Corporate Politics Is the Cockroach of Capitalism

If you’ve doomscrolled LinkedIn recently, you’ve concluded AI will replace leadership in the modern capitalist system. Surely you’ve read the baloney about “1 employee billion-dollar unicorns are coming!” Here’s the inconvenient truth: the replaceable part of “leadership” already got replaced. Decades ago. Leadership got a lobotomy in the 1970s. What’s sitting in the corner office isn’t a visionary decision-maker. It’s a cockroach with a big package of stock incentives.

We’ll start with the layup. Automation (inclusive of AI) has been eating the bottom end of the labor market for a very long time. Globalism and offshoring has made it seem particularly nasty, but those aren’t the real point. The real point is that the machine can do rote, deterministic labor far more efficiently than a human. Dig a hole with: hands -> shovel -> horse and plow -> tractor. One human gets a lot more done with a tractor. Endgame: dark factories. There. That’s enough on this point. We have all known the punchline for a long time: rote human labor is never coming back.

What’s less apparent is this: Automation picked the lock and let itself out the jail cell of deterministic physical labor. What does it do now? Cognition. Epistemic work, formerly the province of the middle tier of human endeavor. We can argue about how much epistemic work it will consume, but whether the human is in the driver seat, or in the side car, or left on the side of the road, but there can be little argument that cognition is squarely in the sights of automation. Automation is pretty good at producing relevant, actionable information, and increasingly, the recommended action itself.

Things that were bastions of human judgement have already fallen. Long ago. Mortgages. Insurance underwriting. Credit scoring. Risk stratification. These were once expert domains. They are now software. Next up (already underway, really) are legal, compliance, finance, software engineering, procurement, marketing, and operations. These functions will undergo the same compression factory work did a century ago. Not because they lack intelligence, but because they are legible - meaning the rules can be expressed.

Here’s the story of one such situation, the epistemic bastion formerly known as Procurement.

Nate is a 45-year-old senior director of procurement at a mid-sized company. A few years ago, he managed a team of six. After two rounds of layoffs, it’s four. Nate is competent. He knows his vendors, understands the regulatory landscape, and has a good instinct for risk. But his job has a fatal property: everything he does reduces to structured decisions over legible inputs. Prices, lead times, quality metrics, delivery risk, historical performance, contractual terms. Procurement software already aggregates these. Decision-support tools already score them. The next step is obvious and already underway: systems that recommend not just vendors, but contracts, order sizes, timing, and contingencies. At first, Nate reviews and approves. Then he handles exceptions. Eventually, the exceptions get rarer. Nate isn’t replaced by a machine because he lacks judgment; he’s replaced because his judgment can be expressed, simulated, and optimized. His role collapses not upward into strategy, but downward into sign-off, and eventually disappears entirely.

What remains is work that is inherently illegible. Work that resists abstraction. “Tacit knowledge resists formalization”. Polanyi, 1958. For lack of a better phrase, we’ll call this remaining kind of labor Residual Work. Residual Work is embodied. You have to literally be there, and it has at least one of two other qualities: a) context is everything (so much so that it’s effectively NP-complete), or b) it acts as a risk or liability sink (like a heat sink).

The surviving middle (the “cognitive middle”) is not: middle management, knowledge work or white-collar generalism. It is: contextual physical labor, local judgment under uncertainty, work embedded with stuff / atoms, work that accepts liability, and work that breaks abstractions.

This looks like regression. Hard Jobs. Yep. That’s the ticket, bucko. This is what’s left after late-stage capitalism leverages automation on every layer of the labor cake. Automation, asserting itself.

You might be looking for some concrete examples of what Residual Work is. Not many prestigious jobs. That’s the point. Here ya go:

Physical therapy
Home inspections
Sex work
Renovation (not new construction)
Nursing (but not doctors)
Trial lawyers

I could go on, but you do the math. I am conflating some of these Residual Work jobs that are liability sinks with those that are NP-complete context problems. You can sort out which is which.

