The 5 Types of Bullshit Job

A guide to recognising yours before AI does it for you

---

I’m reading David Graeber’s Bullshit Jobs for the second time. The first time, about nine months ago, it changed how I thought about everything. A lot of the ideas that went into the article that changed my life (”The Death of the Corporate Job”) were shaped by this book. It gave language to something I’d been observing but couldn’t quite articulate: that a huge proportion of modern work serves no real purpose, and the people doing it know

.

Coming back to it now, with nine months of building Rumbo and hundreds more conversations about career confusion behind me, I have a deeper appreciation for what’s in these pages. The ideas have only become more relevant. And there’s a dimension that makes the whole thing even more urgent: AI.

For those who aren’t familiar with the work: Graeber was an anthropologist at the London School of Economics who wrote an essay in 2013 called “On the Phenomenon of Bullshit Jobs” for Strike! Magazine. It got over a million views, was translated into seventeen languages, and he turned it into a full book in 2018. He passed away in 2020, at 59.

The starting point is a prediction that Keynes made in 1930: that technology would advance enough for countries like Britain and the US to achieve a fifteen-hour working week by the end of the century. That obviously didn’t happen. The argument is that instead of converting the gains of automation into more free time, we’ve invented entire categories of work that serve no real purpose. A YouGov survey cited in the book found that 37% of British workers didn’t think their job made a meaningful contribution to the world.

A bullshit job is defined as a form of paid employment so completely pointless, unnecessary, or pernicious that even the employee cannot justify its existence, even though they feel obliged to pretend that this is not the case.

Over the past year I’ve had hundreds of conversations with people about their careers, and the ones who are struggling most are in easy jobs. The pay is fine, the hours are reasonable, and they still feel like something is slowly being drained out of them. They can’t point to anything specific because there’s nothing specific to point to. The problem is the absence of anything meaningful, which is much harder to diagnose than the presence of something bad.

Six years after Graeber’s death, there’s another dimension to all of this. The data so far shows that AI hasn’t had the dramatic impact on employment that one might imagine. But anyone who has spent real time using these tools knows this technology has the potential to reshape how we work entirely. AI doesn’t care about maintaining the illusion of productivity. It doesn’t need to look busy, it doesn’t need a desk, and it can do a frightening amount of what the book describes faster and cheaper than a person.

The phenomenon is broken down into five types, built from over 250 testimonies collected from workers around the world. I wanted to walk through each of them, because I think they’ll resonate with anyone trying to make sense of their own career confusion. But I also want to look at each one through the lens of what’s coming, because if AI develops the way many people expect it to, it’s going to force a reckoning with all five.

1. Flunkies

Flunky jobs exist only, or mostly, to make someone else look or feel important. The people in them are given minor tasks to justify being there, but the real reason for the role is to serve as a prop for somebody else’s status. Think of a small company that doesn’t need a receptionist because the phone rings once a day, but hires one anyway because it makes them look like a proper firm. Or the doorman at a luxury hotel whose function is to signal to guests that they’re the kind of person who deserves to have a door opened for them. Administrative assistants, store greeters, people whose calendars are empty but whose desks need to be occupied because the person above them needs to look like someone with people working under them.

The insidious thing about flunky roles is that they corrode you precisely because nothing is obviously wrong. There’s no villain, no unreasonable workload, no crisis. Just a slow accumulation of days where you contributed nothing, surrounded by people who never question it. The longer you stay, the harder it becomes to imagine doing anything else, because the job has quietly convinced you that you don’t have anything real to offer.

AI probably won’t eliminate flunky jobs directly, because flunkies exist to satisfy egos, and egos don’t respond to efficiency arguments. But it will make the emptiness harder to hide. When an AI assistant can manage a calendar, draft correspondence, and handle scheduling in seconds, the person whose job was to do those things on behalf of someone important will find it increasingly difficult to pretend there’s a reason for them to be there.

2. Goons

Goons are people whose jobs have an aggressive element to them, and who exist mainly because other people employ them. The clearest example is armies: countries need armies only because other countries have armies. If nobody had one, nobody would need one. The same logic extends to lobbyists, corporate lawyers, telemarketers, and PR specialists. These positions exist because competitors or adversaries have their own goons, which creates an arms race of mutual aggression that produces nothing of value for anyone outside the game.

It always makes me think about how much of the professional services economy is essentially defensive. Billions spent on corporate lawyers because other companies have corporate lawyers. PR teams countering the PR teams of competitors. Whole industries that exist to cancel each other out.

