As AI agents start moving money, finance risks becoming the slowest part of every business. Keeping up takes rails that stay open and connect to each other.
At 6:40pm on a Friday evening, an AI agent compares three suppliers, selects one and approves the order in seconds. Then the payment sits in a bank queue until Monday. That delay will get expensive for any business running on agents. Gartner expects agents to make at least 15% of day-to-day work decisions autonomously by 2028, and a growing share of those decisions involve spending money. Software now runs on a continuous clock, but the financial system still runs on the working week.
For a long time, these limitations were tolerable because people moved money and kept office hours too. When sourcing, approval and invoicing took a week, two days of settlement was a small part of the overall cycle. Now agents compress everything else to seconds, and settlement becomes the bottleneck.
Stablecoins offer a way to move money at any time, without relying on traditional banking hours, but speed is only part of the problem. As agents make payments across different markets and services, they may also need to move between blockchain networks, each with its own pools of liquidity. Money that moves instantly on one network can still be difficult to access on another. Making payments faster will count for little if the money isn’t where it needs to be. The answer is infrastructure that can settle payments continuously and economically, without making businesses wait for traditional banking hours.
Stablecoins run on software’s clock
Stablecoins address the gaps that hold agents back. They settle around the clock, move across borders in seconds and connect to software through an API, so an agent can pay a supplier on a Sunday night as easily as it queries a database.
Payment networks are adapting. Mastercard, Visa and American Express have all launched products that let AI agents pay with cards, under spending rules set for each agent. Stablecoins go further. They connect to software through an API, so an agent can pay a supplier on a Sunday night as easily as it queries a database. On low-cost networks, fees are small enough to charge per API call, per second of computing or per unit of data, payments too small and too frequent for card fees to make sense. And because the same digital dollar works in every market, businesses can run payments and treasury on one currency instead of rebuilding arrangements country by country.
Faster rails bring a new kind of fragmentation
Stablecoins fix the speed problem, but they don’t live in one place. They already run on dozens of blockchains, and payment companies, exchanges and big tech firms are all competing to set the standard for agent payments. As agents pick routes based on cost, speed and availability, a company’s dollars can end up spread across several networks without anyone deciding to go multichain.
A finance team could find its cash split across five chains, with each holding only part of what the business needs. That recreates the problem treasury teams already face with bank accounts in different countries, just on faster technology. Speed counts for little if the money for the next payment is sitting on the wrong network.
The pool of digital dollars keeps growing
Stablecoins are already widely used. USDT, the largest, has around $184 billion in circulation as of October 2026, much of it held by people and businesses in markets where local banks make holding dollars or sending money abroad slow and expensive. Agents need the same things: dollars that are always available, cheap to move and accepted wherever the counterparty sits.
As more money moves across more networks, each one becomes another place liquidity can get stuck. Bridges can move tokens between chains but add steps, fees and security risks. Liquidity pools work only when there’s enough on both sides. Another approach makes the same dollar token accessible from multiple networks, so it can move between them without a separate bridge for each transfer. USDT0 takes this approach and has processed more than $100 billion in cross-chain transfers since launching in January 2025. As financial activity becomes more automated, money needs to follow the transaction wherever it happens.
Machine speed still needs human boundaries
Handing execution to software makes accountability more important. Before an agent touches company money, someone needs to decide what it can do on its own, what needs approval, how much it can spend, who can pause a transaction and what happens if a network goes down. Every action needs to be traceable afterwards.
Onchain infrastructure can help here. Transactions on public networks are visible as they happen, and rules governing what an agent may do can travel with the payment itself: a cap per counterparty, a daily limit, a list of approved suppliers. On connected rails, those rules can hold across every network the money reaches. People set the policy and handle the exceptions; software does the routine work.
Finance becomes part of the automation
When money can move as quickly as the decisions being made around it, finance becomes part of the automation. An agent can pay for computing as it is used, settle a supplier invoice when delivery conditions are met, or rebalance cash across markets without waiting for a bank transfer or a treasury team to intervene. It does not remove the need for financial oversight. It changes how much routine work people need to do themselves.
That takes rails that run continuously and liquidity that follows the transaction across every network. Businesses that build on them will find settlement has stopped being the slowest step in their operations. Everyone else will still be waiting for Monday.
