Sign in Subscribe
Chapter 08

Your Org Chart Is a Liability * AI Takeoff and the Arrival of the Company as Computer

Two employees, $20.000 in startup capital, $1,8 billion in projected revenue. Medvi did not beat its competitor by building a better team. It beat them by not needing one. The Company as Computer has arrived, and your org chart is now your biggest liability.

Peter van Hees

• 6 min read
Your Org Chart Is a Liability * AI Takeoff and the Arrival of the Company as Computer

Two years ago, Sam Altman reportedly entertained a provocative bet: the arrival of the first one-person billion-dollar company. At the time, it sounded like Silicon Valley theatre, part prediction, part provocation. But bets like that matter because they reveal when people closest to the frontier have already seen a shift that the rest of the market still mistakes for hype.

Now that architecture has a visible shape. Two employees, $ 20.000 in startup capital, and $ 1,8 billion in projected 2026 revenue.[1] Medvi did not beat its competitor by building a better team. It beat Hims and its thousands of employees by not needing them.

Forget the founder story. What matters is not the entrepreneur. What matters is that the economics of the firm have changed, and most org charts have not.

AI Takeoff is not the moment AI gets impressive. It is the moment the old logic of headcount stops making economic sense.

The Frame * What 'AI Takeoff' Actually Means

AI Takeoff is the phase in which AI stops acting as a productivity tool inside the firm and starts acting as a production system that changes what kind of firm is possible. Capability improves fast, cost falls fast, and AI begins helping build, optimise, and operate more AI. The result is not just smarter software. The result is a new economic threshold: firms can generate meaningful scale, speed, and leverage with radically fewer people than the managerial systems of the last era assumed.

The Indictment * Medvi as Exhibit A

Matthew Gallagher built Medvi, a GLP-1 telehealth company, in two months.[1] In simple terms, Medvi built a digital front end for people seeking weight-loss medication, connecting demand, clinical review, and fulfilment through a largely AI-orchestrated operating model.

He used 12 AI tools to dissolve most business functions into software, leaving humans concentrated mainly in brand strategy and customer acquisition. In its first full year, the company generated $ 401 million in sales with 250.000 customers.[1] Net profit margin: 16,2%. His competitor Hims, with thousands of employees, managed 5,5%.[1]

Gallagher's own description is revealing: "It's not an A.I. company, but I did it with A.I."[1] He architected a system, not a technology company. He orchestrated agents, not employees.

This is what I have described as the Company as Computer: the idea that an organisation can be architected as a portfolio of automated functions orchestrated by a lean human layer, with little legacy overhead required. I explore this logic in more depth in my book AI Agents: They Act, You Orchestrate.

The mechanism is the Functional Dissolution Principle: any business function whose operations can be defined by rules and whose outcomes can be verified is a candidate for full automation. Gallagher applied this principle to an entire company. Marketing, customer service, operations, compliance monitoring: dissolved into automated functions, orchestrated by two people from a house in Los Angeles.

Medvi offers an early glimpse of what the company of the future may look like: a dynamic portfolio of automated functions orchestrated by a lean cadre of human strategists. That company generates more than $ 3 million per day in revenue.[1]

The Pattern * Revenue per Employee is the new Metric

Medvi is the sharpest illustration, but the pattern is structural. AI-Native startups consistently generate $ 2 million to $ 4 million in revenue per employee. The average public SaaS company employee only generates $ 300.000.[2] That is a 7x to 13x efficiency gap, and it is not narrowing.

The data is consistent across sectors and business models. Cursor reached $ 1 billion in annual recurring revenue with approximately 300 employees, producing $ 3,3 million per employee.[3] Lovable hit $ 400 million in annual recurring revenue with 146 employees, or $ 2,7 million per employee.[2] These are not isolated anomalies. They are firms that built every process around AI from day one, carrying zero legacy organizational overhead.

Revenue per employee is becoming one of the clearest metrics of competitive fitness in the Agent-First Era. Every company below the AI-Native threshold carries structural overhead that compounds against it every quarter. You are not just less efficient. You are subsidizing your competitor's cost advantage with every payroll cycle you run.