AI is likely to supercharge the goon economy before it reduces it. If your competitor can now deploy AI-powered PR, AI-generated legal research, and AI-driven telemarketing at a fraction of the cost, the arms race gets faster and cheaper. More goons, not fewer. At least in the short term.

3. Duct Tapers

Duct taper jobs exist because of a glitch or fault in the organisation. These people are there to solve a problem that shouldn’t exist in the first place. The clearest examples are people whose whole role is to clean up after sloppy or incompetent superiors. One testimony comes from a woman called Magda whose job was to proofread research reports written by her company’s star researcher-statistician, a man who didn’t know the first thing about statistics and couldn’t write a grammatically correct sentence. A lot of these jobs come from processes that could be automated but haven’t been, either because nobody has got around to it, because the manager wants to keep as many people under them as possible, or because of some structural mess nobody wants to touch.

The frustration of duct taping is particular, because you can see the root cause of the problem every day, but fixing it is never your job. You just keep patching.

This is where AI is going to hit hardest and fastest. Duct taping is, almost by definition, the kind of repetitive, pattern-based, fix-the-same-problem-again work that AI excels at. The code that should have been written properly, the reports that need reformatting, the data that needs cleaning. If your whole job is patching things that a better system wouldn’t break in the first place, AI is either going to replace you or replace the broken system that created your role.

4. Box Tickers

Box tickers exist to make it look like something useful is being done when it isn’t. Survey administrators, in-house magazine journalists, corporate compliance officers. One of the examples in the book is a woman hired to coordinate leisure activities at a care home. Her real job was to conduct elaborate surveys of the residents about what entertainment they wanted. She compiled the surveys into reports. The reports were filed. The residents got the same entertainment regardless. The whole exercise existed so the care home could show, on paper, that it was consulting its residents.

I think this is the category most people in corporate environments will recognise instantly. The quarterly reports nobody reads. The training modules you click through to get the completion certificate. Entire departments whose output is paperwork about the work, rather than the work itself.

AI can generate box-ticking output at industrial scale. Compliance reports, survey analyses, corporate documentation. The question is whether organisations will use AI to eliminate the box ticking, or just use it to produce more of it, faster. If I had to bet, I’d say most organisations will initially do the latter. The boxes still need ticking. They’ll just tick themselves now.

5. Taskmasters

Taskmasters come in two forms. The first type are people whose entire role is to assign work to others, where they themselves believe there’s no real need for them to be there. If they disappeared, their team would carry on fine. These are the middle managers who spend most of their time checking whether tasks have been completed, even though they have no reason to think anyone would behave differently without them.

The second type are worse. Their main role is to create bullshit tasks for others, or to supervise bullshit, or to generate entirely new bullshit jobs. They’re the ones producing the strategic vision documents and mandatory reporting frameworks that fill other people’s calendars with work that everyone privately knows is pointless. The first type of taskmaster is useless. The second type makes other people’s working lives worse.

AI project management tools are already starting to handle task allocation, progress tracking, and workflow optimisation. For the first type of taskmaster, the one who’s mostly just monitoring, AI makes the redundancy obvious. For the second type, AI might make things worse. If you can now use AI to generate strategic vision documents in minutes, a bullshit generator with access to AI becomes a more prolific bullshit generator.

---

The diagnosis in this book is brilliant. But the people I talk to every week aren’t asking why the system produces bullshit jobs. They already know. They’re asking how to get out, and more importantly, what to move towards. Because leaving a bullshit job without any sense of what you’d find meaningful just puts you at risk of landing in another one. The problem repeats if you don’t understand what kind of work would feel different for you.

AI is going to accelerate that question for millions of people. If your job is one that AI can do faster and cheaper, you’re going to be forced to confront what so many of the testimonies in the book already describe: the work was never necessary. The difference is that now it won’t just be a feeling. It will be a fact, with a redundancy notice attached to it.

That’s a lot of what we’re building at Rumbo. The whole premise is that self-understanding has to come before job searching. If you don’t know what pulls you in, what holds your attention, what problems you’re drawn to, you’ll end up optimising for salary or status or convenience, and those are exactly the forces that lead people into bullshit jobs in the first place. AI is going to make that cycle more visible and more urgent, but the answer is the same: start with who you are.

If you haven’t read the book, I’d really recommend it. And if you’re in one of these roles right now and trying to figure out what comes next, Rumbo might be a good place to start.