Traditional finance was designed around the way humans live: banking hours that match work schedules, payment systems tied to national borders, credit cards that assume you can physically walk into a store. Nikil Viswanathan, CEO and co-founder of crypto infrastructure company Alchemy, thinks that entire framework is about to become a bottleneck. The reason? AI agents don’t sleep, don’t carry wallets, and don’t care what time zone you’re in.
“You can argue that crypto was built for AI agents, not humans,” Viswanathan told CoinDesk in an interview ahead of his appearance at Consensus Miami next month.
The argument cuts against years of crypto industry effort spent trying to make Bitcoin, Ethereum, and other networks more accessible to everyday users. Seed phrases, private keys, gas fees, all of it has been treated as friction to be smoothed away through better UX. But Viswanathan is suggesting we had the frame backwards. Those features aren’t bugs to be fixed for human adoption. They’re native capabilities for a different kind of user entirely.
Traditional Finance Was Designed for Human Bodies
Viswanathan’s thesis starts with a simple observation about why financial systems work the way they do. Banks close at 5 PM because bankers go home. International wire transfers take days because paperwork moves through physical offices across time zones. Credit cards require photo IDs because merchants need to verify that the person standing in front of them matches the name on the account.
None of these constraints are arbitrary. They’re adaptations to human limitations, the fact that we get tired, that we occupy physical space, that we can only be in one country at a time.
AI agents face none of these limitations. An agent executing trades doesn’t need to pause at midnight. An agent purchasing cloud computing resources doesn’t care whether the server farm is in Virginia or Singapore. An agent managing a portfolio of stablecoins has no concept of weekend banking closures.
“All transactions for agents are online. They’re inherently global,” Viswanathan said.
The mismatch becomes obvious when you try to imagine an AI agent using traditional financial rails. How does a software program walk into a bank branch to open an account? How does it provide a home address for a credit card application? How does it sign physical documents?
These aren’t hypothetical problems. As autonomous AI systems begin participating in commerce (buying and selling services, paying for compute, executing financial strategies without human intervention), they need payment infrastructure that doesn’t assume a human body on the other end of the transaction.
Why Crypto’s Complexity Becomes a Feature
For years, the crypto industry has treated user experience as its Achilles’ heel. The average person doesn’t want to manage a 24-word seed phrase. They don’t want to worry about sending tokens to the wrong address with no recourse. They definitely don’t want to interact directly with smart contract code.
Viswanathan flips this entirely. “Agents read in zeros and ones. That’s their native language,” he said. “That’s also the language of crypto.”
A human finds private key management burdensome. An AI agent can generate, store, and use thousands of private keys without breaking a sweat (metaphorically speaking). A human struggles to understand smart contract logic. An agent can parse Solidity code in milliseconds and execute complex transaction sequences programmatically.
The properties that make crypto difficult for retail adoption, its code-native architecture, its permissionless access, its lack of intermediaries who could reverse transactions, are precisely the properties that make it powerful for machine-to-machine commerce.
Viswanathan compared the shift to the transition from physical mail to email. Writing a letter, buying a stamp, and waiting days for delivery wasn’t inefficient because postal workers were lazy. It was inefficient because the system was built around the physical movement of paper. Email didn’t just speed up the same process. It operated on fundamentally different infrastructure that computers could use natively.
Crypto, in his view, does the same for value transfer. “You can write code to manage a crypto wallet,” Viswanathan said. “You can’t write code to manage a bank account in the same way.”
The Layered Architecture of an Agent-Run Economy
So if crypto is built for agents, not humans, where does that leave the billions of people who might actually want to use financial services?
Viswanathan describes a three-layer architecture. At the base sits financial infrastructure, both traditional rails and crypto networks. On top of that, AI agents operate as the primary users, managing wallets, executing transactions, optimizing capital flows in real time. Above the agent layer sits a human interface, the simplified apps and dashboards through which people actually interact with their money.
“Just like computers operate the internet and humans use it, agents will operate finance,” he said.
This mental model has significant implications for how crypto products might evolve. Instead of trying to make raw blockchain interaction more user-friendly, builders might focus on creating agent-accessible infrastructure and letting AI handle the complexity on behalf of human users. Your financial AI assistant manages your on-chain positions. You tell it what you want. It figures out the most efficient way to execute across whatever protocols make sense.
