Millions of Profitable Autonomous Companies Within a Year, Argues Founder
6 min read · 14 sources
- An essay claims millions of profitable fully autonomous companies will exist within a year, reversing its author's stance from six months ago
- AI-native services target the $4.6T services market, with AI law firm Crosby charging $250-$1000 per contract
- Anthropic committed $100M in October to train 10,000 forward deployed engineers by end of 2027
- OpenAI is rolling out visual ads in ChatGPT, shown to some users during image generation
- Pave data from 10,000+ companies shows sales is 18.4% of AI startup headcount versus 20.7% in traditional tech
The pitch is no longer that AI will augment your company. It’s that AI will be the company. A new essay from The General Intelligence Company argues that within a year there will be millions of profitable autonomous companies - organizations where every management decision is made by agents, with humans only at the very bottom of the hierarchy for specific tasks. That’s a reversal from the author’s position six months ago, when he argued human oversight was still necessary.
Whether or not you buy the timeline, the essay crystallizes what’s actually changing in how software and organizations get architected. The coordination layer - collective purpose, management structure, internal communication, state - is becoming a software problem. That matters even if you think “millions” is fantasy, because the infrastructure being built to get there is real and shipping today.
GC AI’s Cecilia Ziniti sold the first $16M with herself and three reps, closing an $80,000 deal in 21 days because she didn’t know it was supposed to be impossible.
The $4.6 Trillion Services Market Is the New SaaS TAM
The AI-native services model targets money that software never touched. Companies like Crosby, an AI law firm, charge $250 - $1,000 per contract review. Harper, an AI insurance brokerage, handles 1,000+ customers a month versus 20 - 30 for a human team. The prize is the $4.6T enterprises spend on salaries and services, not the $200B SaaS market.
The hard part isn’t getting off the ground - it’s scaling without proportionally more people. Revenue comes immediately, but genuine product-market fit is unproven. The model only works if delivery quality holds as volume grows, and that’s a distributed systems problem wearing a business-model costume. If you’re building one of these, the question isn’t whether the first 100 customers are happy; it’s whether customer 10,000 gets the same output as customer 1.
You're Hiring Too Many Reps
The math from Euclid’s Verticalist Summit is brutal and simple: divide each closed deal’s annual contract value by the human hours spent winning it, including sales engineering and legal. If that number falls as you add reps, fix pricing and packaging before hiring more. The panel’s examples include $15,000 - $20,000 annual contracts closing in six to seven days with a couple of calls and no solutions engineer.
The standout case is GC AI’s Cecilia Ziniti, who sold the first $16M with herself and three reps - closing an $80,000 deal in 21 days because she didn’t know it was supposed to be impossible. The lesson isn’t that sales is dead; it’s that the unit economics of human-led sales have changed when AI does the discovery and the first-draft work.
The Forward Deployed Engineer Is a GTM Function Now
Palantir made the FDE famous; Anthropic is industrializing it. In October, Anthropic committed $100M to train 10,000 forward deployed engineers by end of 2027. The role sits with sales, maps manual work, builds workflows in the client’s CRM, trains reps, and hands over. Success is scored on manual hours saved, team usage, workflows live, and self-sufficiency.
This is the model for selling AI-driven GTM systems: the engineer is the product’s first power user and its best demo. For engineers, it means the boundary between “writing code” and “closing deals” is dissolving. The five-stage process - sit, map, build, train, hand over - is a repeatable playbook, and the $100M bet says Anthropic believes it’s the distribution channel that matters.
Claude Gets Spreadsheets, Whiteboards, and a 9,000-Connector Marketplace
Anthropic shipped three new artifact types: Claude Sheets for collaborative spreadsheets, Claude Whiteboards for meetings, and Claude Data (formerly Dashboards) with Snowflake and Salesforce connectors. Interactive connectors now let partners embed apps like Adobe and Atlassian directly inside Claude, with Salesforce, NetSuite, and Workday plugins in beta. The Claude Marketplace lists over 9,000 connectors and plugins built on MCP.