The Engine * Why the Gap Accelerates

The capability-per-dollar of AI is compounding on two axes simultaneously. Frontier training compute grows approximately 4,5x per year.[4] Pre-training compute efficiency doubles roughly every 7,6 months.[5] These are separate forces: more raw compute deployed, and better algorithmic efficiency extracting more from each unit of compute. They compound on each other.

For business leaders, this matters for one reason: when capability rises and cost falls at the same time, the minimum efficient size of a company starts collapsing.

Greg Brockman, president of OpenAI, named this phase: "We are in this early phase of takeoff of this technology. Takeoff is [when] the AI gets better and better on this exponential curve."[6] When AI improves the tools used to build AI, the pace of capability growth becomes non-linear. The S-curve Brockman describes is recursive acceleration, where better models produce better training methods that produce better models.

The revenue-per-employee gap is a compounding divergence, not a stable difference. Every month you delay restructuring, the cost advantage of your AI-native competitor grows. The tools available to a two-person team next quarter will be more capable than what Gallagher had when he built Medvi. The gap does not plateau. It steepens.

The Counterargument * Why Dismissing Medvi is the Wrong Move

The objections are predictable. Medvi is a reseller. CareValidate and OpenLoop supply the doctors, pharmacies, and compliance infrastructure. Gallagher built a marketing front end on top of human-staffed services. The FDA issued a warning letter in February 2026 for unsubstantiated health claims.[1] And for every Medvi, thousands of AI-powered ventures fail.

These criticisms are all valid. They all miss the structural argument.

Gallagher did not need to own the supply chain because AI allowed him to orchestrate it. The Functional Dissolution Principle applies to his firm's functions, not his suppliers' functions. Pointing out that CareValidate employs hundreds of people does not invalidate the thesis any more than Amazon’s warehouses invalidate the idea that e-commerce changed retail economics.

The survivorship bias objection is the weakest. Dismissing the pattern because one example has vulnerabilities is the strategic equivalent of dismissing e-commerce in 1999 because Pets.com failed. The structural data exists independent of any single company. AI-Native firms across every sector, from developer tools to design platforms, are producing the same 7x to 13x revenue-per-employee ratios.[2] Nearly a third of small and medium-sized enterprises globally now use generative AI.[7] The Functional Dissolution Principle is diffusing across the entire economy.

The Diagnosis * The Org Chart Must Now Defend Itself

I have been prosecuting the org chart by pointing at AI-Native competitors. The twist is closer to home. The Functional Dissolution Principle doubles as a diagnostic you must apply to your own company, function by function.

Any operation definable by rules and verifiable by outcomes is already a candidate for dissolution. Customer support workflows, financial reporting, routine legal review, content scheduling, data pipeline management: these functions are automated elsewhere, today, by firms paying a fraction of what you pay in salaries and coordination overhead.

Stop asking whether AI will replace your company. Start asking which parts of your company already have no economic reason to remain human-run.

The company is no longer best understood as a hierarchy of roles. It is becoming a portfolio of functions, some human, many machine, all orchestrated for outcomes. The firm has already been reinvented. Your org chart just hasn’t admitted it yet.

This article introduces the Company as Computer thesis and the Functional Dissolution Principle from AI Agents: They Act, You Orchestrate by Peter van Hees. The book maps 18 chapters across the full architecture of the Agent-First Era, from the Autonomy Spectrum that separates real agents from glorified assistants, to the Silicon Salary Model for budgeting your synthetic workforce, to the Human Premium Stack that defines where durable human value survives. If the economics in this article rattled your assumptions about headcount, the book gives you the complete operating manual. Get your copy:

🇺🇸 Amazon.com 🇬🇧 Amazon.co.uk 🇫🇷 Amazon.fr 🇩🇪 Amazon.de 🇳🇱 Amazon.nl 🇧🇪 Amazon.com.be

References

Sign up for more like this.

Enter your email
Subscribe