Our take: headcount is the new technical debt — every hire is a loan against your future velocity, and the interest is coordination. For a generation of operators, team size was the scoreboard: more reports meant more importance, a bigger department meant a bigger mandate. AI broke that equation from both ends at once. On one end, a small senior team with real AI leverage now produces what used to take a department. On the other, the coordination cost of a large team didn't go anywhere — it still compounds with every hire, every layer, every sync meeting. The result is a market where the org that hires by default is out-shipped by the org that builds by default, and where boards have quietly changed which number they read first.
The coordination tax, in plain math
This isn't a study you have to trust — it's arithmetic you can check. Communication paths in a team of n people grow as n(n-1)/2. Salaries scale linearly; coordination scales quadratically. That gap is the tax, and it gets paid in meetings, handoffs, alignment docs and decisions that die in transit.
| Team size | Communication paths | What coordination looks like | What actually ships |
|---|---|---|---|
| 3 | 3 | A conversation | Nearly everything the team produces |
| 5 | 10 | A standup and a channel | Most of it — overhead is real but small |
| 10 | 45 | Standups, syncs, a planning ritual | Good weeks; process starts eating time |
| 30 | 435 | Program managers, alignment meetings, roadmap reviews | A fraction — much output is coordination about output |
| 100 | 4,950 | Layers whose full-time job is other layers | Individually heroic, collectively slow |
What AI flipped
The quadratic tax always existed — Brooks wrote it down in 1975. What changed is the other side of the trade. Hiring used to be the only way to add capacity, so the tax was worth paying. AI ended that monopoly:
- A senior builder with AI leverage covers ground that used to require a hire — so the marginal hire competes with a system that costs a fraction and adds zero communication paths.
- Systems don't need onboarding, don't attend syncs, don't misremember decisions, and don't leave with the context.
- The honest pre-hire question changed from 'who do we hire?' to 'what do we build?' — and 'hire' has to win that comparison on merit now, not by default.
- Revenue per employee — long a footnote — became the cleanest single readout of whether a company runs on leverage or on payroll. Boards noticed.
- None of this means never hire. It means every hire should clear a higher bar: does this person bring judgment and leverage a system can't, or just capacity a system already can?
How to operate on this
- 1Before every requisition, force the alternative: what system, automation or AI workflow would make this hire unnecessary? Write it down. Compare honestly.
- 2Track revenue per employee quarterly and treat a decline the way you'd treat rising churn — as a structural warning, not a footnote.
- 3Structure work in pods of 3-5 senior people with full ownership of an outcome, and give them real AI leverage instead of junior support layers.
- 4When a team feels slow, resist the reflex to add people — adding people to a coordination problem is pouring interest onto debt.
- 5Pay down existing headcount debt the way you'd pay down code debt: consolidate, automate the coordination layer, and redeploy the best people into building.
