Where the headcount still earns its salary

A study that only argues one side is propaganda, so here is the honest counter-case. The lean stack is not a universal answer, and three situations genuinely justify a hire over a tool.

Deep specialist craft that an agent cannot yet judge. Brand and visual design at the top end, a senior performance buyer placing six-figure budgets, an enterprise sales closer reading a room — these are judgement-dense roles where the cost of getting it wrong dwarfs the salary. A lean operator can get to good with tools; world-class in a single discipline still tends to need a human who has done it for a decade.

Throughput that genuinely exceeds one orchestrator. A founder plus agents has range but finite hours. If a channel is working so hard that it needs eight hours a day of hands-on execution by itself, that is a real capacity ceiling, and a hire (or a contractor) buys you the hours an agent cannot conjure.

Relationship-owned revenue. Partnerships, key accounts, and high-touch customer success run on trust that accrues to a person. An agent can prepare, draft, and remind, but the relationship itself is not a thing you can put on a free tier.

The point is not that headcount is obsolete — it is that headcount should be the deliberate, expensive choice you make at a specific ceiling, not the default you reach for because operating the tools used to require it. Hire when you have proven the function works and the only constraint left is human hours or human judgement. Until then, the tools and the operating model carry it. The same proof-before-spend logic that governs the free-tier dividend applies to people: pay on proof, not on plan.