← All episodes 
Episode
Revenue per employee: the X-ray that shows who's really using AI
Anthropic pulling in 9 million dollars per employee. Nvidia at 5 million. Microsoft at 1.2. Caio and Marina break down the metric that separates the companies that actually changed how they operate from the ones that just slapped AI on a slide — and why the team of the future might be hiding in your token bill.
In this episode
01 The hook: the question that changes everything
- Caio opens with the provocation: the question is no longer how many people you have, it's how much revenue each person generates with AI alongside them
- Marina pushes back like 'wait, weren't valuation and user count what mattered?' so Caio can say there's a more honest metric showing up
- Drop the shocking number right away: Anthropic estimated at around 9 million dollars in revenue per employee, according to Epoch AI. Make clear it's an estimate, private company
- Marina does the mental math out loud to give it scale, and the hook becomes: how is this even possible?
02 What this is and why it matters again now
- Caio explains it simply: total revenue divided by number of employees. Marina asks 'that's it?' and Caio cuts in saying yes, but what it hides is the gold
- Make clear what the metric does NOT measure: not profit, not culture, not quality. It measures operational leverage, how much a company does with few people
- The historical shift: in the cheap-money era, growing meant hiring. More revenue, more people, more layers, more meetings. AI broke that logic
- Marina brings the investor's question: why put 100 people on it if 20 with good agents deliver almost the same thing?
03 The ranking that gives away the game
- Caio lays it out flat: Anthropic around 9 million, OpenAI around 5.5, Nvidia 5.1, Meta 2.5, Google 2.1, Microsoft 1.2 per employee. The scale is brutal
- Marina notes the contrast: Microsoft has 228 thousand people and sits at the bottom. Anthropic with a few thousand is at the top. Small body, giant economic force
- Nvidia's special case: it doesn't use AI, it sells the infrastructure. Scarce product, high margin, global demand. A different kind of leverage
- Caio nails it: revenue per employee has become the proof of leverage in the AI era
04 The augmented employee (and what changes in practice)
- Caio cuts the fear: the point isn't AI replacing everyone. It's that each person starts operating with way more capacity, surrounded by agents and copilots
- Concrete examples: an analyst doing research, the report, the presentation and automation alone. A dev prototyping, reviewing and testing faster. Marketing doing the brief, copy, landing and campaign with far less dependence
- Heavy data point: Google says 75% of new internal code is already AI-generated and human-reviewed. Marina reacts and Caio explains it changes the unit of measure — it's no longer how many people write code
- Marina pulls the uncomfortable angle: Amazon, with Andy Jassy saying in an official memo that they'll need fewer people in certain roles thanks to efficiency gains from agents
05 The twist: the team disappears from the org chart and shows up in the bill
- Caio drops the Mercor case: the CEO said the startup already spends more on internal agent tokens than on salaries. Recruiting, accounting, fraud, interviews, all with AI
- Marina has the realization: so the company of the future doesn't spend less, it swaps human cost for compute cost. The operational team isn't in the org chart, it's in the API bill
- Caio puts the risk on the table: you can inflate the metric in a dumb way, cutting too many people, hurting service, burning out the team, creating dependence on fragile automation
- The right question isn't just how much each person generates, it's whether that revenue is sustainable, safe and scalable. Marina agrees that otherwise it's just a vanity metric
06 Closing: the most honest metric of the AI era
- Caio sums up the thesis: the leaders won't be the biggest by headcount, but the ones who create more value per person, with smaller, more technical, more leveraged teams
- The practical takeaway for the founder or executive listening: take your company's revenue, divide by headcount, and ask if that number is climbing or you're just hiring to grow
- The line that sticks: revenue per employee reveals who used AI to change how they operate and who just put AI on a slide
- Caio closes with a gentle invite to apply this to your own company before comparing yourself to the giants, and teases the next episode on how to redesign processes to truly get leverage