As artificial intelligence agents evolve from answering questions to taking actions and spending real money, a contest is forming over which financial infrastructure will power machine-driven commerce — and who will control it. Experts say stablecoins and blockchains could become the payment rails that AI agents use to buy data, computing power and other digital services. The emerging battle centers on whether agentic finance develops on open networks or becomes concentrated among a small number of banks, payment providers and technology platforms.

'Follow the Agents'

Cathie Wood has spent years telling investors to "follow the developers" when trying to figure out where technology is headed next. Now the ARK Invest chief executive thinks investors may need to follow something else entirely: the agents.

"We're probably going to be talking more and more about 'follow the agents,'" Wood said during a panel at Robinhood's Summit in Houston on Wednesday. She was referring to AI agents — software that can carry out tasks on a person's behalf rather than simply answering a question or generating text.

The comment was brief, arriving near the end of a wider discussion about AI, private markets and technology investing. But it points to a question that is becoming harder to ignore as companies race to build AI agents: What happens when AI stops simply giving people answers and starts spending their money?

Why Agents Could Be the New Demand Signal

Developers have long helped show which technologies are gaining momentum, because engineers tend to gravitate toward tools they find useful. If millions of AI agents begin making their own choices about which software, services and networks to use, their activity could offer another way to see where demand is heading.

But those agents will also need a way to pay — and how that question gets answered is already shaping up as a contest between open and closed systems.

Chalom: Keep Agentic Finance Open

In a post last month, Joseph Chalom, co-CEO of SharpLink and the former head of digital assets at BlackRock, argued that the financial system used by AI agents should not end up controlled by a small number of banks or technology companies.

"A world full of intelligent agents means nothing if a handful of companies decide where your money can go," Chalom wrote in the final installment of a three-part series on agentic finance.

In Chalom's view, the issue is not simply whether an AI agent can spend money. It is how much power people hand over to that agent, and who controls the financial system behind it. A person might authorize an agent to spend up to $500 booking a hotel, for example, without granting it unlimited access to a bank account. The user should also be able to cancel that authority and review a record of what the agent did.

Chalom also argued that people should be able to move their agents between financial providers rather than becoming locked into one company's system. An agent should be able to carry its identity, financial information and permissions to another provider, in much the same way a person can take a phone number from one carrier to another.

Where Crypto Fits In

That is where he sees a role for crypto.

Chalom said open blockchains such as Ethereum (ETH, $2,709.61), the network that underpins much of the decentralized finance ecosystem, could provide a common financial network that different agents, and companies can use. Instead of every AI company building its own closed payment system, agents could move money across a shared network without a single technology company or bank sitting in the middle of every transaction.

The argument also gives Wood's idea of "following the agents" a financial dimension. If agents begin carrying out more tasks on their own, investors could watch not only which AI models and software they choose, but also how they pay for things and which financial networks they route their spending through.

BlackRock Sees Machine-Grade Payment Demand

BlackRock drew a similar connection in a September paper examining the overlap between AI and digital assets. The asset manager argued that AI agents could create new demand for payment systems built for machines. An agent might need to pay for an API call, purchase data from another service or rent computing power — all without waiting for a person to approve each transaction.

Stablecoins and blockchains could be one way to handle those payments, BlackRock said. Stablecoins, cryptocurrencies typically pegged to assets such as the U.S. dollar, can move around the clock, while blockchain-based payment protocols can let software send small payments directly to other software. Coinbase's (COIN) x402, for example, is named for the web's long-unused HTTP "402 Payment Required" status code and is designed to let machines pay for online services such as data or API access.

Early Signals on Crypto Rails

There are already signs of agents showing up in other parts of crypto. Coinbase CEO Brian Armstrong said on X that "Grok is the leading client for agentic traders on Coinbase currently," though he did not provide figures or further details on the activity.

Crypto Won't Have the Market to Itself

Stripe, Visa, Google and OpenAI are among the companies developing ways for agents to make purchases, and BlackRock said traditional payment systems will remain important.

That competition could make Wood's advice to "follow the agents" especially relevant for crypto investors. If AI agents become major economic actors, watching where they transact could offer a new way to measure whether stablecoins and blockchains are gaining real-world use — or whether most agent activity stays on traditional payment networks.

This article is based on reporting first published by CoinDesk.