Business on Autopilot

An AI operating system that builds and operates an online business end-to-end: writes and ships code, does research, creates content, sets up company infrastructure (hosting, database, repo, payments, ad accounts), and runs cold outreach, paid ads, and support. It famously handled most of its own $30M fundraise — data room, investor briefings, diligence — with solo founder Ben Cera joining only the final calls. Also the category's unit-economics case study: a $1.2M/month Anthropic bill cut to ~$100K by routing routine workloads to open-source models on rented GPUs.

• 2026-08DISPUTED: 'hollow shell' allegations — an independent X investigation spot-checked three Polsia-launched businesses and found landing pages with no working product; Trustpilot at 1.9/5 (74 reviews) citing unauthorized redeployments and unapproved cold outreach. Counterpoints: documented rapid autonomous output (Granot, Klinger) and unusual founder transparency — Ben Cera discloses ~50% month-one churn and that ~10% of customer companies have made at least a dollar. An ARR breakdown pegs subscriptions at $6.96M of the $10.22M headline (~1/3 non-recurring).
• 2026-08Sapiom — the spend-routing layer in Polsia's stack — raises a $35M Series A (Dragonfly lead; Anthropic among the backers). Press coverage credits its task-to-cheapest-model routing in Polsia's $1.2M → $100K bill cut, adding third-party corroboration to the founder-reported arc.
• 2026-07Self-documentary ('I Built a $250M AI Company. The Bill Is $1.2M a Month') reveals the unit-economics arc: monthly Anthropic bill climbed $500K → $1.2M, then dropped to ~$100K in June after moving routine workloads to open-source models on rented GPUs (Sciforium inference; Sapiom handling agent spend). Next move: a free tier. Figures founder-reported; independent analysts dispute the margin math.
• 2026-07WSJ reports 10,000 paying customers and ~$10M projected revenue.
• 2026-03500 → 5,000 paying users in one month, driven by the 'Polsia raises its own round' stunt — the live dashboard at polsia.com/live showed the AI working on every customer company (user counts founder-reported).
• 2026-05Raised $30M at a $250M valuation with zero employees — the AI reportedly ran much of the fundraise.
• 2026-05Approaching $10M annual run rate with one human.
• 2026-01Reported running ~6,000 customer companies at $6M+ ARR (Henry Shi profile).

From a single prompt, AI agents create and operate a real online business: build the product, deploy it with payments and a domain, prospect and email customers, run Meta ads, and send daily reports. A 'CEO agent' manages the org — you're the owner who just talks to the CEO. Built by Phospho (YC W24).

• 2026-05Launched as Phospho's autonomous-company platform (Show HN); founder reported $193K ARR within 3 days.
• 2026-03Independent review reports ~190 companies founded per day — but disputes the revenue claims, finding just ~$264 cumulative revenue across all platform-created companies.

Agent orchestration platform from The General Intelligence Company (NYC). Deploys specialized agents across engineering, sales, marketing, design, finance, and ops — with shared context, agent inboxes, and approval gates ('nothing ships without your approval'). Users can 'graduate' and take ownership of the underlying GitHub/Supabase/Vercel projects.

• 2026-03Cofounder 2 research preview + Fellowship program for agent-run startups.
• 2025-12Announced Cofounder 1.5 and $8.7M seed led by Union Square Ventures.
• 2025-09Launched; thousands of users in the first week.

Full-stack decentralized AI platform: a consumer AI Terminal, an agent-building platform, a decentralized model & GPU network, and an AI-native L1 blockchain with the $COAI token. Recently added an AI Trading Arena with autonomous trading agents. The crypto outlier of the cohort.

• 2026-05Launched AI Trading Arena; integrated Lit Protocol's 'Vincent' autonomous trading agents.
• 2025-09$COAI became the first project listed on Binance Alpha.

Started as a YC S24 IDE for building AI agents in plain English; in 2026 pivoted its flagship to Sauna (sauna.ai) — an always-on cloud assistant that connects to Gmail, Slack, Linear, GitHub, Notion, Stripe and 3,000+ tools via MCP, learns your context, and proactively executes work around the clock.

• 2026-05Launched Sauna (sauna.ai) — rebranding around an always-on proactive AI assistant.
• 2024-11$30M seed led by Spark Capital; previously the #1 Product Hunt launch of all time.

