The crypto industry has spent years talking about mass adoption as if it would arrive through more users downloading wallets, learning private keys, and manually clicking through blockchain interfaces. But a very different vision is now taking shape. According to Vibhu Norby, who leads product strategy and AI adoption at the Solana Foundation, the future of on-chain activity may not be primarily human-driven at all. Instead, he believes AI agents, bots, and large-language-model-based wallets could dominate blockchain usage within the next two years.
That is a bold claim, but it captures a much larger shift that is beginning to define the next chapter of crypto infrastructure. The discussion is no longer only about people trading tokens or sending funds between wallets. It is increasingly about software agents acting on behalf of users, businesses, and digital services—executing payments, interacting with protocols, creating wallets, making decisions, and coordinating economic activity across blockchain networks without requiring constant human clicks.
If that model takes hold, it could radically change how blockchains are used, how payments are structured, how products are designed, and even how people think about the role of a wallet. The user interface, as Norby puts it, may increasingly disappear into language. In practical terms, that means users may simply tell an AI what they want done, while the agent handles the technical execution in the background.
This is not just a speculative concept anymore. On Solana, AI agents are already reportedly making millions of transactions, especially around small, usage-based digital services. That suggests the industry may be moving toward a model in which automated, machine-driven microtransactions become one of the most important use cases for public blockchain infrastructure.
From passive assistants to active economic actors
The shift matters because AI agents are no longer being framed merely as smart assistants that answer questions or organize information. They are increasingly being described as actors capable of doing real work. In crypto, that means they can move from recommendation to execution.
A traditional assistant might suggest a trade. An AI agent in a blockchain environment might actually place it.
A normal app might remind a user to pay for a service. An AI wallet could pay for that service automatically when certain conditions are met.
A human user might once have had to compare fees, sign transactions, and navigate interfaces manually. An agent-based product may eventually do all of that through natural-language instructions and machine-readable infrastructure.
This is the core of the thesis. If agents can reliably perform blockchain actions, then the center of gravity moves away from human input and toward automated execution. The blockchain becomes less of a visible destination and more of a settlement layer operating beneath intelligent software.
That is why the prediction sounds so striking. Saying that 99.99% of on-chain transactions could be driven by agents, bots, and LLM-based products is really another way of saying that blockchains may become machine economies before they become fully user-friendly human economies. Instead of waiting for billions of people to behave like blockchain power users, the industry may get billions of software actions flowing through networks first.
Why Solana sees itself at the center of that shift
Solana’s positioning here is not accidental. Its broader pitch has long centered around speed, throughput, and low transaction costs—three features that become especially important in an AI-driven transaction environment.
If the future consists of agents executing countless small payments, calling APIs, settling pay-per-use services, and interacting with on-chain applications constantly, then the cost and latency of those transactions matter enormously. A blockchain that is expensive or slow becomes harder to use for frequent, automated machine activity. A blockchain that is cheap and fast becomes much more attractive as a background infrastructure layer.
That helps explain why Solana is leaning aggressively into this story. According to Norby, the network is already seeing major traction in so-called agentic payments through x402, a framework for internet-native payments that can enable machines to pay for digital resources directly. His claim that Solana now accounts for a large share of these agentic payment flows reflects an attempt to place the network at the center of a new economic model: one where software pays software.
This idea may sound futuristic, but the commercial logic is actually straightforward. Today, many digital services are bundled into subscriptions or fixed payment plans. In an AI-native environment, that model could start to shift toward usage-based payments. Instead of paying a flat monthly fee for access, an agent might pay per query, per transaction, per data retrieval, per compute task, per minute of service, or per API call.
That would create a much more granular digital economy—and blockchains capable of handling constant low-cost settlement could become its core plumbing.
The UI disappearing into language
One of the most important phrases in this vision is the idea that the user interface is disappearing into language.
That phrase deserves attention because it points to a major design shift. For years, crypto products have struggled with usability. Wallets, seed phrases, network settings, fee structures, and transaction signing flows have remained obstacles for mainstream adoption. Even relatively experienced users often find blockchain interfaces cumbersome compared with the smooth experience of traditional fintech apps.
AI changes that equation by introducing language as the primary interface layer. If users can simply say, “Send this payment,” “Move funds into a yield strategy,” “Buy this asset at a certain price,” or “Create a compliant wallet for this use case,” then the product no longer needs to expose every underlying technical step to the user.
In that model, blockchain becomes abstracted. It is still there, still crucial, still doing the actual work—but it fades into the background. The AI layer becomes the visible face of the interaction, while the chain becomes an invisible execution and settlement network.
