AI VS STAFF COST 2030
AI vs staff cost: AI beats the payslip in 50% of roles.
Key figures
Share of jobs where AI is already cheaper than the human doing them: 23% in 2026, 50% by 2030 on my call, humans still cheaper in the rest.
Cost per digital task: a human worker about $25, an AI agent $1 to $2.40; per customer service contact: $3 to $6 human, $0.25 to $0.50 AI; a mid-level employee $120,000 to $150,000 a year fully loaded, an AI employee $3,000 to $6,000, payback 6.7 months.
The token bill: a 4-person AI startup paid $113,000 for 1 month of tokens in March 2026, double the month before, $340,000 a year per head, 5 times a $65,000 salary, not the 21 times the viral chart claims.
The other side of the ledger: IBM replaced hundreds of HR staff with AI and hired more people in sales and engineering; 95% of enterprise AI pilots returned nothing.
AI is cheaper than staff in 23% of jobs today. I forecast 50% by 2030. That is the whole card, and both halves matter: the sellers only show you the first, the skeptics only the second.
The sellers' math: a human charges about $25 a task on digital work, an AI agent $1 to $2.40. A customer service contact costs $3 to $6 with a person, $0.25 to $0.50 with a bot. A mid-level employee costs $120,000 to $150,000 a year fully loaded, an AI employee $3,000 to $6,000. Payback on a scoped deployment: 6.7 months.
The tape's math: MIT ran the sums on every role in 2026. AI wins on cost in 23% of them. In the other 77%, the human is still cheaper, because judgment, rework and review get added back.
And tokens are the new payroll. Swan AI, 4 people, paid Anthropic $113,000 for 1 month in March 2026, double the month before, and the CEO called it his proudest invoice. The viral chart says 21 times a salary. The math says 5 times per head, for a company whose agents earn its whole revenue.
Jensen Huang wants a $500,000 employee to burn $250,000 in tokens. Chamath Palihapitiya: "my costs go up 3 times every 3 months, my revenues do not." 95% of enterprise AI pilots returned nothing.
The honest line is not "AI replaces staff". It is "AI replaces the routine 30% of every job first", my call from 15 years ago. IBM did exactly that: replaced hundreds of HR staff with AI, hired more in sales and engineering. The consulting version is my Nº 40 card, the analyst pyramid.
Who wins: the company that scopes the agent, meters the tokens, and redeploys the human. Who loses: the CFO who buys "95% cheaper" from a vendor deck and gets the invoice.
