recently canvassed the top merchants across the U.S., Europe, and Latin America to assess where stablecoin acceptance actually stands today. Direct acceptance remains limited: just 4% of large U.S. merchants, 8% in Europe, and 12% in Latin America accept stablecoins as an e-commerce payment method. That said, around half of large merchants across these regions have agreed to participate in crypto gift card programs, making indirect acceptance far more common than direct on-chain or PSP-intermediated flows.   The cost case is real. Stablecoin PSPs offer acceptance costs of 0.8–1.4%, comparing favorably to credit cards at 1.8–4% in US. But cost alone is not sufficient to drive mass adoption. Only around 5% of U.S. adults hold stablecoins today, and just 2% used any form of crypto to make a payment in 2024. The addressable market is largely limited to those already receiving stablecoins through payroll, remittances, or other inflows.   The near-term opportunity is concentrated in specific verticals: high-risk, luxury, travel, and cross-border commerce. Consumer adoption will follow business adoption, not the other way around. Merchants can afford to evaluate their own case carefully. Banks and money-movement PSPs should be acting now.   Full article here: https://lnkd.in/dJHHJMKz   Authors: Will Hay, Alessandro Mighetto, and Jordan Matar   #Stablecoins #Payments #MerchantAcceptance #Fintech #NewWaysToPay  

One of the more common hypotheses we get approached with, when it comes to stablecoins, is that people will start using them to pay for stuff online. We did the math, and found that a) this probably won't meaningfully be the case anytime soon, and yet b) there are surprising hot spots where signs of traction can be seen. My general perspective continues to be that in payments acceptance, we'll continue to diverge into a more 'horses for courses' approach - particularly playing out behind the scenes, behind the consumer initiation layer - with stablecoins playing a role alongside real-time payments, cards and whatever comes next. #stablecoins #payments

Looking forward to this one. Flagship Advisory Partners works with numerous PSPs and enterprises on loyalty and value-added service strategy development, and for good reason. The average US shopper is enrolled in about 8 programs, actively uses maybe 5, and leaves roughly $10B in rewards on the table every year. Spend influenced by loyalty is even greater and is a revenue accelerator across the ecosystem. As such, issuers, wallets, and PSPs are all continuing to invest in loyalty, with Adyen's first-ever acquisition being a loyalty platform. Plenty to unpack in this conversation, while learning more from Accrue and SNIPES on the impacts they see/feel from embedded payments/loyalty strategy. Come find us in Nashville if you want to talk more about this or other embedded payments/finance topics. #MAG2026 #payments #fintech #loyalty #embeddedpayments

The software market has been rocked this year on the fears of disruption from AI, but there's nuance to be considered that the market tends to overlook. Some thoughts: - SMB software is less threatened, on the whole, than enterprise - SMB capacity for investment, development and implementation is a major barrier to entry - and yet this is not reflected in the market. - There are sectors where software *is* compliance; in healthcare, legal and FS, there are layers of software that provide assurance and have the history of certifications, audit and relationships to be able to provide assurance. i.e. you can't vibe-code credibility. - We've spoken at length about how vertical SaaS is becoming the primary distribution channel for payments - and this in itself is a defensive moat. Not only does churn collapse to low single digits when you embed payments in software anyway, but AI can't stand up banking relationships, ensure PCI compliance and handle money movement. As we're seeing with the early stages of agentic commerce, payments are too critical to surrender control to AI just yet. More thoughts below in our latest piece written by Joel Van Arsdale and Charlotte Al Usta.

For all of the conversation about PayPal's change in CEO, what I've taken away from it is how real the erosion of a fintech's core value proposition can be - even if it is a giant of the industry. My colleague Ben Brown and I have written a deep dive on the subject (stay tuned for the performance analysis and creative bets on the future), but ahead of that, it's worth reflecting on how strong payments businesses are constantly having to reposition themselves - often with big bets - to address changing market fundamentals. 🔃 Klarna was astute in recognizing the necessary regulatory and competitive headwinds in the BNPL space and has made a strategic effort to reposition itself as an AI-enabled shopping assistant/platform, and in many ways a digital bank, including savings and card issuing. There's early evidence of this success; their BNPL merchant fees have gone from 75% of revenue in 2020 to 57%, whereas ad revenue went up 15X - from $13M to $180M. 🔃 Visa is fascinating to me - a behemoth in a cards industry with a deep defensive moat and value proposition in the US, one could argue that it had little reason to worry about growing A2A payments and Pay By Bank. However, Visa is now live as a key enabler of A2A payments in the UK for Pay By Bank, and is unbundling its core value proposition (brand trust, dispute resolution, rules, governance) from cards and actively selling it over A2A rails. That's a recognition, and bet, on its USP. 🔃 Nuvei's story is relevant, from a different angle - sometimes bravery isn't in the product pivot, but structural. Going private in 2024 (less than 4 years after its landmark Canadian IPO) was a bold bet on its core value proposition, removing the limitations that being public might have placed on its ability to go after new alternative payment methods and rails. Public companies can be squeezed into chasing volumes or margins - would PayPal benefit from a similar approach? That's a sneak peak from our POV.... In today's fintech world, incumbency only buys time. Commitment to product-market fit and pivoting to meet customer need is what brings survival and success.

