Culture ‘Everybody’s Replaceable’: The New Ways Bosses Talk About Workers - Step it up, stop complaining—and make way for AI. CEOs are no longer lauding employees as the talent.

Illustration: Daisy Korpics/WSJ, iStock, Pixelsquid

By Chip Cutter
May 11, 2025 11:00 pm ET

Corporate America’s long-running war for talent sounds more like a war on the talent these days.

Not long ago, bosses routinely praised workers as their most prized asset, so much that some hoarded new hires before having enough for them to do. Today, with a giant question mark hanging over the economy, executives are pulling no punches in saying employees need to work harder, complain less and be glad they still have jobs.

“Work-life balance is your problem,” Emma Grede, co-founder of the shapewear company Skims and CEO of clothing label Good American, said this month. After recently cutting more than a 1,000 jobs, Starbucks CEO Brian Niccol said remaining corporate staff needed to step it up and “own whether or not this place grows.” JPMorgan CEO Jamie Dimon, in a profanity-laced internal meeting, told employees lamenting a return-to-work mandate that he didn’t care.

“I’ve had it with this kind of stuff,” he said. “I’ve been working seven days a week since Covid, and I come in, and—where is everybody else?”

The shift in tone marks a shift in power now that companies are shrinking their white-collar staff. With jobs harder to find, many workers are seeing perks disappear and their grievances ignored.


JPMorgan CEO Jamie Dimon. Photo: Al Drago/Bloomberg News

The latest episode happened at a contentious all-hands at Uber last month. The company had just changed the requirements to get a monthlong paid sabbatical to eight years of working at the ride-hail giant, from five years. A decision to require people to work at least three instead of two days in the office also drew complaints. CEO Dara Khosrowshahi suggested those unhappy with the changes deal with it.

“We recognize some of these changes are going to be unpopular,” he said in comments originally reported by CNBC. “This is a risk we decided to take.”

How much more license do bosses have to talk tough to staff? Take the outrage in 2023 when the head of furniture company MillerKnoll told staffers worried about bonuses to “leave pity city.” That comment, made in a video call, immediately went viral, sparking days of headlines and worker backlash. CEO Andi Owen quickly apologized, and said her comments were insensitive.

After the Uber town hall, on the other hand, Chief People Officer Nikki Krishnamurthy issued a memo saying the company would speak with some staff for being disrespectful in voicing their displeasure.

Workers like Donnie Donselman, who recently worked for a technology-services firm, can sense the new power dynamic. As he applies for new tech jobs, the 47-year-old has noticed that many companies now want applicants to do so many tasks, a position is essentially “three jobs” in one.


Uber CEO Dara Khosrowshahi recently said some of the changes implemented by the company will be unpopular with employees. Photo: Kent Nishimura/Bloomberg News

“They want it all,” he said.

In his job search, he tries to suss out the culture of a company because he has noticed the tough-talk language from CEOs and finds it worrisome. “All you’re doing is putting fear in people, and you’re not going to get good results from that,” said Donselman, who lives near Lexington, Ky.

Behind CEOs’ more brusque tone lies a disconnect between employees and executives, said Michael McCutcheon, an adjunct professor in applied psychology at New York University and an executive coach.

Some employees are operating like it is “still 2021,” when they could name their demands because of labor shortages and a surge in worker resignations, he said. Now bosses face a global trade war and sinking consumer confidence and feel they must ask more of employees to survive.

“This is a matter of pragmatism,” McCutcheon said.

President Trump and his billionaire adviser Elon Musk have helped set the more-aggressive tone in their bid to slash the federal workforce.

“Everybody’s replaceable,” as Trump put it shortly after the inauguration. Musk called his February demand that federal workers email what they accomplished that past week a “pulse check” to prove they did any work.


Tobias Lütke told Shopify employees that the company won’t make new hires unless managers can prove AI isn’t capable of doing the job. Photo: Dustin Chambers/Bloomberg News

Advances in generative AI also play a role. Shopify CEO Tobi Lütke recently told employees that the e-commerce company won’t make new hires unless managers can prove AI isn’t capable of doing the job. Other business leaders are warning their staff to adopt more AI—or else.

