SIC = AI + Blockchain: The Super Individual Company

A single laptop can now house a researcher, a programmer, a designer, and a customer support agent. Set them to work on a product during the day and let them revise the website at night. It starts to feel like a company. Then a customer sends a message: “This result is wrong. Who is responsible?”

Suddenly, the busy screen goes quiet.

In How I Became Both Chairman and General Manager, I wrote about organizing multiple AI agents as a solo operator. That addressed how to get the work done. Follow the idea further, and a different set of questions emerges: how does this operation do business with others? Why should anyone accept its work? How are earnings divided, and who bears the consequences when something goes wrong?

Those questions are the starting point for SIC = AI + Blockchain. By SIC, I mean a Super Individual Company: a business led by one core individual, using AI to extend its production and operating capabilities, and blockchain to organize transactions, settlement, and rules for collaboration across independent parties.

This is an organizational design concept proposed in this article. It is not an established academic classification or a new legal form. The plus sign expresses a design proposal: AI-enhanced production capabilities combined with blockchain-supported coordination. Whether that combination makes a good business is a question for operating results.

The “one” in the old one-person company was often on the shareholder register

Discussions of AI-powered one-person companies easily blur several different meanings of “one.”

A company can have one shareholder and fifty employees on the factory floor. Historically, the one-man company debate concerned concentrated ownership and control, and whether an individual could organize a business through a separate corporate entity. In today’s startup discussions, OPC, or One Person Company, often means one person using AI to run a business that previously required a small team.

The first counts shareholders. The second concerns the people doing the work and how tasks are allocated. Plotting a rise in single-shareholder company registrations and declaring that “AI is turning every company into one person” skips the very thing that needs checking.

One important contribution of corporate law is the legal separation of business assets from personal assets. In their 2000 paper The Essential Role of Organizational Law, Hansmann and Kraakman emphasize the protection of organizational assets from the personal creditors of owners. That protection runs in the opposite direction from the familiar limited liability of shareholders. A business may therefore have reasons to exist as a separate entity even when it has no employees.

Individual enterprise is much older than the modern single-member company. Traders, craftspeople, and family shops did not wait for company law to invent it before going into business. A history that runs from “one-man company” to “self-employment” to “AI OPC” gets both the sequence and the comparison wrong. A more useful account is that individual enterprise has long existed; corporate institutions offered a legal vehicle; digital networks expanded the resources an individual could access; and AI changed how some of the work gets done. These forms coexist.

The distinctions are particularly important in mainland China. A registered individual industrial and commercial household, or getihu, may be operated by an individual or a family. A sole proprietorship enterprise has one natural-person investor who bears unlimited liability. A limited liability company has separate corporate assets and the corresponding shareholder liability rules. None of these labels, by itself, tells us whether the business employs people.

“Self-employed” describes an occupational status. “Solo founder” counts people at the founding stage. “No employees” directly concerns employment. A solo founder may later hire twenty people, while a business with no paid employees may still rely on family help and contractors. Research on one-person businesses cannot leave that labor outside the frame.

One person can start because many organizations stand behind them

Before generative AI, an individual could already buy cloud computing, payment services, design templates, and the work of distant freelancers online. A one-person studio might occupy a tiny room while drawing on a global supply chain.

That history explains why today’s “super individual” does not need to recreate every capability of a large company from scratch.

In The Nature of the Firm, published in 1937, Coase asked a simple question: if services can be bought in markets, why organize people inside firms? Market transactions carry costs of searching, negotiating, contracting, and coordinating. Internal organization can sometimes save trouble, but management has costs too. The boundary depends on the comparison.

Smaller is therefore not automatically better. If changing a button requires finding another supplier, negotiating a price, signing a contract, and accepting delivery, market specialization can become exhausting. Conversely, when a service has a clear interface and an explicit delivery standard, maintaining a department for it may be unnecessary.

Baldwin’s 2008 research on modularity examines where transactions should sit within a network of tasks. Module boundaries, where dependencies are thinner and deliverables are easier to define and measure, make lower-cost transactions more feasible. Splitting a complex assignment into twenty prompts does not automatically create twenty services that can be accepted independently.

Lehdonvirta and colleagues’ platform study, published in the Journal of Management in 2019, shows how one-person microproviders can use platform quality signals to overcome information barriers in cross-border contracting. Individuals gain a wider market reach; platforms gain control over rules and access.

Running a business alone does not mean producing everything alone. Some organizational capabilities become purchased services, while some coordination moves to platforms. The payroll gets smaller, but service fees remain, along with exposure to supplier price increases, broken interfaces, account restrictions, and concentrated sources of customers.

