Most people stopped carrying cash years ago. Credit cards won that battle, but they're about to lose the next one.
The reason comes down to a core design limitation of card rails: They're based on a "pull" model. When you swipe at a checkout counter, the merchant pulls money from your account. This requires authorization windows, fraud checks, charge-back mechanisms and interchange fees of 2% to 3% to cover the risk. The architecture made sense when payments happened a few times a day and the primary threat was a stranger stealing your physical card.
It makes far less sense when the "spender" is a software agent executing thousands of transactions per minute.
AI agents are entering enterprise finance faster than most realize. Gartner
The pull model breaks down here. A few seconds of authorization delay is acceptable for a human at a checkout counter. For an agent executing machine-speed operations, it's not. The 2% to 3% interchange fee that's tolerable on a $50 purchase becomes economically devastating when applied to thousands of sub-dollar transactions. The entire category of micropayments, entailing fractions of a cent for individual API calls or data queries, simply cannot exist on card rails.
The agentic economy requires "push" payments with near-zero fees. Stablecoins on Layer 2 networks or high-throughput chains (which power everything from cross-border payroll to real-time API monetization) can deliver transaction costs measured in fractions of a cent with sub-second finality. There are no authorization delays, no interchange, and no charge-backs, because the sender, not the recipient, initiates the transfer.
The clearest example of this infrastructure is
This is the infrastructure AI agents need. When an agent calls a paid API, it doesn't create an account, enter card details, or wait for authorization, because it's able to
Conditional logic becomes native to transactions, such that a smart contract can hold funds in escrow and release them only when external conditions are met. This kind of programmable money, where payment terms are embedded in the transaction itself, is impossible with cards, which can only authorize fixed amounts at a single point in time with no mechanism for conditional release.
The company's recently released 2026 trends report is about "moving and changing the narrative from being Eeyores to being Tiggers," Abbott said in the latest American Banker podcast.
Enterprise adoption is accelerating, and corporate payouts are emerging as the first major category to shift to stablecoins.
The main driver is economic. Visa just launched a stablecoin payout pilot targeting business disbursements. Revolut partnered with Polygon to power stablecoin-based remittances and payments and has processed over $690 million in volume since integration. Stripe
Building on blockchain APIs is cheaper and faster than integrating legacy SWIFT/IBAN systems for global coverage. For enterprises running agent-driven operations where thousands of automated transactions execute daily, this cost differential compounds.
The GENIUS Act, passed with bipartisan support, establishes federal rules for issuing, redeeming, and backing fiat-pegged stablecoins in the U.S. This matters for agents and enterprises because it creates a clear legal framework for programmable money to operate within regulated finance.
Before GENIUS, enterprises faced uncertainty about whether stablecoin-based treasury operations or automated payments would trigger regulatory risk. Now there's a defined path. Within weeks of passage, Visa and Mastercard embedded stablecoin rails into their networks. The infrastructure layer aligning with the regulatory layer means enterprises can build agent-driven payment systems with confidence that the underlying rails are compliant.
The result will be fewer geographic restrictions, no FX conversion fees, no settlement delays tied to banking hours. Add programmability like automated compliance checks, conditional releases, and real-time streaming and you have infrastructure that legacy rails simply cannot match.
Cards won the war against cash because they were faster and more convenient for humans. Stablecoins will win the next transition because they are faster and more capable for machines.
The pull model that made cards work, and offered authorization, fraud checks and charge-backs, becomes nothing but friction when the payer is a trusted software agent. The settlement delays become bottlenecks that break machine-speed workflows.
Enterprise finance is heading toward a world where AI agents manage treasury, execute payments and settle transactions autonomously. The infrastructure those agents need: push payments, sub-cent fees, programmable conditions, instant finality is what stablecoins were designed to deliver.
Cards were built for humans standing at checkout counters. The agentic economy needs programmable, instant, borderless money, and that's exactly what stablecoins deliver.