The point for operators: context-switching is the tax you pay for a fragmented toolchain, and these artifacts are designed to collect it. Whiteboard discussion to spreadsheet to dashboard without repeating context is the kind of workflow consolidation that actually saves hours, not minutes.
OpenAI Turns on Visual Ads and Launches Business Hub
Source: searchengineland.com ↗
Two moves from OpenAI today. First, visual ads are rolling out in ChatGPT - some users see ads during image generation, and they disappear once the images are done. Second, Businesses in ChatGPT lets businesses claim their profiles through Business Hub and see engagement insights. Business Agents allow eligible US e-commerce and online service businesses to interact with users directly. Conversations and Business Agent chats are not shared with businesses, which is the privacy line that makes the whole thing palatable.
Cursor Projects Goes Multiplayer
Cursor’s Projects feature now lets teams run a fleet of coding agents through a shared coordinator. A task board shows who and what agents are working on, PRs link to tasks, and a unified context system holds requirements docs, specs, and diagrams accessible to all agents. Agent-to-agent communication lets coordinators plan around each other in real time - on-call handoffs become seamless.
This is the first workspace designed for team collaboration with multi-agent coding. The shared task board turns agent work into something a human manager can actually supervise, which is the missing piece for trusting agents with production work.
The Sales Team Myth, Crunched
The narrative that AI startups run lean sales teams doesn’t survive contact with Pave’s data from 10,000+ companies. Sales is 18.4% of headcount at AI hypergrowth companies versus 20.7% in the rest of tech - a gap of just 2.3 points. The real divergence is elsewhere: customer support is 1.8% versus 4.4%, while engineering is 47% versus 33.9%. ICONIQ’s data showing 43% fewer GTM FTEs at $10M - $25M ARR is because AI-forward companies are smaller overall, not because sales is being cut.
The takeaway for org design: AI is reshaping engineering and support roles far more than sales. The sales playbook hasn’t changed as much as the hype suggests, and large deals still need humans to navigate procurement.
Designing for the Agent That Buys Your Product
Resend’s Agent Experience work is the new SEO. In a study, Claude Code picked Resend first in 64 of 102 email recommendations, beating SendGrid (which came first 7 times but appeared 55 times as an alternative). Resend has 4.3 million users and 120,000 paying customers, and its MCP server grew from ~20 to over 80 tools. The trick: adding .md versions of pricing pages so agents get prices as a Markdown table in one request.
Revolut's Uncompromising Scale
Source: cheekypint.substack.com ↗
Nik Storonsky runs a bank that adds 100,000 customers per day and hit 80 million total, with zero growth budget until 2022. He’s now targeting Amex Platinum and Chase Sapphire in the US, uses strict retention curves for new bets, and is cutting 210-day compliance approval cycles with AI. For fintech engineers, the lesson is the “new bets” framework: build, measure against a retention curve, kill fast.
Networks for Agents, and the Great Product Fallacy
Two quick ones. a16z’s piece on agent networks sketches Uber-like marketplaces where agents negotiate on price and availability - a restaurant offering early-seating deals to flexible groups. Building these requires new infrastructure for machine-to-machine negotiation. And Jason Cohen’s counterpoint to the “product beats sales” crowd: a company that fired its entire sales team dropped from $5M to $1M in revenue. Great products make great sales teams more valuable, not unnecessary.
You May Also Like
OpenAI Severs Cursor Ties, Self-Improving AI Runs the Lab, and Power Can't Keep Up
OpenAI is ending its contract with Cursor on November 12 following SpaceX's acquisition, citing contract violations by Elon Musk's companies. Anthropic …
AI compute costs explode, Anthropic loses $567 M; Zapier self‑heals, Notion pivots to agents
GPU rental rates have surged to $12‑$13/hr, pushing Anthropic’s Q3 compute bill to $5 bn and flipping a $974 m profit into a $567 m loss. The top 1% of AI …
Killing the Per-Seat License, Vercel Virtual Models, and Anthropic's One-Month Sprints
The per-seat SaaS model is breaking under AI automation, with hybrid pricing surging from 25% to 37% over the past year according to Kyle Poyar's 2026 data. …