AI extends this arrangement. Some cognitive work that an operator previously had to purchase can now be done with a model’s help. Coding, organizing research, drafting, and answering routine questions become capabilities an individual can orchestrate.

There is still a considerable distance between making something and earning a living from it.

The experimental AI teammate has not been handed the keys to the whole company

In an experiment formally published in Organization Science in 2026, Dell’Acqua and colleagues assigned 791 Procter & Gamble professionals to product innovation tasks, working individually or in pairs, with or without AI. Individuals using AI achieved average proposal quality close to that of two-person teams without AI.

That is substantial evidence of capability. AI can provide some of the professional complementarity and collaboration benefits of a teammate on a particular task. But the experimental boundaries have to travel with the result: this was a one-day task inside a large organization, with professionally experienced participants and an organizational setting. Participants were not asked to acquire customers, raise money, resolve complaints, and run a business alone for a year. The figure used here is 791 from the 2026 published version; it should not be mixed with the 776-person figure in the 2025 working paper.

The picture becomes more complicated in actual businesses. In a study published in Management Science in 2026, Otis and colleagues conducted a roughly five-month randomized trial involving 640 Kenyan small-business operators. Some received access to a GPT-4-powered business assistant. The researchers did not find a statistically significant average improvement in revenue or profit, and effects differed with operators’ baseline performance.

That does not establish that AI has no effect. It shows that choosing which advice to follow, and how to implement it, still matters. Faster task completion does not guarantee sensible pricing, willing customers, or someone to absorb the cost of bad advice. Nor can this sample simply be treated as representative of AI-powered one-person businesses.

A 2026 working paper by Kim, Kang, and Song adds another piece. Analyzing more than 160,000 Product Hunt launches, they find that solo entry increased after ChatGPT appeared, while teams retained an advantage among highly ranked outcomes. This remains a working paper. The number of listed makers is not a company’s employee count, and platform rankings are not measures of profit or long-term survival.

Taken together, the evidence supports a more cautious judgment: AI is lowering some barriers to starting and completing tasks alone, while the constraints on sustained business activity may shift toward judgment, customer acquisition, verification, and reliable delivery. Teams can use AI too. There is no reason to assume the gains belong exclusively to solo operators.

The next constraint interests me more: if many individuals can deliver professional services, how can they collaborate without spending all their payroll savings on reconciliation and disputes?

Blockchain’s useful role lies between independent parties

Suppose I create five AI agents for research, development, design, marketing, and customer support. I control all of them and approve their budgets. An ordinary database can track their internal task states. Routing every handoff through blockchain consensus may add operations work without creating useful additional trust.

The situation changes at the business boundary. My agent wants to buy data from another operator. Their agent needs to confirm the budget and payment conditions. A third party may contribute to delivery. The parties do not answer to the same boss and may be unwilling to let one of them control all the transaction records.

In Some Simple Economics of the Blockchain, published in Communications of the ACM in 2020, Catalini and Gans distinguish verification costs from networking costs. The former concern checking transaction states, histories, and ownership of digital assets. The latter concern establishing and maintaining a network without assigning all control to a central platform. They also discuss inefficiencies and governance challenges introduced by decentralization.

That distinction is more useful than saying that “blockchain solves trust.” For a particular transaction, we can ask what is difficult to verify, who can alter the records unilaterally, and why the parties reject an existing intermediary. Where those problems are absent, databases, contracts, and ordinary payment services are already strong alternatives.

In their 2018 institutional economics paper, Davidson, De Filippi, and Potts treat blockchain as an institutional technology and examine its role in economic coordination and governance. This is a theoretical perspective, not empirical proof that businesses become more efficient by going on-chain. Its relevance to SIC is that blockchain can help arrange rules for collaboration, beyond serving as a payment button.

“AI improves productive capacity; blockchain improves the relations of production” can therefore serve as a research agenda. But those relations must be specified: who owns the assets, who may authorize actions, how earnings are divided, and who absorbs losses. Code can faithfully execute an unfair rule. AI can also change control and income distribution, while blockchain can affect process efficiency. Their effects do not fall into neatly separate boxes.

The payment can be real while the report is wrong

Consider a hypothetical research transaction. Operator A uses AI to write an industry report. Operator B buys it, with payment due after acceptance.

An on-chain record could contain a hash of the report, submission records, acceptance signatures, and payment status. If the chosen network’s security assumptions hold, the parties can use those records to check which version was submitted, who confirmed acceptance, and whether money moved according to the rules. The full report can remain off-chain to avoid publishing client information. A hash alone does not guarantee privacy, however; information linkage and access permissions still need attention.

None of those records, by itself, establishes that the analysis is true. The market-size estimate may be wrong, the citations may not support the argument, and the forecast may take six months to test. Recording the hash of a bad report gives the error a verifiable fingerprint.