“AI is coming for your jobs. Heck, it’s coming for my job too. This is a wake-up call,” Micha Kaufman, CEO of the freelance marketplace Fiverr, wrote in a staff memo last month. Those “who will not wake up and understand the new reality fast are, unfortunately, doomed.”

Employees will someday have their moment in the sun again, said Charles A. O’Reilly, a professor of management at Stanford.

“When the market turns around, and job opportunities are plentiful, then CEOs will start to talk more about how important employees are, and employees will take advantage of it, ” he said.

For now, though, some executives say fewer, not more corporate staff, will help them run more efficiently. On Thursday, Match Group, which runs dating apps Hinge and Tinder, became the latest company to say it planned to thin its managerial ranks in sweeping layoffs. About one in five managers will be cut, and Match’s CEO, Spencer Rascoff, told investors the company is stepping up efforts to cut costs and rewire the organization to focus on its products.

“We lit a fire under the team here,” Rascoff said.

Source (Archive)

 

[Otterly]

You aren't alone, and the effect is scaling across the county.
I see a lot of old heads saying things like "The men who know things are retiring and there is no one to replace them!" like it's the young guys' fault. Sure, there's an element of prestige, but this complaint betrays a fundamental misunderstanding of how the world works now. If a young man could become a power plant operator today, and know he had that job for life and be able to afford the things the old man had, as well as a pension, that young man would be willing to literally stab someone to get it.

There's nobody to replace them, because those men who know things are insufferable boomer twits who still try to play 1980s power games with each other and everyone else around them, and nobody wants to fucking deal with their ghastly attitudes anymore.

I know a bloke who is now retired but he retired several years later than he wanted to because there was no one to replace him. He did something very important with nuclear control systems. Top bloke, had been asking for a replacement for years, it management wouldn’t let him hire anyone to pass the knowledge on. In the end he put his resignation on the table, after warning them for years he was going to and they flipped out and kept him on at frankly obscene pay. I remember him telling me how much it was if he stuck out the years, it was a lot.
But then he ran into another problem. The guy management gave him was a graduate in the speciality he needed but he was useless. They fired him, and this process repeated several times. Now this bloke is a really decent chap, he’s very chill and he wasn’t the problem, but they could not find anyone who they thought could replace him at all. In the end they got someone older who did replace him but he’s also looking at retirement and so it’ll happen again.
There is a huge amount of skill being lost. And management have lost the ability to hire good people because HR is in the way. Together it’s a disaster. Nobody is being trained on the job anymore which is what a lot of technical jobs need.
My own job is one of them. You need the base degree type knowledge but you’d fail just coming straight from college. You need to come in from colleague or life and sit as my sidekick for a few years and then you’ll get it, but nobody will hire to do that.
I bet there’s a lot of people like my friend out there.
He retired and did something pretty fun with the money.

 

[Chunky Salsa]

It’s a plague in the whole industry just now. What we do is really complex, and we used to do it with a million people and paper and now all the people are too expensive so we are using The Magical Systems but they are mainly crap. Exec level fell for the ai meme without understanding which bits actually work with it and which don’t. All they care about is line go up and shiny presentations.

My department has been asked to build an LLM that will produce technical information via RAG, and also produce diagrams, charts, and tables using the same. We have said that this will not work. We make measurements in the KeV scale, we work with intensely complex models that have to be validated properly or nothing will function. Management isn't listening and referred to it as 'whining'. I truly hate the way that people see AI and think that it's a magic wand; when in reality it's a cheap stage magicians trick, but most of the time he saws the poor girl in half.

 

[CEO of Gay]

Nah jeets are just retarded, not really intentionally malicious

¿Por Qué No Los Dos?

I’m just having a tea break and ironically I’m using an AI augmented system to model out something for a client.
It’s already inflated one number by twenty odd time, for no reason I can see, it just read ‘3’ in the input data and decided I actually meant 75. I’m not sure why. It’s not something I’m happy using, but I am no longer allowed to use the excel model I had before - despite the fact the excel model had its guts on show and I could track back the provenance of every number and see where it came from and what had been done to it. The new system is very pretty, it makes much prettier pictures than the excel, but it’s effectively a black box, and I can’t see under the hood.
Perhaps I’m just an idiot, or a Luddite, but the combo of that and the weird errors makes me a bit unsettled.
And yes! I am responsible for the numbers it spews forth! Isn’t that nice? I am reminded that I must check the working, but when I ask how I can check the working when I can’t open it up, I get angry replies.

