Teams became optional

For a hundred years there was exactly one way to do more, and it was to hire someone to do it. That is the whole story of building a company until about the year before last. You could not clone yourself, so you paid other people to be the extra hours. Capacity was the constraint, hiring was the release valve, and the number of people who worked for you became the readout everyone trusted, because it was the one number that actually moved when the business grew.

So headcount stopped being a cost and started being a scoreboard. How many people work there is still the first thing anyone asks about a company, before what it makes or whether it makes money. Forty people is a real business. Four is a hobby with a logo. You raised on the headcount, you got written up on the headcount, and somewhere in there you started hiring to hit the number instead of to get the work done, which nobody admits, because the number was the thing that looked like winning.

A proxy works right up until the thing it stood for comes loose from it. Headcount stood for output. It was a good proxy for a century because the link was mechanical: more hands, more done, no other way to buy the "more." Then the tools got free and competent in the same eighteen months, capacity fell off a cliff, and the link snapped. The number that used to track output now tracks two things only, cost and coordination load, both of which go the wrong way. A large team no longer tells you a company does a lot. It tells you a company pays a lot and spends its Tuesdays in meetings about the meetings.

The evidence that headcount and output came apart is older than the AI panic, which is the useful part, because it means this was never an AI story to begin with. WhatsApp had somewhere around fifty-five people when Facebook bought it for roughly nineteen billion dollars, and those fifty-five were serving something like four hundred and fifty million users. Instagram had thirteen when it sold for about a billion. Craigslist ran a chunk of the planet's classifieds on a staff you could seat at one long table. None of that was AI. All of it was the link breaking on its own, years early, in a handful of companies that happened to sit on software's zero-marginal-cost trick before the trick was handed to everyone. The tools did not invent the decoupling. The tools mailed it to the rest of us.

The trap is a live one, because the instinct outlived the mechanism it was built on. Hiring still feels like progress after it stopped being progress. You add a person and it registers as the company getting bigger, more real, more serious, the same warm hit it gave in 1990, except now the person is often doing work a tool would do for free and doing it slower, and you are paying salary plus the overhead that rides on salary to feel a feeling. That is the hire that adds a headcount and subtracts your focus, made because the org chart wanted to grow, not because a job needed a mind. The proxy is still whispering, and the proxy is wrong.

So the question stopped being how many people. It became what capability is missing, and whether a hire is the only way to get it. Capacity is free now, which means capacity is never the reason to hire, ever again. Delete it from the list. What survives are the things the tools cannot supply, and the honest list is short. Parallelism, when the work needs a body in two places at once and you have one body. Independent error-correction, a second mind with its own stakes that catches what yours misses, which is not a model you prompt to disagree with you (see why the solo operator has no team to average out a bad bet). Human-held trust, the enterprise buyer signing a seven-figure contract who wants a person accountable and not a founder and a stack. And durable throughput a person owns end to end, when the work truly exceeds what one life can hold. Four things. Everything the old instinct files under "hire" that is not on that list is now a tool, and calling it a hire is paying salary for something free.

The strongest version of the other side, and I will not shave it down: teams do things one person structurally cannot, and no amount of free tooling closes the gap. You cannot sit in two customer meetings at the same hour. You cannot hold ten enterprise relationships at the depth each one demands. You cannot cover a support desk around the clock out of one body that needs to sleep. A team disagrees back with its own stakes in a way a compliant model never will. And a buyer moving real money wants a human on the hook, not a solo operator gamely insisting the automation has it. "Optional," on this reading, is a founder's cope for being unable to afford or manage people, and it caps you at exactly the ceiling teams were invented to break through.

Every word of that is true, and every example in it is a capability, which is the whole point rather than a refutation of it. Read the counter's own list back: parallelism, independent judgment, human-held trust, throughput one life cannot sustain. Those are the four. The counter did not attack the thesis, it filled in the test for it. The claim was never "never hire" and never "teams do nothing." The claim is that the burden of proof flipped. The old world hired by default and needed a reason not to. The new world stays lean by default and needs a reason to, and the reason has to be a capability a person supplies and a tool cannot. Notice what the counter never once reached for: more hands, more output, more capacity. It couldn't, because the tools retired those, so even the best case against optional is built entirely from the four things that make a hire defensible. If your reason is on the list, hire, guilt-free. If your reason is "this is what growing companies do," you are paying salary to a proxy.

Which is why the capability test runs before any hire, and it is four questions.

Name the missing capability, not the missing capacity. One line: what specifically cannot get done, and why. If the honest answer is "there is too much of it for me," that is volume, and volume is what the tools are for. Stop, go automate. Only keep going if the gap is a kind of work, not an amount of it.

Run the tool test. Could a competent person with today's tools do this without the hire? Try it for a week before you post the job. Most of "I need to hire a marketer, an analyst, an ops person" is "I have not pointed the tools at this yet." If the tools close the gap, the gap was capacity, not a role you needed to fill.

Check it against the four survivors. A hire earns its cost only if the capability is one tools cannot give: you have to be in two places at once, or you need a second mind with its own stakes, or you need a human the market will trust and hold accountable, or the work truly exceeds one life and someone has to own it end to end. On the list, hire. Off the list, it is a tool.

Price the coordination cost you are buying. A hire is not salary. It is salary at roughly one-and-a-quarter to one-and-a-half times once you count benefits and overhead, plus the months before they are up to speed, plus the standups and reviews and the way your own day changes. Brooks worked out the ugly version of this fifty years ago: communication channels grow faster than people, so adding a body to a late job can make it later, not sooner. If the capability is real but the coordination cost eats the gain, the answer is a contractor, an advisor, a bounded engagement, not a permanent seat. Buy the capability at the smallest commitment that supplies it.

Do this, and it costs you one honest afternoon. Name your last three hires, or the three you are about to make, and for each one write the single capability the tools could not have supplied. Not the job title. The capability. If you can name it for all three, your instinct is calibrated and you can trust it. If you get to one where the truest sentence is "that is what a growing company does," you did not find a capability. You found the proxy, still whispering, and you were about to pay it a salary.