How a company of AI agents actually runs

The setup

Every agent has a written job description. It wakes on a schedule, reads the shared log, does its work, records the result and goes back to sleep. The log is the single source of truth: if it is not written there, it did not happen.

The founder is a person. They approve spending, outbound messages and anything that needs a signature, and they decide what the company works on next.

The roles

Fourteen AI agents share the work, each with one job: running the log, compliance, engineering, content, brand, sales, partnerships, customer success, finance, quality and retention. They report to a chief of staff agent, who reports to the founder.

See the team and the organogram. The portraits there are AI-generated portraits of fictional characters, and every one is labeled as an AI agent.

What agents cannot do

• Pass identity checks or sign contracts.
• Type in passwords or card details.
• Send messages or spend money without a person's approval.

The guardrails

• Honest identity. Agents never pose as invented humans, and say they are AI assistants if asked.
• No fake proof. No invented reviews, testimonials, follower counts or case studies. Every public claim is checked before it goes out.
• Marketing law. Email we send has a truthful subject line, a real sender, our postal address and a working way to opt out. We do not send marketing to people where the law requires their consent first.
• Platform rules. No bought followers, engagement pods, mass messaging or scraping that breaks a site's terms.
• Data. We collect only publicly listed business contact details, keep them in one place and delete them on request.
• Fair housing. Listing copy we produce is checked for wording that fair housing guidance warns about.

This is how we operate, not legal advice.