I am hearing more and more anecdotes like this from medical and scientific research corners. I find myself awaiting the Therac-25 moment of LLM "AI".

Urgh. A lot of other people have to use it, so it’s less a crosshairs thing and more a useless inability to understand how different levels of the company process things. I have escalated the particulars upwards with my usual laundry list of WHY the thing is a problem and WHAT needs to fix it and HOW but I will as usual be ignored.
It’s a plague in the whole industry just now. What we do is really complex, and we used to do it with a million people and paper and now all the people are too expensive so we are using The Magical Systems but they are mainly crap. Exec level fell for the ai meme without understanding which bits actually work with it and which don’t. All they care about is line go up and shiny presentations.

And this is why. When one who understands their field, or understands how LLMs function raises issues with the middle management or execs why this AI tool is a poor fit and will produce poor results, the anecdotes I hear sound like the scene from Idiocracy. Brando's got what plants crave. It's got electrolytes.

 

[CryptoHermit]

My department has been asked to build an LLM that will produce technical information via RAG, and also produce diagrams, charts, and tables using the same. We have said that this will not work. We make measurements in the KeV scale, we work with intensely complex models that have to be validated properly or nothing will function. Management isn't listening and referred to it as 'whining'. I truly hate the way that people see AI and think that it's a magic wand; when in reality it's a cheap stage magicians trick, but most of the time he saws the poor girl in half.

I'm sat on meetings where these faggots are floating the idea of using AI in my spectroscopy lab. Not looking forward to it.

 

[AnotherPleb]

They won't have a choice, because if they tell the vast majority of the population (who will be out of work in this scenario) to fuck off and die, that population will kill them in response. The only way forward if most of the population is out of work is either welfare or disguised welfare. Anything else results in violent rebellion.

That's what the fuck off island vaults are for.

 

[rel=alternate]

The only thing more complicated than a codebase is a human, and most AIs struggle to deal with them. Besides the fact that AI can barely handle some of the most simple customer service tasks, the flip side is that most humans can't stand dealing with AI. Unless its an X-rated anime chatbot waifu, they aren't going to put up with the lackluster attempts of an AI to mimic human interaction.

It'd be cheaper and more effective to just hire, train and retain human workers to deal with human customers, but most management teams have no idea how to manage humans - or AI.

I'll give one quick example which I think makes it easy to understand where I am coming from.

Customer service can be seen as resolving complex issues, which it can be. And AI does not perform well at this. Or more so, it makes too many mistakes because it does not grasp the context or the way people can talk when angry.

It can really derail conversations by interpreting things literally which makes it obvious you're talking to a bot, making a bad situation worse.

However, when it comes to answering questions about a product or service, updating or cancelling an order, sending an invoice, sending a new order confirmation etc... AI will perform really well as long as you have trained it well and put some limitations where clear customer validation is required. On your end, you would never know it was an AI you talked to and performed the task.

And it does it instantly, which is pretty good customer satisfaction. It does not sleep, take breaks or make mistakes (unless confronted to a new scenario you did not identify). It also costs a fraction of the salary of one employee.

It will never completely eliminate the need for humans, but I think it's going to change a lot things for a lot of jobs.

 

[Lobster Cult High Priest]

if you're going to apply somewhere apply through the company website, I am 99% sure there are no real jobs on Indeed. I don't know if it's all just a scam to collect personal information or if these companies get a tax break for having "open positions they're hiring for" but something fucky is going on when 75% of the jobs listed are up for 6+ months with thousands of applicants and never filled or are taken down and reposted every few weeks with no changes to it.

if you apply to jobs on Indeed you might as well just make a sign out of cardboard and panhandle on the side of the street.

No HR department is able to keep up with the Jeet spam from Jeets flooding every posting with their Shit covered resumes that they aren't remotely qualified for

We need a Great Firewall for India.