This is a profound change because it could solve one of crypto’s longest-running problems: the complexity of direct interaction. Instead of making humans better at using blockchains, it may be easier to make AI better at using them on behalf of humans.
Solana is building infrastructure, not just making predictions
What makes Norby’s comments more notable is that they are not presented as theory alone. His team has reportedly been building infrastructure to support exactly this transition.
In early February 2026, a new product group was assembled with the goal of creating AI-ready interfaces for both enterprise users and ordinary consumers. That initiative led to the Solana Developer Platform, a suite of APIs designed around payments, tokenized assets, and compliance tooling.
That matters because if AI-driven on-chain activity is going to scale, it cannot depend only on grand ideas. It needs standardized tools. It needs APIs. It needs machine-readable systems. It needs compliance layers. It needs ways for institutions to connect without having to rebuild everything from scratch.
This is where the conversation becomes much more practical. Solana is not only arguing that AI agents will matter. It is trying to make itself the easiest place for those agents to operate.
That effort becomes even more interesting when major financial names such as Mastercard and Western Union are mentioned as beginning to integrate with the platform. Whether those integrations remain limited or expand more significantly over time, the symbolic importance is clear. Solana is trying to position its infrastructure not just for crypto-native developers, but for globally recognized financial actors who may want programmable payment rails compatible with AI systems.
Machine-readable skill files and autonomous learning
Another technical detail highlighted by Norby may sound small, but it points to a much bigger direction. Solana reportedly became the first major blockchain to place a machine-readable skill file at the root of its website.
That may not sound exciting to non-technical readers, but it matters because it means AI agents can more easily learn how to interact with the network on their own. In practical terms, an agent can autonomously understand how to create wallets, execute transactions, and work with on-chain programs without requiring a human to handhold every step.
This is crucial for the broader AI-agent thesis. If blockchain systems want machines to use them natively, then they must become readable and understandable to machines at the protocol and developer-experience level. Human-readable documentation alone is no longer enough. The infrastructure has to become agent-friendly.
That is what makes this moment feel different from past crypto hype cycles. This is not just another story about token speculation or consumer wallets. It is about machine usability, protocol accessibility, and the possibility that blockchains become coordination layers for software-driven economies.
A broader ecosystem is already taking shape
The movement is also not limited to the Solana Foundation.
Outside Solana, the ecosystem of AI-agent development tools in crypto is expanding quickly. Open-source frameworks such as ElizaOS are gaining major attention, with strong traction on GitHub. That kind of developer interest matters because infrastructure shifts tend to begin with tool adoption before they show up fully in end-user behavior.
Meanwhile, other ecosystems are reporting significant levels of autonomous activity. The Virtuals Protocol ecosystem, for example, has highlighted millions of jobs completed by autonomous agents. That kind of metric suggests that the idea of AI-driven execution is no longer confined to experimental demos. It is beginning to become operational.
At Coinbase, the groundwork has also been laid through products and frameworks designed to let AI interact with blockchains more directly. AgentKit and x402 are examples of how large crypto platforms are trying to prepare for a world where software, not just people, becomes a major source of transaction activity.
This broader ecosystem matters because it makes the trend more credible. If only one network or one executive were making these claims, the argument would look narrow. But when multiple major players are building tools for AI-blockchain interaction at the same time, it suggests the industry sees a genuine infrastructure opportunity.
Why AI-driven transactions could explode so quickly
The prediction that AI agents could dominate on-chain activity within two years sounds extreme, but there are structural reasons it could happen faster than many people expect.
First, machines scale much faster than humans. A user might manually perform a few on-chain actions per day, or perhaps far fewer. An AI system operating across services, wallets, markets, and workflows could generate hundreds or thousands of economically meaningful actions in the same period.
Second, machine-driven payments fit naturally with the internet’s shift toward modular digital services. If applications increasingly rely on APIs, compute services, micro-access, premium data feeds, and real-time usage metering, then automated payments become much more useful than traditional human-managed billing systems.
Third, once agents start interacting economically, they tend to create network effects. If one service can pay another directly, and that second service can pay a third, a chain of automated coordination begins to emerge. In that world, blockchains no longer serve merely as speculative environments. They become active settlement layers for machine-to-machine commerce.
Fourth, AI systems do not complain about user experience the way humans do. They do not get tired of signing transactions, switching interfaces, or comparing payment rails. If the infrastructure is machine-readable and reliable, they can use it far more efficiently than the average retail user ever could.
That changes adoption math dramatically. Mass human onboarding is difficult. Mass machine execution could happen much more quickly once the tooling is ready.
