Payment Networks Face an AI Reckoning as Agentic Commerce Threatens Card Rails

On a Monday in late February 2026, shares of Visa, Mastercard, and American Express dropped between 4% and 7% after an analyst firm published a scenario document describing how artificial intelligence agents could systematically route around the card network economics that have underwritten decades of payments industry profits. The report was framed as a thought experiment. The market treated it like a warning.

The Citrini Research Scenario: What Agentic Commerce Does to Interchange

Citrini Research co-authored a document with Alap Shah, Co-Founder and CEO at Littlebird, titled "The 2028 Global Intelligence Crisis," framing a near-term scenario in which agentic AI effectively neutralizes the 2-3% card interchange fee structure [3]. The mechanism is straightforward: AI agents executing autonomous purchasing decisions will optimize for the cheapest available payment method, and stablecoin settlement on Solana or Ethereum L2 networks delivers near-instant settlement at a cost measured in fractions of a penny per transaction, compared to card interchange averaging 2-3% [3].

"In machine-to-machine commerce, the 2-3% card interchange rate became an obvious target." [3]

Citrini Research, "The 2028 Global Intelligence Crisis"

The scenario projects that by 2027, Mastercard's quarterly reporting will show the first visible signs of agent-led price optimization pressure in discretionary categories, representing what the report calls "the point of no return" for card network economics [3]. Shares of Visa fell approximately 5%, Mastercard 6%, and American Express more than 7% on the day the report circulated, while the S&P 500 declined more than 1% [4]. By the following session, all three had edged higher in overnight trading, and retail sentiment on Mastercard moved to "extremely bullish" on StockTwits [3].

Who Actually Holds the Risk

The market reaction may have been technically misdirected. Louis Amira, CEO of Circuit and Chisel, a company building payment infrastructure for AI agents, draws a precise distinction: Visa and Mastercard are network operators taking a minimal cut measured in basis points, while the majority of the 2-3% interchange fee is absorbed by card issuers, specifically card-centric banks including American Express, Synchrony, Capital One, and Discover [4].

"Agentic commerce routing around interchange posed a far greater risk to card-focused banks and mono-line issuers, who collected the majority of that 2-3% fee and had built entire business segments around rewards programs funded by the merchant subsidy." [3]

Citrini Research, "The 2028 Global Intelligence Crisis"

Visa and Mastercard's revenue model as technology companies, routing transactions and providing fraud infrastructure, is less exposed to the specific fee compression Citrini describes than the issuer layer. The counterargument does not eliminate the risk: if agentic commerce bypasses card rails entirely, volume losses affect the networks regardless of their share of the fee [4]. However, the gradual nature of any infrastructure shift, and the fact that firms are already exploring stablecoin settlement for 1-2% cost reductions without abandoning cards, suggests adaptation is possible before any collapse scenario [4].

Sources: Yahoo Finance / Bagalkote, Citrini Research via StockTwits [4][3]

American Express faces the compound risk Citrini identifies. Its customer base, concentrated in white-collar professionals, is precisely the workforce facing the largest AI-driven productivity shifts. Agents routing discretionary purchases around AmEx's premium interchange model could compress revenues from both sides: fewer high-spending customers and lower per-transaction economics [3].

Mastercard Moves First: Agent Pay and the Crypto Defense

Mastercard is not waiting for the scenario to materialize. The company has already completed what it describes as Europe's first live payment initiated by an AI agent within a regulated banking framework, executed in partnership with Santander through its Agent Pay technology [1]. The demonstration establishes that Mastercard's infrastructure can be the rail for agentic commerce rather than a victim of it, positioning the network as compatible with, rather than threatened by, autonomous payment flows [1][5].

The strategic two-track is deliberate. Agent Pay addresses the AI threat by making Mastercard the infrastructure layer that AI agents call, rather than the fee structure they optimize against. Simultaneously, the MetaMask Card partnership and broader crypto integration push extend Mastercard's presence into self-custody digital asset spending, stablecoin settlement pilots with Circle, Paxos, and Nuvei, and the Multi-Token Network for tokenized deposits [1][5].

"These moves sit at the core of its business as a global payments infrastructure. Mastercard is positioning its network as compatible with AI advancements, addressing risks of AI bypassing card networks." [1]

Yahoo Finance / Simply Wall St analysis

Mastercard has also created a new Director of Crypto Flows position, signaling organizational commitment to building crypto payment systems as a durable revenue line rather than a strategic experiment [5].