So many gems in our latest Payments on Fire podcast with Ran (Goldi) Goldshtein from Fireblocks, covering Stablecoins and Tokenized Deposits. The value in stablecoins resides in the time value of money: "And if you can find economic value in that speed, then that is the real value that stablecoins will give you as a rail. So for example, in the importer exporter use case, the fact that someone can pay their exporter on a container that just hit the port and they don’t need to wait two more days, that’s two days of storing that container in port that costs $25-50,000, whatever. So they might have saved $50,000, but not on the transfer, but the economic value that the velocity of money created for them." Banks don't want to use stablecoins, because it requires 1-1 reserve backing. This is why Tokenized Deposits are more interesting to them: "Banks do not want to hold your money one-to-one reserved. They are not on the stablecoin camp. That is not their business model." Of the $1.5 Trillion in Stablecoin payment processing, most of it is B2B payments and disbursements: "...about 40% of that is B2B payments, 15% of that is payment acceptance and pay ins, about 25% is payouts, so disbursements, and the rest is remittance and a bit of other stuff." How PayPal's issuance of PYUSD helps unlock efficiencies for PayPal and their cross-border partners that enable Xoom "Now with PYUSD, which is a stablecoin that PayPal created, by the way, as an issuer, they fund [XB partner] every 24 hours, which means that they can fund them on a weekend. They can fund them and keep the rest of the three days they used to send their way still in their account, making interest or doing whatever they want to do with their money." I highly recommend listening to this episode. Link to podcast and transcript in the comments.

OpenAI is improving at building (and leveraging) a persistent and comprehensive picture of who you are, by taking data from conversations you have with it and updating a profile it builds in the background on your work, hobbies, and other unique attributes like travel preferences. The implications for agentic commerce are intuitive: The better ChatGPT knows you, the more relevant the results will be for product and services discovery. I predict they will build a "payment preference" model that understands what payment methods you like to use for different commerce experiences. Persistent and evolving preference data is a key component of the “Missing Context Layer” I wrote about in my recent series on agentic commerce. OpenAI is developing this in the cloud layer as opposed to the device layer, but my expectation is that the device layer will also have this capability in the future. We are already seeing major improvements in the technology to enable this on-device. Researchers at Stanford and Lambda recently published a framework called OpenJarvis that runs the entire personal-AI stack: inference, memory, and learning on the device itself, with competitive accuracy and much lower cost. It connects to your email, calendar, and messages through permissioned integrations, and the profile it builds stays local to you. The takeaway is that the persistent picture OpenAI is assembling in the cloud can increasingly be assembled and held on the device instead. That sets up a more important question, and it is not really a technical one: who do we want holding this model of us? Three answers are taking shape. 1. OpenAI is building it in the cloud, tied to the conversational relationship you already have with ChatGPT. 2. Apple and Google are positioned to build it at the operating-system layer, where they have native access to device context. 3. And an open, on-device approach, the direction OpenJarvis points to, would keep the model in the hands of the consumer. For payments and commerce, this is not an abstract debate, because whoever holds the model shapes the decisions that happen before the transaction: what products you consider, which option is recommended, and ultimately which payment method is selected. In my series I argued that the operating-system layer held the structural advantage, since the relevant data lives on the device. A cloud assistant with permissioned access to your accounts complicates that view, it can reach much of the same financial context without ever owning the device. In the end, the trade-off comes down to trust. A comprehensive, cross-domain profile held in one company's cloud is a different proposition than the same profile held on your own device. As this capability matures, I expect the question of where the model lives, and who controls it, to matter just as much as how well it works. Sources: Decoder, MarkTechPost, OpenAI, (links in comments)