For infrastructure providers like Alchemy, which supplies APIs, node services, and developer tools for blockchain applications, this represents both opportunity and strategic positioning. If agents become the primary consumers of crypto infrastructure, the companies that make their services agent-friendly could capture an outsized share of transaction flow.
The AI-crypto intersection has already attracted significant market attention. Earlier this year, AI-focused cryptocurrencies rallied sharply after Nvidia CEO Jensen Huang made bullish comments about autonomous AI agents, with major tokens posting double-digit gains.
What Alchemy Actually Does
For readers unfamiliar with the company, Alchemy operates as a pick-and-shovel provider for the blockchain industry. Rather than building consumer-facing applications, it supplies the underlying infrastructure that developers need to build blockchain-based products.
The company offers APIs that let applications read and write data to various blockchains, node infrastructure so developers don’t have to run their own blockchain nodes, and data services that power everything from decentralized finance protocols to NFT marketplaces to blockchain-based games. Major crypto applications rely on Alchemy’s infrastructure without most end users ever knowing the company exists.
This positioning gives Viswanathan a particular vantage point on where crypto development is heading. Alchemy sees what developers are building before those products reach the market. And increasingly, what developers are building involves AI integration.
The company has been expanding its toolkit to support Solana and other high-throughput chains alongside Ethereum, reflecting developer interest in networks that can handle higher transaction volumes at lower costs. For agent-driven use cases that might involve millions of small transactions, network efficiency matters enormously.
Unanswered Questions About Agent Finance
Viswanathan’s vision is coherent, but it prompts skepticism he didn’t directly address in the interview.
First, there’s the regulatory dimension. Financial regulators have spent decades building frameworks around the assumption that humans are the end users of financial services. Know-your-customer requirements assume there’s a customer to know. Anti-money-laundering rules assume someone can be held accountable for suspicious transactions. How do these frameworks adapt when the entity moving money is a software agent acting on someone’s behalf? Or worse, acting autonomously without clear human oversight?
Second, there’s the security problem. If AI agents control wallets with significant value, they become high-value targets. Compromising an agent’s decision-making process could be more lucrative than compromising a private key. The attack surface shifts from cryptographic security to AI safety.
Third, there’s the question of how quickly this transition actually happens. Viswanathan frames agent-operated finance as something between inevitable and imminent. But agents that can reliably handle complex financial operations without human supervision remain largely theoretical. Current AI systems excel at many tasks but still make errors that would be catastrophic in financial contexts. The gap between “AI can help you manage your crypto” and “AI agents operate finance” is substantial.
None of these objections necessarily undermine the broader thesis. They’re complications, not contradictions. But they suggest the path from here to an agent-operated financial layer will be messier than the clean architectural diagram implies.
Looking Ahead to Consensus Miami
Viswanathan will expand on these ideas at Consensus Miami next month, where the intersection of AI and crypto is expected to be a dominant theme. The conference has increasingly featured discussions about how autonomous systems might interact with blockchain infrastructure, reflecting broader industry interest in the topic.
For crypto builders, the practical takeaway might be less philosophical. If agents are going to be significant users of blockchain infrastructure, then designing for agent usability matters. That means APIs that are easy for software to interact with, transaction formats that agents can parse efficiently, and protocols that support the kind of high-frequency, low-value transactions that machine commerce might generate.
Whether Viswanathan’s vision plays out exactly as described, the underlying trend seems hard to dispute. AI systems are increasingly capable of taking actions in the world, including financial actions. Those systems need payment rails that work for them. And crypto, whatever its struggles with human adoption, offers properties that traditional finance simply doesn’t.
The industry spent a decade trying to make crypto more human-friendly. The next decade might be about making it more agent-friendly instead.
Related Reading
- Technology news
- More on Alchemy
- More on AI Agents
- More on Nikil Viswanathan
- More on Crypto Infrastructure
Sources
- https://www.coindesk.com/tech/2026/04/25/crypto-is-built-for-ai-agents-not-humans-says-alchemy-s-ceo
Reader note: this article is journalism, not a recommendation to buy, sell, or hold any asset. Do your own research before acting on any of it.