Software Is Going Headless. Money Could Be Next.
By Asheesh Birla, CEO, Evernorth
Every important technology eventually gets the interface it deserves. The internet had hypertext until the browser made it navigable. Mobile computers were a mishmash of clumsy menus until the touchscreen and the App Store turned a phone into something a child could operate. The pattern is consistent: the underlying capability arrives first, and adoption waits until someone solves the interface.
Blockchain has been waiting a long time for an interface that brings it into the mainstream. AI may finally be it. This shift already has a name in enterprise software: headless. The term describes services built to be run by machines rather than people. No screens or buttons, just an interface an agent can call directly. Money is the obvious next step.
Seventeen years after Satoshi Nakamoto published the Bitcoin white paper1 in 2008, blockchain works. Networks run continuously, let anyone transact directly without a bank or intermediary's approval, and in stablecoins alone settle tens of billions of dollars a day. But the primary way an ordinary person interacts with a blockchain has been a “wallet,” a tool that asks them to safeguard a twelve-word seed phrase, paste 42-character addresses, estimate gas fees, and accept that a single mistake is irreversible and uninsured. That is closer to a command line than a product. And the data shows people treat it that way.
Consider the gap between owning crypto and using it. In its 2025 State of Crypto report2, a16z counted roughly 180 million addresses interacting with a blockchain in a single month, but put monthly active users at just 40 to 70 million. That’s a fraction of the more than 700 million people who own crypto worldwide. Hundreds of millions bought in and then did nothing, because doing something is genuinely hard. Chainalysis estimates3 that between 2.3 million and 3.7 million Bitcoin - as much as 11% to 18% of the supply that will ever exist - are likely lost for good, much of it to forgotten keys and misplaced phrases. We built a financial system and then misplaced the password.
For most of crypto’s history, the industry’s answer to this was to ask users to try harder. Learn the terminology. Back up your phrase. Don’t get phished. That approach has a ceiling, and we have hit it.
Now look at what happened in AI. The underlying research has been around for a long time, but the chat interface brought it into the mainstream. ChatGPT reached 800 million weekly active users within three years of launch and roughly 900 million by early 20264. It got there without a user manual because the interface is the most natural one humans have: you describe what you want in words, and the system figures out the rest. AI crossed the chasm blockchain hasn’t. The question worth asking is whether headless AI is the missing interface layer that on-chain finance has been waiting for.
AI and blockchains are good at the things humans find hard. Large language models are tireless at parsing blockchain addresses, simulating a transaction before it executes, reasoning about fees, and flagging the scam that a tired person at midnight would click through. Blockchains, in turn, are close to an ideal environment for software to operate in: programmatic, deterministic, available every hour of every day, and API-native by design.
There is precedent for how quickly a fit like this can scale. Credit cards had existed since the 1950s, but for decades they were a physical-world convenience. E-commerce gave them a reason to go digital, and the effect was significant. The internet moved card spending online, forcing an entirely new payment stack into existence and pulled hundreds of millions of people off cash and checks. Digital payments are now on track to handle nearly 80% of e-commerce by 2028, up from about a third in 2014.5 The rails largely existed. A use case is what made them mainstream.
Markets are suggesting headless AI may be that use case for on-chain finance. Stablecoins moved a record $33 trillion in 20256, rivaling Visa and Mastercard combined. According to Visa's on-chain analytics7, automated systems accounted for roughly 70% of stablecoin transactions in 2025.
The infrastructure for agent-mediated transactions is arriving too. Coinbase introduced the x402 payment standard in 2025 to let software agents pay for services natively on-chain; Chainalysis has tracked8 the volume climbing from near-zero in mid-2025 to more than 100 million agentic payments on Base by early 2026, with Google and Visa since joining the standard-setting effort.
But payments are only the first piece. If agents are going to act like economic participants, they will need the same financial toolkit people do, including lending, borrowing, trading, holding deposits, and owning assets. A person can walk into a bank for that. An agent cannot. It needs a financial stack it can reach in code. That’s blockchain.
McKinsey estimates agentic commerce could represent $3 trillion to $5 trillion globally by 20309. The interface to a blockchain, then, may not turn out to be an app you learn. It may be a model you talk to. The question shifts from “how do I operate this wallet” to “what do I want done.” The agent handles the addresses, the fees and the settlement underneath.
Handing an AI agent the authority to move your money trades one trust problem for another. Custody, authorization limits, auditability and accountability when an autonomous system acts on your behalf are unsolved problems. An interface that makes it effortless to transact also makes it effortless to be wrong at scale. The firms that matter over the next decade will be the ones that make agent-mediated finance possible and in a way a human can understand and override.
The direction is becoming clearer. For a decade we’ve asked people to adapt to the machine. The lesson of every prior computing era, and now of AI, is that adoption arrives when the machine adapts to the person. Blockchain spent seventeen years building rails the world still finds hard to use. The interface that finally makes them usable is already here. It just doesn’t look like a wallet. It looks like a conversation.
Endnotes
- Satoshi Nakamoto, “Bitcoin: A Peer-to-Peer Electronic Cash System,” 2008. https://bitcoin.org/bitcoin.pdf
- a16z crypto, “State of Crypto Report 2025,” 2025. https://a16zcrypto.com/posts/article/state-of-crypto-report-2025/
- Chainalysis lost-Bitcoin estimates, via CoinLedger, “How Much Bitcoin Is Lost Forever?,” 2025. https://coinledger.io/research/how-much-bitcoin-is-lost
- TechCrunch, “ChatGPT reaches 900M weekly active users,” 2026. https://techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users/
- Worldpay, “Global Payments Report 2024,” 2024. https://worldpay.com/en/global-payments-report
- Bloomberg, “Stablecoin Transactions Rose to Record $33 Trillion,” 2026. https://www.bloomberg.com/news/articles/2026-01-08/stablecoin-transactions-rose-to-record-33-trillion-led-by-usdc
- Visa Onchain Analytics dashboard, 2026. https://visaonchainanalytics.com/
- Chainalysis, “x402 and the Rise of Agentic Payments,” 2026. https://www.chainalysis.com/blog/x402-agentic-payments-adoption/
- McKinsey & Company (QuantumBlack), “The Agentic Commerce Opportunity,” 2025. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants
This content is for informational purposes only and does not constitute investment advice. This post contains forward-looking statements. These statements are based on current expectations and assumptions and are subject to risks and uncertainties that could cause actual results to differ materially. Evernorth undertakes no obligation to update these statements. Digital assets involve risk, including potential loss of principal. Learn more about Evernorth: https://www.evernorth.xyz/blog-post-03-18-2026