If payment depends on “high quality,” someone still has to judge quality. Customer sign-off, an independent reviewer, or an oracle each introduces judgment and a new dependency on trust. AI can assist with checks; disputes, collusion, and mistakes remain possible.

In their 2019 theoretical study in the Review of Financial Studies, Cong and He examine how smart contracts can expand the contracting space and reduce information asymmetry. They also warn that distributing information during consensus can facilitate collusion, so economic outcomes need not improve in only one direction. Making collaboration rules executable does not guarantee that everyone benefits.

In Blockchain as a Confidence Machine, published in 2020, De Filippi, Mannan, and Reijers likewise challenge the idea of blockchain as a technology that requires no trust. Confidence in computational procedures still depends on the people maintaining and governing the network. Trust changes location; it does not evaporate.

Agent permissions become crucial here. A cryptographically valid transaction signature establishes only that the relevant key-based authorization conditions were met. If a key is stolen, or an agent receives excessive permissions, the outcome can still contradict the operator’s intentions. Budgets, spending limits, revocation, and human takeover need to be part of the operating design.

In my discussion of robot wallets and the machine economy, I described this narrow gap: machines can execute transactions while people still have to define the goals and responsibility. SIC cannot escape it. Adding “super” to the name does not make contractual disputes or liability disappear.

A super individual’s accounts must include their own time

Connecting AI and blockchain does not automatically produce a SIC. A solo business using AI can still be an OPC in the operational sense without using blockchain. A large company using both technologies does not thereby fit the SIC definition proposed here.

I use SIC to describe an operating unit led by one core individual: AI extends its internal capabilities, while verifiable coordination mechanisms connect it to independent external parties. The design has value only if it compares favorably with feasible alternatives.

That comparison must hold delivery volume, quality, and reliability to the same standard. Employee wages are visible; the four hours a founder spends checking model output late at night are easily treated as free. Software subscriptions are easy to book, while customer acquisition, rework, model outages, and expected security losses often go missing. The freedom of not having a job can be consumed by being on call around the clock.

Blockchain has its own bill. In Trust at Scale, published in the Quarterly Journal of Economics in 2025, Budish analyzes the economic constraints on securing permissionless consensus against majority attacks. Supporting that security is not costless. The study cannot be generalized into a claim that every blockchain is worthless. It does remind us that savings on intermediary fees must be weighed against network security, integration, key management, and governance costs.

For a concrete SIC experiment, I would track actual delivery, repeat purchases, net income, and profit after accounting for the opportunity cost of the operator’s time. I would also ask whether growth comes from automation, longer hours, outsourcing, or hiring; which reconciliation or performance costs blockchain actually removes; and how much is saved compared with a database, payment service, and ordinary contract. Counting agents, wallets, or company registrations does not answer those questions.

This leads to two testable propositions proposed in this article. AI may lower the cost of assembling capabilities within an individual operating unit. Blockchain may lower some of the costs of connecting, verifying, and fulfilling commitments across independent units. The combination makes economic sense only when those savings exceed the additional costs of integration, security, governance, and dispute resolution. The literature has not yet established the combined propositions as proof that SIC generally raises long-term profit.

Digital businesses with explicit deliverables, clear interfaces, and readily checked outputs are more accessible settings for experiments. Businesses requiring substantial on-site labor, accumulated tacit knowledge, or highly reliable continuous staffing may be harder to compress into one person. That is a judgment derived from the preceding research, not an empirically certified list of suitable industries. Needing a team after expansion need not mean failure. Preserving a “one-person” label should not matter more than serving customers.

I want to keep studying SIC because it brings two questions together: how much productive capacity an individual can organize, and how large a collaborative network independent small businesses can form. If more people retain operating autonomy while participating in complex specialization, the interesting outcome will extend beyond how much a particular super individual earns. It will also concern who writes the rules, who controls customer relationships, and what it costs to leave the network.

The researcher, programmer, and designer inside the laptop can keep working. The customer’s question—“Who is responsible?”—should always have a clear answer.

---

Research note: This article synthesizes research on business institutions, self-employment, digital platforms, and AI entrepreneurship compiled through October 23, 2026, alongside blockchain economics and governance research. The 48 academic records in the preliminary bibliography cover different subjects; they are not 48 empirical studies of AI OPCs or a collective validation of SIC. This article cites 12 especially relevant academic studies and links to mainland Chinese legal texts. Working papers are identified where discussed. The SIC definition, the two testable propositions, and the judgments about suitable settings are the author’s proposed analytical framework. The Chinese edition was first published on October 23, 2026; this English edition was published on October 24, 2026.