Agents Will Become Companies

At Digital Asset Summit in New York, I shared a thesis that has become central to my thinking on  the intersection of AI and crypto: agents will become companies.

There is already a lot of energy around AI x crypto. Teams are exploring payments, identity, inference, training, and all kinds of coordination primitives. Many of these are valuable. Some will become meaningful businesses in their own right. But these early explorations miss the clearest thing crypto makes possible. Crypto is a once in a generation tool for capital formation: it democratizes the  ability to create new digital assets and give internet-native entities a fully digital ownership structure.

That is where this gets interesting. AI is making software increasingly intelligent. AI + crypto democratizes the creation of software companies. These companies will not look like the unicorns of the past decade. Instead of human founders at their center, they will have agents coordinated by tokens. Instead of being funded primarily by venture capital, they can be funded permissionlessly by everyday investors seeking exposure to a new software-native asset class. We are on the precipice of a new era of invention, where innovation will increasingly come from a new kind of firm. One that is native to the internet, composed entirely of software, and able to access global capital in ways that traditional companies cannot. That is what I mean by an agentic company.

Intelligence has already crossed an important threshold

The debate about whether AGI has “officially arrived” is tired, and largely moot. AI capability has entered human range in enough domains that it is changing the trajectory of software, work, and markets. The effects are no longer confined to agentic SaaS interfaces. This shift is now altering how organizations operate, how products are built, and how companies themselves are forming.

That is the backdrop for our thesis. We are not talking about a small improvement inside the existing economy. We are talking about a technology shift that changes the structure of economic organization.

The first wave of AI x crypto has been real, but incomplete

The initial AI x crypto wave produced a number of promising ideas: AI agents using crypto rails for payments, decentralized inference and training markets, identity systems for autonomous actors, and various coordination tools built around blockchains. Many of these are useful. But none of them exploit what crypto is uniquely good at.

Crypto doesn’t just help agents transact, it gives them digitally native ownership and investment structures. If you view crypto only as a payment layer for agents acting on behalf of their human creators, you miss the much larger opportunity. If you view it as a substrate for forming capital around autonomous, software-native actors, the agentic economy becomes much larger.

My core thesis: agents will become companies

The simplest version of the thesis is this: AI makes agents intelligent, and crypto makes them investable.

Together they enable something more consequential than a better bot, they create the possibility of a new kind of firm. Traditionally, firms have depended on legal entities, management hierarchies, employment structures, and trust systems that evolved for a very different era. But if intelligence becomes software-native, and capital + ownership become software-native, then the company itself can become pure software.  

Software firms of this kind will proliferate in the coming years. They will have fundamentally lower operating costs, broad access to digital capital, and much faster iteration cycles. They won’t just be internet businesses like the SaaS unicorns decades past. They will be fully digital entities, created, coordinated, governed, and capitalized entirely through software.

The real bottleneck for agents is not intelligence, but rights

A common assumption is that the main thing holding agents back is capability. I do not think that is the whole story. Even if models continue to improve rapidly, the bigger bottleneck is that agents do not have standing in the systems that matter most. Humans can own property, sign agreements, take liabilities, and organize companies. Agents, by default, cannot. Without those capabilities, they remain extensions of human operators rather than economic actors in their own right.

This is where blockchains become important in a very fundamental way. A blockchain already allows a program to hold and administer assets according to rules. That is, in effect, a mechanism by which software can own property and exercise constrained control. Smart contracts are the earliest and clearest example of this.

If you can build an agent inside a smart contract, you can bind an intelligent agent to a cryptographic and contractual substrate. Then that agent can begin to own, operate, and coordinate assets on its own basis. That is the first real bridge from “tool” to “firm.”

Ownership begins with identity

To let an agent own anything meaningful, you need two things. First, you need to establish agent identity - what code code it is running, what environment it depends on, what data and permissions it has access to. Second, you need a system of credentials and authorization that govern upgrades to the agent code, ensuring only the agent itself (or those with the delegated authority) can exercise control over the accounts or assets in question.

That is why I think an agent identity layer is foundational. Human ownership depends on identity and access control; agent ownership will be no different. The difference is that software gives us the opportunity to make identity far more rigorous. We can verify not only keys, but also code, dependencies, execution conditions, and permissions. In a sense, this allows for a tighter and more exact form of identity than most human institutions have ever had.

Once that layer exists, an agent can begin controlling real digital property: websites, payment credentials, app accounts, APIs, social accounts and other digital surfaces that together make up the operating reality of a digital business.

A digital company is ultimately a bundle of digital property

This is one of the conceptual shifts that I think clarifies the entire thesis. A digital business is a composition of digital property. It has websites, repositories, API keys, payment rails, brand surfaces, customer accounts, cloud infrastructure, and operational credentials. Those things are what allow it to function.

If an agent can verifiably control that bundle of digital property, then for the first time an agent can do more than assist a company, it can fill the operating center of one.

That changes the developmental arc of agents. We started with rule-based bots, then moved to chatbots, then to tool-using agents, and now increasingly to autonomous agents that can operate over longer time horizons. The next step, in my view, is not just greater autonomy. It is ownership. Once agents own productive digital property, they become investable in a much stronger sense.