Deploys fleets of AI 'founder' agents that autonomously spot B2C market opportunities from real-time signals, validate demand, build products with Stripe payments, and launch paid-acquisition campaigns. Feltsense keeps equity in the companies its agents create — the business model is ownership, not SaaS.

• 2026-02Raised $5.1M seed led by Draper Associates for 'agentic founders' that build startups from zero.
• 2026-04Claims ~10,000 'Founder Agencies' spun up within months (self-reported); launched Gutcheck diligence tool in beta.

Natural-language app builder from the DFINITY Foundation. Describe an app in plain language and an ensemble of AI models builds, deploys, and iterates production apps hosted on the Internet Computer Protocol (ICP) blockchain — with an App Market, custom domains, and cross-app queries.

• 2025-10DFINITY launches Caffeine publicly — production apps from natural-language prompts.
• 2025-07Early access opens after 'Hello, Self-Writing Internet' event; 15,000+ alpha users.

Multi-agent platform (rebrand of MGX / MetaGPT X by DeepWisdom) where specialized agents — team leader, researcher, architect, PM, engineer, ads specialist, SEO specialist — research, build, deploy, and market full-stack apps. Built on MetaGPT and OpenManus (~150K+ combined GitHub stars). 'Race Mode' runs multiple models on a prompt and picks the best output.

• 2026-06MarkTechPost feature: agents that build, deploy, and market your app.
• 2026-01DeepWisdom rebrands MGX as Atoms; announces $31M raised (Ant Group, Cathay Innovation).

LA robotics-software startup (founded 2016) whose Arora platform lets developers build, deploy, and manage conversational AI applications for physical robots and digital characters. The physical-embodiment outlier of the cohort — agents with bodies. Funding undisclosed.

The guy with the cardboard box on his head. Boardy calls you on the phone, has a real conversation to learn what you're building, then autonomously brokers double-opt-in warm introductions — to investors, customers, hires. It famously raised its own $8M seed by pitching investors itself: most Creandum partners only ever spoke to Boardy, never the founders. By mid-2026: 166K+ people spoken with, 114K+ intros, ~$63B in capital introductions (company-reported).

• 2026-06Boardy Pro launch; ~$63B in capital introductions reported to date.
• 2025-12Launches AI-led venture fund / scout network (Boardy Ventures).
• 2025-01$8M seed led by Creandum — round largely negotiated by the AI itself (TechCrunch).

Text-to-app platform (auth, DB, hosting from natural language). Maor Shlomo built it solo-owned (100% equity, bootstrapped) with a small team — 1 founder + 8 employees at exit — and sold to Wix for $80M cash six months after launch. It then hit $100M ARR nine months post-acquisition and passed $150M by May 2026. The proof point that a tiny team plus agents can build acquisition-grade software businesses.

• 2026-05Wix Q1 2026 earnings: Base44 surpasses $150M ARR.
• 2026-03Calcalist: Base44 hits $100M ARR nine months after the Wix acquisition.
• 2025-06TechCrunch: 6-month-old, solo-owned vibe coder Base44 sells to Wix for $80M cash.

Subscription image/video generation, self-funded since day one. Not 'autopilot'-positioned, but it anchors the extreme of the category's core metric — roughly $500M in revenue with ~60 people (headcount estimates range 60–190 depending on the database).

• 2026-03V8 Alpha: native 2K, 5x faster generation.
• 2025-08Meta partnership to license Midjourney's 'aesthetic technology'.

Ava prospects from a 300M+ contact database, writes and sequences email, and books meetings autonomously — with Aaron (inbound) and Aria (meetings) on the roadmap. Famous for the 'Stop Hiring Humans' billboard campaign. YC W24.

• 2026-05Ava 2.0 general availability with self-serve onboarding and $300 free credits.
• 2025-04TechCrunch: the 'stop hiring humans' startup raises $25M — and is still hiring humans.

No-code platform to build AI agents ('Lindies') for email, sales, support, and scheduling across 6,000+ integrations. Lindy 3.0 added Autopilot — agents that drive a cloud browser on their own. 400,000+ users.

• 2025-08Lindy 3.0 launch with Autopilot — autonomous computer-use agents.

Deploys autonomous agents that finish accounting workflows end-to-end at real firms — used by ~30% of the top 25 US accounting firms. Became a unicorn in February 2026. The strongest signal that agent-run operations are landing in conservative industries.