My colleagues Ryan and Joseph shared our quarterly update on US card issuing performance. One of the more interesting points is made at the end of the article, where the guys mention how BNPL credit is moving against the traditional credit trend. Klarna specifically saw allowances grow 48% in 2025 against 29% receivables growth, lifting its implied loss rate 60 bps to 4.5%. Flagship Advisory Partners https://lnkd.in/eBC2-nDU

𝘞𝘩𝘰 𝘪𝘴 𝘵𝘩𝘪𝘴 𝘢𝘤𝘵𝘰𝘳, 𝘢𝘯𝘥 𝘤𝘢𝘯 𝘐 𝘵𝘳𝘶𝘴𝘵 𝘵𝘩𝘦𝘮? Despite programmable payment rails, AI agents, and stablecoins, IME the main merchant onboarding bottleneck hasn't budged: TRUST INFRASTRUCTURE Fintech is focused on the rails that move money. But equally important is the ability to make confident decisions at scale. Agentic commerce will only work if the ecosystem can answer my opening question instantly and reliably. Tahoe has become a thinking ritual for me. Same chair. Same spot. Any time of year. Five years ago I was sitting in this Adirondack chair when I realized I was genuinely excited to build in payments again. Over a drink with my friend Billy Conway, who pointed out the excitement in my voice and encouraged me to go build again. That conversation tipped me from curiosity into commitment. All in on Kompliant. I sit here again, unplugged, musing on what the next five years might bring to fintech. Some things that I’m watching closely:  • 𝗔𝗴𝗲𝗻𝘁-𝗱𝗿𝗶𝘃𝗲𝗻 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗲: Agents will initially transact on card rails, then increasingly on stablecoins.  • 𝗖𝗮𝗿𝗱𝘀 𝗮𝘂𝘁𝗵𝗼𝗿𝗶𝘇𝗲 𝗺𝗼𝗻𝗲𝘆. 𝗦𝘁𝗮𝗯𝗹𝗲𝗰𝗼𝗶𝗻𝘀 𝗺𝗼𝘃𝗲 𝗺𝗼𝗻𝗲𝘆: Stablecoins do not need to replace cards. They may make them work better.  • 𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗮𝗯𝗹𝗲 𝗽𝗮𝘆𝗺𝗲𝗻𝘁𝘀 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝘀: Initiatives like x402 from Cloudflare and Coinbase suggest a future where agents can pay APIs and services directly w/o accounts or traditional onboarding.  • 𝗔𝗴𝗲𝗻𝘁-𝗳𝗶𝗿𝘀𝘁 𝗳𝗶𝗻𝘁𝗲𝗰𝗵 𝘄𝗶𝗹𝗹 𝗯𝗲𝗰𝗼𝗺𝗲 𝗮 𝗰𝗮𝘁𝗲𝗴𝗼𝗿𝘆: But, again, the critical unlock across all of this is trust infrastructure. Processors and banks are not just enabling payments. They are underwriting risk. And when agents begin transacting autonomously, identity, verification, and risk intelligence will become even more important than the rails themselves. ASK I would love to connect with operators and builders who are thinking deeply and working on these problems from inside the system:  • Agent commerce infrastructure  • Federated identity & trust layers  • Merchant risk intelligence (at scale)  • Programmable payments standards  • Stablecoins & card networks Comment or DM if that's you 🙏 [Tahoe photo taken a week ago with my youngest]

Card payment performance is often a (lack of) enhanced data problem, and the payments ecosystem is not leveraging network tokens and 3-D Secure effectively to improve it. My colleagues Christopher Uriarte and Justin Pituch have written a great report on how tokenization and 3-D Secure should be complementary components of merchant risk management. Not all payment tokens carry equal trust. Any merchant that receives a PAN can request a network token, even if that PAN was obtained fraudulently. Without authentication at provisioning or at transaction time, the credential itself gets more secure, but the identity behind it remains unverified. Issuer readiness for network tokens is still uneven. Many issuers don't yet consume the full set of token metadata: assurance levels, domain controls, that's supposed to improve decisioning. Some don't support network tokens at all. Merchants are still running PAN-based fallback strategies to specific issuers because tokenized auth rates underperform. Removing 3DS in a tokenized environment doesn't reduce friction, it increases it. Visa's data shows that including priority data elements in authentication requests can lift fraud detection rates by up to 65%. Strip that layer out and issuers compensate with tighter thresholds, more declines, and a worse customer experience. The industry narrative that tokens will eventually replace authentication has it backwards. As credentials get more secure, the relative value of strong authentication signals goes up, not down. Merchants optimizing for approval rates need both layers working together.