The term one-person company, OPC for short, sounded a few years ago more like a freelancer's self-comforting slogan. But by 2026, it's starting to genuinely compete with traditional organizations. Some solo founders, armed with nothing but themselves and a bench of AI agents, are now pulling off what used to require a department of a dozen people. More precisely, a variant has emerged: the one-team company, OTC, where a handful of people, backed by clearly divided AI agents, absorb the workload that used to belong to an entire functional department. Indie developer Pieter Levels is the most frequently cited example of this path: by his own public revenue figures and a 2025 interview, he runs a portfolio of products solo, pulling in more than $3 million a year in revenue, with zero full-time employees.
AI-native doesn't mean using a lot of AI tools
A lot of teams think they're already AI-native because they subscribe to five or six SaaS products with AI features bolted on. That's actually the most fragile form it can take. Each tool's AI capability is siloed, memory doesn't carry over between them, permissions are managed independently, and the moment one vendor changes its pricing or kills a feature, the whole workflow breaks. Being genuinely AI-native means designing a bench of agents as part of the organizational structure itself, not stapling a few AI plugins onto an existing process.
What a real one-person company's day actually looks like
Break it down and a well-run one-person company typically leans on a few clearly scoped agent roles day to day. A research agent continuously tracks industry news and competitor movement, handing the founder a summary every morning. A monitoring agent watches for anomalies in pricing, inventory, or customer feedback, and only raises a flag once a threshold is crossed. A drafting agent turns the founder's rough ideas into a weekly report, an email draft, or a first pass at social copy. The founder, for most of the day, is really doing one thing: making judgment calls. Deciding which item the research agent surfaced is worth digging into, which anomaly the monitoring agent caught deserves priority, which draft from the drafting agent is good enough to send as-is. That's the real operating logic of a one-person company. The human handles judgment, the agents handle execution and standing watch.
AI-agent-run one-person companies tend to post noticeably higher profit margins than comparable businesses run with a full-time team. The gap isn't driven by revenue scale, it's the near-zero headcount cost structure underneath it.
Why owning beats subscribing
There's an easily overlooked fragility here. If a one-person company depends on five or six disconnected SaaS AI features, its moat is actually paper-thin, any competitor can subscribe to the same tools and copy the same workflow. What makes this kind of organization genuinely hard to replicate is the founder treating the agent bench as part of their own asset base, with memory accumulating in their own workspace rather than scattered across a vendor's servers. That's also why owning your own AI stack, rather than renting fragmented SaaS AI features, becomes the dividing line for whether a one-person company survives long term. Rented capability can be rented by anyone else at the same price. But a bench of AI avatars that's been running for two years, that remembers every client preference and every past decision, can't be casually copied. A case like Pieter Levels' is worth studying not because of the specific revenue figure itself, but because it proves that the high margin a low headcount structure makes possible isn't a brief windfall, it's a structural advantage that holds up over time. That's the real reason one person replacing a department is something that can actually last.