Interestingly (and a bit sadly, I’m afraid), leadership is Residual Work.

But let’s be clear about what leadership really IS and IS NOT.

You think, because you read, or watch tv, or otherwise listen to the mythology of capitalism, that companies and institutions are run by leaders. Smart, wise, capable people.

That story is comforting. It sells ad space. It’s also mostly false.

News flash: Institutions don’t actually run on wisdom. They run on forecasts. Demand forecasts. Revenue forecasts. Risk forecasts. Growth forecasts. Budget forecasts. Headcount forecasts. Healthcare cost increase forecasts. Usually made under time pressure, with partial information, and a generous helping of motivated reasoning, plus “this worked ok last year”.

Leadership, in practice, is the act of choosing which forecast to believe. But more importantly: which forecasts a) will survive a board meeting where, “Susan better not shoot this down”, b) will resonate with the street, given all the things we’ve told them in the last 4 quarters, c) make me and my team look competent and trustworthy.

A little context: The period stretching from the end of World War II to the early 1970s is widely categorized by economic historians as the golden age of Managerial Capitalism. In this golden age, the public corporation was more than a bundle of assets. It was a durable social institution with obligations that extended far beyond the next fiscal quarter. But, in the 1970s it got brain damaged.

So now go on and fire up your Amazon: Rakesh Khurana’s From Higher Aims to Hired Hands and William Lazonick’s theory of Predatory Value Extraction. The leadership lobotomy involved the deliberate severing of managerial “higher brain” to a single neural pathway: the maximization of shareholder value.

The result was that this new-style corporation, the post-1970s company, got leaner and more responsive to market signals, let’s grant that. But it is clearly prone to short-termism, stripped of organizational memory and unmoored from the economic health of its employees and communities.

What’s left? Leadership is an optimization game played with unwritten (illegible) rules. Ie., just politics. You might put lipstick on the pig and say “Politics is the art of getting things done,” and you wouldn’t be wrong. But what leadership IS NOT is clearer than ever: It’s not “hard calls” and “vision”. Anyone who’s worked at a Fortune 500 knows this viscerally.

I have kids in elementary, middle school and one (soon two) in high school. We talk a fair bit about what they might do for work. This is an entirely different discussion than I had growing up. The “doctor, lawyer, astronaut, finance bro” choose-your-own-adventure has been steamrolled by automation.

Here’s a made-up story to illustrate what I mean by Residual Work:

Janet is a 38-year-old career switcher who just got her general contractor’s license. She does bathroom renovations. On paper, this looks automatable. In reality, every job is an adversarial battle. Old houses hide rot, mold, illegal wiring, out-of-date plumbing, and DIY homeowner “changes”. Materials arrive late or wrong. Codes vary by locality and enforcement, by inspector. Clients change their minds mid-demo. Schedules collide. The tile guy got injured and he’s out. Every project requires on-site judgment under uncertainty, and when something goes wrong, Janet owns it. She absorbs the risk the system can’t price and the context no model can fully capture. This is Residual Work. It survives not because it’s inefficient, but because it breaks abstraction. You can simulate a bathroom. You can’t simulate this bathroom, in this house, with this client, under these constraints. Janet’s job endures because it is embodied, illegible, and accountable in the real world of misbehaving atoms.

So what do I tell my kids? I tell them:

DO NOT optimize for prestige. Prestige is a map of what’s about to be automated.
Get comfortable being in the room when things go wrong.
Ambiguity is a feature, not a bug. If the job has clear rules, it’s already being automated.
The system needs someone to blame. Understand that, and decide if you can live with it.
Learn a trade that touches atoms in unpredictable environments.
Only consider leadership if you love politicking.

Also, I can’t not name the elephant in this room. There’s a stark class dynamic, isn’t there? Residual Work jobs are disproportionately filled by women, minorities, immigrants, and people without “significant” credentials. The residual economy isn’t just hard, it’s already staffed by people with limited options.

Tell your kids: Hard jobs are what’s left. Make peace with that. Find dignity anyway.