The economic model behind agentic payments
At the center of this conversation is a concept that may become increasingly important: pay-per-use digital economics.
Traditional software monetization often depends on subscriptions, bundles, tiers, and long billing cycles. That works reasonably well for human users, but it is less natural for AI agents operating in real time. Agents do not necessarily need monthly bundles. They need instant access to resources as required.
That could mean paying for:
- one specific computation,
- a single API response,
- one unit of data,
- one compliance check,
- one transaction execution,
- one retrieval from a service,
- or one limited-duration task.
This model is economically powerful because it aligns payment directly with usage. It also reduces friction for experimentation. Instead of committing to a large contract or recurring subscription, software can pay for exactly what it uses, when it uses it.
Blockchain rails are especially appealing here because they can enable direct, programmable, low-friction payments without requiring traditional card networks or account relationships for every tiny interaction. If that model scales, agentic payments may become one of the first truly native use cases where blockchain infrastructure feels economically superior rather than just philosophically interesting.
The institutional angle could matter just as much as the retail one
A lot of the public conversation around AI agents in crypto focuses on wallets, trading bots, and consumer products. But the enterprise angle may actually matter just as much.
Financial institutions, payment companies, compliance providers, and cross-border money networks all have reasons to care about AI-ready transaction systems. If AI can automate parts of payments, treasury movement, customer flows, or settlement logic, then blockchains with strong API layers may become attractive infrastructure even outside the usual crypto-native user base.
This is where compliance tooling becomes especially important. For agentic systems to operate at scale in the real economy, they cannot exist only in anonymous or purely experimental environments. They need ways to function within institutional rules, reporting requirements, and regulatory expectations.
That is why the mention of compliance APIs is so significant. It suggests that the infrastructure is being designed not only for open experimentation, but also for regulated integration. If that succeeds, the market for AI-driven blockchain transactions becomes much larger than retail crypto activity alone.
The bullish vision—and the reality check
Norby’s prediction is clearly optimistic, and it reflects a vision of the future in which blockchain activity becomes deeply integrated with AI systems. But like most ambitious predictions, it deserves both attention and caution.
The bullish case is compelling. AI agents could unlock entirely new categories of blockchain usage. Automated microtransactions, real-time machine payments, language-based wallet control, and software-driven economic coordination all make intuitive sense in a digital economy. Solana’s low-cost, high-speed design also fits naturally with this narrative.
But the reality check is just as important.
Predictions that 99.99% of on-chain activity will be machine-driven within two years assume several things go right at once. The tooling has to mature. Security has to hold up. Agent behavior must become reliable enough to trust with real value. Institutions need to integrate more deeply. Regulatory friction cannot become overwhelming. And users must actually want a future in which software increasingly manages their financial actions.
That is not impossible. But it is also not guaranteed.
There is also a philosophical question underneath all of this. If blockchains become dominated by machine activity rather than human interaction, how will that change the way people value decentralization, transparency, and user sovereignty? The answer may not be simple. A machine-driven chain economy could be efficient, but it might also look very different from the original vision many crypto users had in mind.
Brian Armstrong and the wider industry agree on the direction
One reason this trend deserves serious attention is that Norby is not the only major figure making this argument. Coinbase CEO Brian Armstrong has also suggested that transaction activity may increasingly be dominated by AI agents rather than humans.
When leaders from different corners of the industry converge around the same basic thesis, it usually means something structural is developing. They may not agree on timelines, architectures, or winning networks, but they are increasingly aligned on one idea: blockchain rails may become much more important when AI starts acting economically on its own.
This is not just a crypto trend or an AI trend. It is potentially the convergence of both.
And when two major technology waves start converging, the market often underestimates how quickly the practical use cases can begin to compound.
Conclusion
The prediction from the Solana Foundation that AI agents, bots, and LLM-based wallets could drive nearly all on-chain transactions within two years may sound aggressive, but it captures a real shift now forming across the crypto industry. The future being described is one where humans do not disappear from finance, but where software increasingly acts on their behalf—making payments, running wallets, interacting with protocols, and coordinating digital services at machine speed.
On Solana, this trend is already being framed as a live transition rather than a distant possibility. Millions of agent-driven transactions, growing infrastructure, machine-readable interfaces, and enterprise-facing API products all point to a network trying to position itself for a world where blockchain is less a manual user experience and more an embedded execution layer for AI.
Whether the 99.99% figure proves accurate or not, the broader direction is becoming harder to ignore. If AI agents become major economic actors, then the next era of blockchain adoption may not be defined by people clicking buttons on wallets. It may be defined by software paying software, settling usage in real time, and turning blockchains into the invisible financial rails of an automated digital economy.