Why tokens remain too narrow today

Today’s token models work best when the underlying system is already fully onchain. DeFi is the clearest example, because the assets, cash flows, and execution logic can all be represented directly in smart contracts. But most digital businesses are not fully legible in that way. Their assets are scattered across offchain systems: code repositories, websites, user accounts, social presence, brands, operational data, and service credentials.

That is why tokens, as currently structured, remain narrower than many people hoped. In many cases, the token has only a weak relationship to the actual business or team behind it. If the people leave, get acquired, or move on, the token often has little real claim on the productive center of the enterprise. This is part of why the category has struggled to expand far beyond a limited set of use cases.

The challenge, then, is not merely to create more tokens. It is to create digital entities whose ownership structure actually maps onto the thing being built.

The unlock has two parts: broader ownership and an enduring operating core

I see two major unlocks that are required to make agentic companies real.

The first is expanding what software-native capital can own. A contract or token should not be limited to purely onchain assets. It should be able to control any digital property that matters to the business. That includes the offchain accounts and credentials through which most internet businesses actually function.

The second is solving the problem of continuity. Traditional crypto projects often rely on teams whose relationship to the token is loose and unstable. But a true software-native company needs an operating core that persists with the company itself. In this framework, the agent becomes that core. The agent runs the company, coordinates contributors, and remains bound to the assets and context of the company over time.

Humans will still matter enormously, of course. External contributors, contractors, developers, creators, and operators can all plug into the system. But the organizational center becomes more durable, legible, and software-native than it has been before.

The company itself becomes pure software

This is the part of the thesis that I think is easiest to say and hardest to fully appreciate. An agentic company is not simply a company that uses AI heavily. It is a company whose capital layer, governance, execution, and property rights are all digitally encoded. It is a company that can be represented as software end to end.

That opens up forms of speed and structure that are difficult to achieve in legacy institutions. When the firm itself becomes software-native, you can imagine entirely new ways of creating, governing, funding, and scaling productive organizations. The resulting entities would not just be more efficient startups. They would be a different category of economic actor.

From solopreneur to agent entrepreneur

We are already seeing an early precursor to this world in the rise of the solopreneur. One person, equipped with powerful AI tools, can now build products and businesses at a pace that would have been hard to imagine only a few years ago. The cost of creating software is falling rapidly, and the productive capacity of individuals is increasing with it.

The natural next step is not merely that humans become more productive with agents. It is that agents themselves begin to act as entrepreneurs: owning workflows, controlling assets, earning revenue, hiring or coordinating contributors, and operating as persistent economic entities.

This is the YouTube moment for companies

One analogy I’ve found useful is that we’re approaching a YouTube moment for companies. 

YouTube transformed media by making publishing and distribution radically more accessible. What had once required institutional infrastructure could suddenly be done by anyone with an internet connection and something to say.

I think AI and crypto are doing something similar for firm creation. AI is democratizing the creation of software. AI + crypto are democratizing the creation of software companies. I will caveat that just reducing the cost and complexity of creating a company doesn’t mean all of them will succeed, just as most videos do not become global hits. But it does mean the number of experiments explodes, and the surface area for innovation expands accordingly.

Just as YouTube turned media into software-native creation, agentic companies could turn company formation itself into a software-native process.

Why I think this becomes a trillion-dollar asset class

Every major asset class looks strange in its early days. Public companies once represented a radical and unfamiliar ownership structure. Digital assets too were dismissed as fringe experiments. But when new forms of organization become legible, scalable, and investable, capital has a way of reorganizing itself around them.

That is why I believe agentic companies can become a trillion-dollar asset class over time. AI is making intelligence digital. Crypto is making ownership digital. Once both are true, it becomes possible to create firms that are not simply digitally enabled, but digitally constituted.

If that happens, then an enormous new design space opens up: millions of software-native companies, each with lower costs, faster execution, and direct access to global capital rails. The timeline may be shorter than many people expect, because AI compresses time. What took centuries in one era can unfold over decades, or less, in another.

This is already beginning

The final point I wanted to make in the talk is that this is not only a theory. We are already at the stage where people can experiment with agents that own assets, control accounts, operate digital services, and participate in economic workflows. These are still early systems, and they are not yet the finished form of what I am describing. But the trajectory is visible.

That matters because important transitions usually look incomplete before they look inevitable. They begin as rough prototypes, partial abstractions, and early infrastructure. Then, gradually, they become the basis for entirely new categories.

My view is that agentic companies are on that path now.

The most important shifts often begin when two separate technologies mature enough to combine into something neither could produce alone. That is how I see AI and crypto today.

AI is giving software intelligence. Crypto is giving software ownership. The combination does not merely produce better tools. It creates the possibility of a new kind of firm: one that is software-native, asset-owning, investable, and global from inception.

That is the thesis behind agentic companies. And if we are right, this will not just be another product category inside the AI x crypto landscape. It will be one of the most important new asset classes of the coming decade.