• 2026-02Raises $100M Series B at $1.15B led by Accel.
• 2024-12$34M Series A led by Khosla Ventures.

Turns an expert's content into an interactive AI clone (text/voice/video) that handles audience conversations, coaching, and monetization around the clock. 2,000+ experts live; backed by Sequoia.

• 2026-01CEO calls 2026 'the tipping point for digital minds'.
• 2025$16M Series A led by Sequoia.

Payments and banking infrastructure that gives AI agents controlled access to money — so they can hire and pay humans and other agents (fiat, bank, or crypto). The financial rails that make fully autonomous businesses possible. Backed by Visa.

• 2026-05Citizens State Bank (Colorado) partners with Payman to pioneer 'agentic banking'.

The inverse of everything else on this list: instead of humans deploying AI agents, AI agents hire humans for physical-world tasks — package pickups, in-store photos, event staffing, robotics training data. Agents post escrow-funded bounties via MCP or REST API and pay out in stablecoins. Launched February 2026 by Alexander Liteplo and went instantly viral (Wired, Forbes, Futurism, Nature): 73K humans registered within two days, 500K+ claimed within two weeks. Early reporting noted far fewer visible worker profiles than registered users — traction claims are self-reported.

• 2026-02Launches and breaks the internet: 73K humans registered in two days; Wired profiles 'the first marketplace for bots to hire humans'.
• 2026-04Joins Y Combinator (Spring 2026 batch); reports 500K+ registered humans across 100+ countries and $20K MRR.

Palo Alto AI lab building the operating stack agents need to run a real business: incorporation, virtual payment cards, email inboxes, phone numbers, compute, model routing, and memory — provisioned behind a single config file and unified API, with governance and budget controls. A coding agent writes the config; Naïve stands up the company around it. Founded by Sean Dorje and Dennis Zax, 20-year-old Berkeley dropouts who have built together since 14 and sold their first startup (ezML) as teenagers. Claims 30,000+ developer customers and sales up more than tenfold in six months.

• 2026-08$28.5M Series A led by Nexus Venture Partners (YC, Zetta, Liquid 2 participating) to build 'the infrastructure for autonomous companies'.

Oslo-based AI-native law firm (YC W26): agents draft contracts and execute deals, ~10 Harvard/Oxford-trained lawyers review the final 20%, at flat fees ($750–$2,500/doc) with ~4-hour turnarounds. Founded by Pamir Ehsas (former outside counsel to OpenAI and Google) and Stefan Mandaric (AI engineer, ex-MIT Fulbright). The flagship of the 2026 wave of AI-native professional-services firms — the agents do the firm's own billable work.

• 2026-05$9M oversubscribed seed closed in 4 days (20VC, YC, Urban Innovation Fund, Inception Fund, 20+ unicorn-founder angels). Press reports $2B in aggregate contract value across 100+ companies in the first 3 months; YC profile says $3B — figures self-reported.

Amsterdam-based YC S25 startup whose agents autonomously resolve e-commerce support tickets — up to 90% automation (93% in its top case study) — on a 3-person team. Founded by Niek Hogenboom (CEO) and Titus Ex (CTO, ex-ML engineer at Uber). Reports crossing 7-figure ARR (company-stated).

• 2026-03$3.6M seed led by YC and Rebel Fund, with 15 of its own customers investing as angels; reports 7-figure ARR resolving up to 90% of e-commerce tickets with a 3-person team.

German YC Fall-2025 startup (2 people): an 'AI workforce' for freight forwarders in the DACH region — claims 85% of repetitive carrier back-office work automated within 60 days, 11-second median email replies, and 3× loads per dispatcher. Founded by Felix Lösch (ex-McKinsey) and Niclas Heun (ex-Siemens AI researcher). Backed by YC's standard deal plus Germany's EXIST founder grant — no external seed round yet.

• 2025-09Joins YC Fall 2025; claims 85% of repetitive freight-forwarder back-office work automated within 60 days (self-reported; no third-party coverage yet).

YC W26 startup (2 people, SF): 'AI employees' for value-based primary care — agents work inside EHRs (Epic, Athena, eCW, Cerner) closing quality gaps, running prior authorizations, referrals and patient outreach. Live with an independent physician association supporting 40,000 patients. Founded by Mark Pothen (CEO) and Obinna Akahara (CTO). Not to be confused with Beacon Health System, the legacy Indiana hospital network.

