The Thermometer Doesn’t Drive the Car

Accounting, Legal, HR, and Sales should never run the company. Anyone from those groups who disagrees should be trusted with decision-making authority even less.

Accounting, Legal, HR, and Sales should never run the company. Anyone from those groups who disagrees with that should be trusted with decision-making authority even less.

That’s the claim. I’m going to take the pushback seriously before dismantling it, because the pushback isn’t stupid. The people making it have watched founders drive companies into walls that a competent GC would have seen coming. They have receipts. The argument deserves a real answer, not a dismissal.

The Steelmans

Legal: Law is the codified residue of prior disasters. A good general counsel has read the case law on whatever you’re about to do and knows exactly how the last three companies that tried it got destroyed. Individual exceptions exist -- the aggressive deal-making lawyer who closes more than they kill, who treats legal as a competitive weapon rather than a defensive crouch. That person is real. The argument from their existence is that legal training isn’t inherently conservative; it’s that most lawyers self-select toward risk elimination, which is different from being structurally incapable of strategic thinking.

Finance: Someone has to be the adult. The graveyard has founders in it -- people who believed their own story right through the cash wall. WeWork burned $47 billion in valuation because Adam Neumann was allowed to run a real estate company as though it were a software company, and anyone who said otherwise got managed out. Theranos ran a decade of fraud because Elizabeth Holmes treated the CFO function as an obstacle. Jawbone spent $900 million and never figured out that it was losing money on every unit. The finance steelman is: you need someone whose entire job is to not believe the narrative.

Sales: The closest thing to continuous, adversarial customer feedback the company has. Losing fifty deals in a row teaches you something real about what the market actually wants versus what product thinks the market should want. Product-led companies that filter out sales signal -- that treat “the customer is wrong” as a default position -- build things nobody buys. The sales steelman is that you need someone in the room who has personally been told no by a human being recently.

HR: Culture is a product. Early Amazon, early Uber, early Activision treated people operations as overhead, and they paid for it in ways that are now well-documented and expensive. You need someone whose job is the human system -- not because people are fragile, but because coordination at scale requires deliberate architecture. The HR steelman is that companies failing to invest in that function don’t get a free ride; they get the dysfunction later, at higher cost.

All four of those arguments are correct.

None of them prove these functions should govern.

The Collapse

Every steelman above proves the same thing: these functions have genuine, necessary informational roles. A GC who spots a regulatory tripwire you missed is doing something real. A CFO who sees the cash wall before the founder does is doing something real. A head of sales who brings back unfiltered market signal is doing something real. An HR leader who identifies that a critical team is six weeks from mass resignation is doing something real.

Having necessary information is not the same as deserving authority over what to do with it.

The failure mode isn’t that these functions exist. It’s that they optimize for their own internal metrics rather than the company’s actual purpose -- and those metrics are structurally opposed to value creation, not just different from it.

Zero lawsuits is not the same as a company succeeding. It’s compatible with a company that ships nothing, takes no risks, and slowly loses its market to competitors who were willing to move. Zero attrition is not the same as a company succeeding. It’s compatible with a company full of comfortable people who’ve stopped doing anything that could get them fired. Closed deals at any margin is not the same as a company succeeding. A sales team that hits quota by discounting to break-even and customizing the product into unrecognizable shapes is actively destroying future optionality. Budget variance at zero is not the same as a company succeeding. It means this year’s resource allocation looks exactly like last year’s, which is fine if nothing changed and catastrophic if it did.

The finance steelman fails for a reason that sounds technical but is actually structural: the companies it cites (WeWork, Theranos, Jawbone) didn’t fail because finance lacked authority. Finance was captured -- by narrative incentive, by equity upside, by the social cost of being the person who says the emperor has no clothes in a room full of believers. Enron had an audit committee. It had outside counsel. It had a CFO (Andrew Fastow, who later went to prison). Captured finance is not the same as absent finance, and giving finance more authority doesn’t fix capture. It just makes the capture more dangerous.

The Netflix culture deck gets cited as evidence that HR generates culture. Reed Hastings wrote it. HR executed it. The cultures that actually worked -- Netflix, early Google, Valve -- were operator-driven. The HR apparatus didn’t generate them; it administered them. When HR tries to generate culture independently, you get engagement surveys that don’t increase engagement and values posters that actively predict toxic workplaces.

The aggressive deal-making lawyer exists. Most lawyers aren’t that. The incentive structure rewards avoiding loss, not creating gain. Individual exceptions don’t rebut structural tendencies; they illustrate the variance around a mean that’s still pointed the wrong direction.

What Happens Next

There is a phenomenon I’m going to call the Accountability Discovery. It has a specific shape, it happens at a specific moment in a specific kind of career, and it explains a lot of what you see when support-function executives reach the chair.

The next post describes it in detail, through four specific autopsies. Same cause of death every time.

A thermometer is necessary. It doesn’t drive the car.