Back to News

Business · Oct 23, 2026

Can an AI agent run a business on its own? What the autonomous microbusiness hype gets right

If you follow AI on X, you have probably seen Alex Finn's posts this month. Finn, who writes and posts about AI agents, said in mid-October that he gave an AI agent the job of building an autonomous microbusiness. In posts between October 13 and October 20, he said the business had made about $1,800, drawn roughly 100,000 visitors, and picked up around 3,000 subscribers. On October 15 he shared a template he calls Side Hustle CEO so others could try the same setup.

These are his own reported numbers, and I have no way to verify them. But the experiment is worth talking about, because it says something real about where AI agents are, and something important about where they are not.

What is genuinely new here

A year ago, an AI could write a product description or a social post. What Finn describes is an agent picking an idea, building a site, writing content, publishing, and adjusting based on results, with limited involvement from him.

Whether or not every detail holds up, the underlying capability is real. In the past few weeks, OpenAI launched dots, always-on agents with their own cloud computer. Anthropic previewed Claude Duet, a proactive agent for small business owners. Google and Microsoft both launched persistent agents for work. The tools to run long, multi-step tasks without constant supervision now exist.

What the hype leaves out

Here is what I would want any owner to keep in mind before trying to replicate a viral experiment.

A microbusiness is not your business. A small digital product with no inventory, no employees, and no existing customers is about the most forgiving environment possible for an AI agent. Mistakes cost little. Your business has customers, contracts, and a reputation.

Revenue is not profit. Reported revenue says nothing about the cost of AI usage, tools, and the time spent setting up and watching the agent.

Survivorship matters. We hear about experiments that worked. We rarely hear about the ones that did not.

Someone is still accountable. If an autonomous agent makes a promise to a customer, publishes something inaccurate, or breaks a platform's rules, the owner is responsible.

The useful lesson for a real business

Strip away the hype and Finn's experiment points at something practical. AI agents are now capable of running a defined, repeatable process end to end when the goal is clear and the stakes of a single mistake are low.

Most established businesses have processes like that. Publishing regular content. Following up with old leads. Monitoring reviews and drafting responses. Researching prospects before a sales call. These are good places to hand real autonomy to an agent, with a person checking the results.

What I would not do is hand an agent your customer relationships, your money, or your public reputation without approval steps. Not because the technology is useless, but because the cost of being wrong is real.

What this means for your business

  • Autonomous agents can now run simple, repeatable processes end to end. That capability is real.
  • Viral experiments happen in low-risk settings. Your business has more at stake.
  • Look for processes in your business that are repeatable, low risk, and clearly defined. Start there.
  • Keep a person in the loop for anything involving customers, money, or your brand.

There is real opportunity in letting agents take over repetitive work. If you want help finding the parts of your business where that makes sense, and setting it up safely, we are happy to talk.

Sources