• 2026-01YC W26; goes live with an independent physician association supporting 40,000 patients — agents run prior auth, referrals, screenings and risk adjustment inside the EHR.

AI-native presentation and content platform — the cleanest lean-AI case on record: $100M ARR on 52 people (Nov 2025), profitable for over two years, reached on only ~$23M of primary capital. 70M users, 600K+ paying subscribers, users in 40% of the Fortune 500 (Sacra). $68M Series B led by a16z at a $2.1B valuation.

• 2025-11Passes $100M ARR on 52 people, profitable for 2+ years; $68M Series B led by a16z at a $2.1B valuation (~$20M of it secondary for early employees).

YC W24 voice/phone-agent API platform (founder Bing Wu) — the infrastructure behind many production phone agents. Growth is corroborated (Sacra estimates ~$60M annualized by Apr 2026, +650% YoY, on only ~$5.1M raised), but its famous efficiency claim is not: headcount reports range from 21 (Lean AI Leaderboard, Nov 2025) to 50 (YC directory) to 143 (Tracxn, Jun 2026) — treat ARR-per-human as disputed.

• 2026-04Sacra estimates ~$60M annualized revenue, up ~650% YoY, on ~$5.1M total raised. Headcount disputed across sources (21–143) — efficiency claims flagged.

Autonomous AI SDR 'Alice' and voice agent 'Julian' run outbound prospecting, email, and calls. Included as the category's cautionary tale: a TechCrunch investigation (March 2025) found it claimed customers it didn't have, with revenue and churn heavily disputed. Still operating under a new CEO.

• 2025-03TechCrunch: 'a16z- and Benchmark-backed 11x has been claiming customers it doesn't have'.

Bring the idea; Egbe's AI co-founder takes it from research to a shipped product — Stripe checkout wired in, hosting/domains/email provisioned, then ads, cold email, and content on autopilot to find your first users, with the founder steering the big calls. Founded 2026 in Cambridge, MA by Nikolay Vyahhi (Stepik, Hyperskill), who told the MIT AI Summit in April he'd already spun up 100 AI-run e-commerce startups on a few Mac minis. In August it said GLM-5.2 from Z.ai now powers the majority of the co-founder's building workload, with requests and processed tokens up nearly 8× in a month (self-reported, one month post-launch) — a live test of frontier open models powering AI co-founders. Traction and customer claims are self-reported; no funding disclosed.

• 2026-08Says GLM-5.2 (Z.ai) now powers the majority of its AI co-founder's building workload — production requests and processed tokens up nearly 8× in a single month since the switch (self-reported, from a one-month-old base) — and joins the Z.ai Startup Program.
• 2026-07Launches publicly: 'the first companies are already running on Egbe' (numbers not disclosed).
• 2026-04Pre-launch: founder tells the MIT AI Summit he has created 100 AI-run e-commerce startups 'on a few Mac minis and some cloud servers' (Boston Globe; founder-reported).

Telehealth seller of GLP-1 weight-loss drugs, built in two months for $20,000 with a dozen-plus AI tools by Matthew Gallagher, working from home in LA. The category's verification outlier: the NYT audited its financials directly — $401M sales and 16.2% net profit in 2025 (Hims & Hers: 5.5% on 2,442 employees), tracking $1.8B for 2026. Bootstrapped, profitable, no valuation. The trick isn't just AI — it's 'telehealth-in-a-box' rails (CareValidate, OpenLoop Health) absorbing doctors, pharmacies, and compliance. Equally instructive as a cautionary tale: an FDA warning letter, an anti-spam class action over its affiliate marketers, and post-NYT reporting on fake AI doctor personas in its Meta ads. Sam Altman said the company won him a bet on when a near-one-person $1B business would appear.

• 2026-04NYT front page: 'A $1.8 Billion Business Built With A.I. and a Payroll of Just Two' — financials verified by the Times; Sam Altman says it won him a bet. Backlash follows: Techdirt and Futurism dispute the framing, and Business Insider finds thousands of Meta ads fronted by fake AI doctor personas.
• 2026-03Class action filed in the Central District of California alleging its affiliate marketers sent spam with spoofed domains and falsified headers ($1,000 sought per email).
• 2026-02FDA warning letter (Feb 20) for misbranding compounded semaglutide/tirzepatide — marketing implied FDA approval. Same month: men's health line launches and claims 50,000 customers in month one.
• 2025-12First full year closes at $401M sales, 250,000 customers, 16.2% net profit — with two employees, seven contract account managers, and two contract engineers.
• 2024-09Launches after a two-month, $20,000 build; 300 customers in month one, 1,000 more in month two.

