BonrockThe One-Person Company: What AI Changes, and What It Doesn't
AI lets one person run what used to need a small team. It does not find customers, carry risk or decide what is worth building. Here is what the data says, and how to set up a one-person company that lasts.
More people are starting companies alone. Carta's Solo Founders Report found that the share of new startups on its platform with a single founder rose from 23.7% in 2019 to 36.3% in the first half of 2025. In China the idea has its own name, 一人公司 or OPC, and its own policy wave: a Fudan University team, in a report published by Xinhua in July, counted 426 OPC communities in 65 cities and 106 dedicated local policies by May 2026, from free computing credits in Shanghai's Pudong district to startup insurance in Ningbo.
One practical attraction is the range of work a founder can now draft with software. Writing code, drafting contracts, answering support email, producing a video, reconciling a ledger: an AI model can now do a first version of each in minutes. That can reduce the work needed to prepare a first version. Checking it, serving customers and operating the systems still take time; there is no fixed number of employees an agent replaces.
What "one-person" really means
The useful definition is not about headcount. The Chinese coverage of the trend puts it well: the test is whether AI has entered the core workflow, not whether there is literally one person (关键不在"一个人"，而在"AI 是否进入核心工作流"). A consultant who uses a chatbot to polish emails is still a consultant. A one-person company is one where agents do a real share of the delivery, the operations and the back office, and the founder's job shifts to deciding, reviewing and selling.
The workload changes with that arrangement, and the founder still has to account for what remains.
What AI does not change
The Fudan report is blunt about its findings. AI lowered the cost of execution, but what decides a company's direction and survival is still the founder's experience, professional skill and access to real business situations (可触达的业务场景). It lists the problems founders actually report:
- You become the "all-round employee", not the boss (全能员工而非老板). Every task that a team would split, from tax filings to customer complaints, lands on one desk.
- Work and life blur. Agents work at night, so the founder starts to as well.
- Costs move from people to computing. Some founders said model and computing bills had become most of their costs, and they rise with use.
- Competition is crowded. When tools make a product easy to build, many people build the same one.
- All the risk sits with one person. Legal, tax and liability questions that a company would share are carried alone.
Money follows the same pattern. Carta found that solo-founded companies were 30% of startups founded in 2024 but received only 14.7% of the cash raised in priced equity rounds that year. Those figures cover companies on Carta and priced equity rounds; they do not describe every solo business. A founder still needs a realistic cash runway, whether it comes from sales, savings or other income.
So the hard part of a one-person company is the same as for any company: finding people who will pay to have a problem solved. AI makes building cheap, which makes it tempting to keep building. It is easy to ship a dozen products and have no customers. A tool is not a business until someone pays for what it does.
How to set one up so it lasts
For a founder planning the operation, these are five questions to settle before adding more automation.
1. Start from a problem you already understand, for buyers you can reach. Your advantage is your years in an industry and the people who already trust you. Sell a service to them first, by hand if necessary. Turn the parts you repeat into software afterwards. That order produces revenue while you learn what is worth automating.
2. Give every agent a job description and a limit. Write down what each agent may do on its own, what it must ask about, and what it must never touch, such as payments, customer data and anything that cannot be undone. We covered how to design these approval points in Let AI Draft, Let People Sign.
3. Use the cheapest layer that works. A scheduled script is cheaper and more reliable than an agent. Keep agents for work that needs judgment, and put a monthly budget and a stop condition on each one. Track computing costs against completed work, including runs that fail.
4. Treat yourself as the single point of failure. In a one-person company, the founder is the one thing with no backup. Automatic backups, written runbooks, password management, two-factor sign-in and monitoring are not overhead; they are how the business survives a week when you are sick or travelling. If an agent can operate your systems, someone else can too, so secure them accordingly (our notes on agent security).
5. Keep a weekly scorecard for customer demand. Ask three questions every week: what did I ship, who saw it, and did anyone pay? If you keep shipping without reaching buyers or making sales, investigate distribution and demand before building another feature.
When one person is no longer enough
A one-person company does not have to stay one person. The aim is to employ people only for work that truly needs a human, such as client relationships, judgment in your field or a second set of eyes on risky decisions, and to let software handle the rest. Many founders add a part-time bookkeeper or a contractor long before they hire staff, and that is a healthy sign, not a failure.
A useful weekly review includes the work you still had to do: answering customers, correcting drafts, handling exceptions and maintaining the systems. Use that record to decide what to automate next and where another person would help.
If you are setting up a one-person company and want the back office, website, payments and agents in place quickly and safely, talk to us.
Found this helpful? Pass it on.
Working on something like this?
We help organizations build AI systems that are secure, private and ready for real load.