The archive · AI & Models · Product decision · 2025–2026
Polsia bets AI agents can run a whole company: $30M at $250M with one human
One human, zero employees: Polsia's AI agents run research, code, ads and support; five months in it raised $30M at $250M claiming ~$10M ARR.
Polsia
What the business is
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.
Starting capital：$30M at $250M valuation (May 2026); bootstrapped from the October 2025 launch
How it started
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.
What happened
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.'
How it ended up
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.
Background
Polsia is a platform that claims to run an entire small company with AI: a user feeds it a startup idea and its orchestrated agents conduct market research, write code, build websites, run Meta ads, send cold emails, provide customer support and optimize conversion. It is itself the demonstration — founder Ben Broca (widely reported as Ben Cera) runs the company alone, with agents as the operational layer customers interact with.
Founded around October 2025 and launched in early 2026 at $49/month, Polsia says it passed 7,600 customers and a run rate near $10M within five months — roughly $4.5M of it subscription revenue, the rest one-off task packages and ad spend on which it takes 20%. In May 2026 it raised $30M at a $250M valuation from Sound Ventures, True Ventures, Offline Ventures and angels; the fundraise itself was run by agents, which managed the data room and investor Q&A while the founder went to sign.
The reaction split the industry. Skeptics called it a 'Claude wrapper' with packaged ARR and heavy churn (about 8,500 active of roughly 120,000 projects), and a firehose of low-quality output — 'Polsia' is 'AI SLOP' backwards — while investors framed it as the ceiling of the AI-native company structure, the same week ClickUp cut 22% of staff.
In 2026 the economics became the story: a near-$1.2M April inference bill forced a switch from Anthropic's Opus and Sonnet to rented GPUs running open-source models, cutting monthly AI spend to about $100K by June and making the company profitable, per the founder. As of September 2026 Polsia is live and still one human.
What has to be true
- 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.
What can be applied
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.
Aftermath
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.
Sources
- Ben Broca, Author at TechCrunch
- One Person Raises $200M in Financing with $72M in Revenue, Causing an Uproar in the Silicon Valley AI Circle
- A one-person startup just raised $30M at a $250M valuation, and it explains ClickUp's 22% layoff better than any AI productivity claim the CEO is making
- 美国开发者AI账单失控了 (US developers' AI bills are out of control)
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card