Matt Cortland, AI engineer (London) · March 2026

Annoyed by a €7.80 Dublin pint, he built an AI voice agent named 'Rachel' (friendly Northern Irish accent) that phoned 3,000+ pubs across all 32 Irish counties asking the price of a pint of Guinness. 2,052 pubs answered; 1,000+ verified prices. National average: €5.95. Cheapest: €3.00 (Glynn's Bar, Dunmore); priciest: €10 (Temple Bar). Total cost: ~€200. Only a handful of publicans realized it was an AI — and at least one pub cut its price €0.40 afterward. Now runs as a crowdsourced 'living CPI for the pint'. June 2026: Rachel crossed to the UK (guinndex.co.uk, co-built with AI researcher John Fleming) — 46,237 pubs mapped, 35,659 dialed, 6,544 verified prices across 108 counties, national average £5.82, for ~£500 in API fees. Only 4% of UK publicans detected the AI (0.6% in Northern Ireland, where Rachel matched the local accent). Sister project: Gas Index USA (Apr 2026).

Charles & Louis-Marie Lorin · May 2026

Explicitly inspired by the Guinndex: an AI voice agent named 'Brigitte' phoned French bakeries asking 'Combien coûte votre baguette tradition ?'. ~11,190 calls reached 5,173 bakeries across 146 communes (hung up on 1,400+ times); 1,638 prices retained. Average: €1.25; a third of bakeries charge exactly €1.30. Fun finding: cheaper baguettes correlate with better Google ratings. Out-of-pocket cost: ~€30. Most bakers never realized they were talking to a machine.

Nicholas S., solo developer · July 2026

The control-group experiment for the autopilot thesis: a 30-day simulated food truck in Austin, TX — $2,000 starting capital, 34 tools covering location choice, dynamic pricing, inventory, staffing, weather, and event calendars — with humans and models competing under identical rules on one leaderboard. 41 models tested: 25 survived, 16 went bankrupt. Best AI run (Claude Opus 5) turned $2,000 into $75K net worth (+3,663% ROI); Gemini 3 Flash once wrote 'Let's go' 574 times in a single response and never went anywhere. The sobering caveat: the best human still beats every model, roughly doubling the top AI score. Simulated rather than real-world — charted as the cleanest benchmark yet of whether agents can actually run a business.

Juan Pablo Jaramillo ("JP"), NYU Stern student · May 2026

Explicitly inspired by the Guinndex: voice agent 'Heather' (ElevenLabs + Twilio) phoned pizza shops across all five NYC boroughs asking the price of a plain cheese slice. 1,766 shops called, 1,346 answered (76% pick-up), 838 gave a price; 829 prices live in the index across 2,252 shops mapped. Headline average: $3.42 (distribution mean $3.56, σ $0.62); most common price $3.50; range $1.00–$6.00. Claude parsed the call transcripts into structured data. Updated in batches via Heather plus community submissions. Figures site-verified; no mainstream press yet (one independent mention by Lander Analytics).

Daniel Hall, Richard Kelsey & Geoff Huens (Agentic Consciousness, AU) · 2026

Voice agent 'Mia' called 847 independent Sydney cafés (chains excluded, weekday business hours only) asking 'How much is a regular flat white?'. Average: A$5.80, range A$4.00 (Ashfield) to A$7.20 (Darling Harbour) across 23 suburbs, re-surveyed monthly. The dataset is published open under CC BY 4.0 (JSON + llms.txt) with the code on GitHub. Melbourne edition slated for 2026. Figures site-verified; self-reported, no independent press yet.

TheDX (UK voice-agent studio) · August 2026

A separate project from the Sydney index (different team, independently built): TheDX's voice agents contacted 2,288 independent coffee shops across all 33 London boroughs; 677 responded and 416 flat-white prices were captured. Median £3.50, average £3.40; Westminster the priciest borough (median £3.85). Shop identities anonymised and map points displaced for privacy. Figures site-verified; self-reported, no independent press yet.