Polsia bets AI agents can run a whole company: $30M at $250M with one human

Polsia is a 'one-person company' platform: users hand it a startup idea and its AI agents handle market research, coding, websites, Meta ads, cold email, customer service and growth, on subscriptions, ad commissions and revenue share.

That an orchestrated team of AI agents can run an entire small company end-to-end, replacing headcount with software and taking a cut of what users earn.

$30M at $250M valuation (May 2026); bootstrapped from the October 2025 launch

Ben Broca (widely reported as Ben Cera) founded Polsia around October 2025 after a year of building with AI agents, frustrated that agents could write code but still couldn't deliver complete company operations end-to-end. He built an 'agent orchestration system for end-to-end company operation' and launched the platform in early 2026 at $49/month.

Growth was fast and noisy: ~7,600 customers within five months, a claimed run rate near $10M (about $4.5M subscription ARR plus one-off task packages and ad spend), and a $30M round at $250M in May 2026 — announced the same week as ClickUp's 22% AI-driven layoffs. Critics called it a 'Claude wrapper' with heavy churn (skeptics counted ~8,500 active of ~120,000 projects) and 'AI slop' output; the founder answered 'most of it is indeed junk, but the quality of the junk is improving every week.'

Still running: after a near-$1.2M AI bill in April 2026, the founder moved inference to rented GPUs running open-source models (MiniMax, GLM); by June 2026 monthly AI spend was down to about $100K, which he said made the company profitable and enabled cheaper or free service tiers.

The zero-employee structure made the company itself the proof: if agents can run the fundraise, the data room and the product, the pitch needs no deck.

Revenue-share and ad-commission pricing aligned Polsia with users' outcomes rather than seat count, unlike classic SaaS.

The contrarian embrace of criticism — 'most of it is indeed junk, but the quality of the junk is improving' — turned a weakness into a memorable brand.

The near-$1.2M April 2026 AI bill became a forcing function: switching to open-source models on rented GPUs cut monthly inference spend about 12x and made profitability the next chapter.

A one-person, agent-run company can raise at a benchmark valuation because the headcount is the story; whether revenue is real, recurring and profitable is what later rounds will test.

As of September 2026 Polsia remains live and run by its single founder. After the near-$1.2M April 2026 inference bill, it moved to rented GPUs running open-source MiniMax and GLM models; by June 2026 monthly AI spend had fallen to about $100K, which the founder said made the company profitable and funded free tiers. Ben Broca is scheduled to speak at TechCrunch Disrupt 2026 about going from $0 to a $10M run rate in the AI era. Open questions remain: whether revenue is recurring, whether churn stabilizes, and whether 'AI-run companies' produce real profits for users.