Visa Leads the AI Race: The Evident Benchmark

The competitive intelligence on AI deployment at the network level comes from a global benchmark published in late February 2026, which ranks Visa as the leader of the payments industry's three-way race for AI supremacy, with Mastercard effectively tied and PayPal in close pursuit [5].

Alexandra Mousavizadeh, co-CEO of Evident, the firm that produced the benchmark, frames the historical context: "Payments firms adopted AI out of necessity long before many other industries. Their business models demanded it. Companies who invested early, like Visa and Mastercard, have gained a clear advantage over their peers, both in AI capabilities and the value their deployments are realising" [5].

Visa's leadership is measured across four pillars: talent, innovation, leadership, and transparency. The network's AI is institutionalized across core transaction infrastructure, with particular maturity in fraud detection, cybersecurity, and network-level risk reduction [5]. Visa, Mastercard, and AmEx together account for nearly half of the payments industry's entire AI talent base, despite the payments sector having over 30% more AI-focused workers than other financial institutions relative to total employee counts [5].

Source: Evident benchmark via Finextra, February 25, 2026 [5]

The AmEx Settlement: Antitrust as Context

American Express entered 2026 also managing a $17.5 million antitrust class action settlement arising from the case Moskowitz, et al. v. American Express Co., filed in the Eastern District of New York [6]. The suit alleged AmEx violated antitrust law through "anti-steering" rules that prevented merchants from directing customers to competing card networks, specifically Visa and Mastercard, effectively forcing merchants to absorb AmEx's higher fees and pass costs to non-AmEx cardholders [6]. AmEx denies wrongdoing.

The settlement covers Visa and Mastercard debit card holders who made purchases from qualifying merchants across multiple US states between 2015 and 2022, with claims submission open through May 19, 2026 and a final approval hearing scheduled for June 17, 2026 [6]. The case is not existential for AmEx, but it provides a textural backdrop: the antitrust scrutiny on payment network fee structures predates the agentic commerce debate by years, and regulatory interest in interchange economics is not a new phenomenon.

The Structural Question

The Citrini scenario does not require AI agents to eliminate card rails by 2028 to be significant. It only requires that a meaningful share of discretionary commercial transactions, particularly those executed autonomously by AI agents in business-to-business or productivity workflows, begin defaulting to stablecoin settlement rather than card authorization. At the volumes that agentic commerce could eventually reach, even a 10-15% diversion of transaction flow would represent a material revenue compression for issuers and a volume signal that forces network-level strategic response.

Mastercard's Agent Pay demonstration with Santander is the clearest evidence that the networks understand the stakes. Making the card network the authentication and authorization layer that AI agents trust, rather than the fee structure they optimize against, may be the durable answer. Whether the timing of that repositioning precedes or follows the first visible volume shifts will determine how the market prices these networks through the next cycle of the technology transition.

References

[1] Yahoo Finance / Simply Wall St, "Mastercard Pushes AI and Crypto Payments While Shaping Future Fee Economics," March 2, 2026. https://finance.yahoo.com/news/mastercard-pushes-ai-crypto-payments-171830746.html

[2] Nasdaq / Zacks, "Blockchain Meets Checkout: Mastercard Plays the Long Game," March 2, 2026. https://www.nasdaq.com/articles/blockchain-meets-checkout-mastercard-plays-long-game

[3] StockTwits / Reuters, "Visa, Mastercard, AmEx Could Be Gutted by AI Agentic Commerce Threat, Citrini Research Warns," February 24, 2026. https://stocktwits.com/news-articles/markets/equity/visa-ma-amex-could-be-gutted-by-ai-agentic-commerce-threat-citrini-research-warns/cZRvmLoR4zg

[4] Yahoo Finance, "Visa, Mastercard Aren't the Real Casualties in Citrini's AI-Stablecoin Scenario," February 28, 2026. https://finance.yahoo.com/news/visa-mastercard-arent-real-casualties-133113889.html

[5] Finextra, "Visa Leads Payment Industry's Three-Horse Race for AI Supremacy," February 25, 2026. https://www.finextra.com/newsarticle/47361/visa-leads-payment-industrys-three-horse-race-for-ai-supremacy

[6] Top Class Actions, "$17.5M American Express Antitrust Class Action Settlement," February 25, 2026. https://topclassactions.com/lawsuit-settlements/open-lawsuit-settlements/17-5m-american-express-antitrust-class-action-settlement