 

[John.Doe]

However, when it comes to answering questions about a product or service, updating or cancelling an order, sending an invoice, sending a new order confirmation etc... AI will perform really well as long as you have trained it well and put some limitations where clear customer validation is required. On your end, you would never know it was an AI you talked to and performed the task.

And it does it instantly, which is pretty good customer satisfaction. It does not sleep, take breaks or make mistakes (unless confronted to a new scenario you did not identify). It also costs a fraction of the salary of one employee.

Actually, it can't do that and no, it does not cost the fraction of a cost of one employee. Do you work for an AI company or something?

 

[rel=alternate]

Actually, it can't do that and no, it does not cost the fraction of a cost of one employee. Do you work for an AI company or something?

Why do you lie about something you don't know?

It can and it does. I know it because I fucking use it. A big part of it is not even new tech, since you train it very much like you train a scraper when it comes to performing actions.

It is also about a third of the cost of an employee if you include taxes and take into account VAT. Unless your customer service is in Bengladesh, then it's super expensive.

As for me working for an AI company, you would think I would at least name drop a couple of these. They have been spamming the market for months with cold outreach by now, they are not super hard to find, so the only explanation is that you have not tried, but just decided it does not exist, actually.

 

[Otterly]

am hearing more and more anecdotes like this from medical and scientific research corners. I find myself awaiting the Therac-25 moment of LLM "AI".

The need for human checking is there and it’s going to be there for a long time. Or else you get multiples or random things. Which is fine for a random graph output but less so for radiation.

Brando's got what plants crave. It's got electrolytes.

I’m laughing because that exact scene was in my head last week. Like water? From the toilet?
The llm stuff is great within a niche. If you want an access bot or something that’s a simple call and response then it can be great. It’s exceptionally good at image diagnostics - the recent one where it can tell the sex of an x ray even from a small portion of an image is a good example.
But our execs do not understand the difference between that use case and what they want it to do which is to provide what people provide, which is thinking about stuff. It simply cannot do that.
The model I was using depends on certain time gap parameters, for example start of treatment cycle to next, or time for validation to whatever. So say you’re modelling out something like how many patients you can treat with a specific cycle based treatment schedule. If you have ten sites in the uk, it puts all of them ‘at once.’ It doesn’t understand that only one patient at a time can use the kit, or that you can’t have one staff member go to ten sites simultaneously in a day to set up the treatment itself. It just knows that the vector will be ready on this date, and so all the patients should be ready too.
It’s DUMB. It only knows what you tell it, and it can’t think.
And that’s the problem. It’s great for closed use cases. Look at those x rays. Pick out the tumours. Look at this array data. Pick out the statistically linked patterns of gene expression only in rhe urea cycle. Absolutely amazing for that. But it can’t think and troubleshoot.
And all the people who can are expensive and are getting ‘streamlined’ away for the systems.
I hate my job now. I used to work for people who were dedicated, smart and I respected. Now I spend my days muttering ‘it’s got electrolytes’ and wondering why people paid ten times what I am are unable to grasp what are simple concepts. The answer of course is that they’re not from the field like they used to be. They’re mba’s and bean counters and line must go up. That’s all they care about.

 

[The-Mad-Asshatter]

Last week I attended a very hoity-toity, invitation-only event with some people very high up the food chain in very, very high market-cap software companies. We're talking people who are in the c-suite and have Ph.D.s in comp sci from some of the best programs in the entire world.

And all of them said the same thing about LLMs and really anything using the Transformer architecture all LLMs are based on:

LLMs have already hit the wall. There is only small, incremental improvement to be derived from it now. Of course there are use cases where it's plenty useful at this level of functional, but there is no way that it's going to continue getting leaps and bounds better...or even small hops, really.

Just try to tell that to the naive tech dreamers that still want to shove it in everything, these people want it to be the magic wonder box that creates utopia and catapults mankind into this super advanced age, and they come at it with almost religious fervor. Ever hear how they gush on about the Singularity?

If AI can't take your job today, it's not going to tomorrow, either. The smartest minds in the room have already started telling the Ph.D. students of tomorrow "don't go into LLM research, it's a trap, look into new architectures and possibilities but don't fall down the Transformer rabbit hole."

I am glad there are sane minds still, it gives me hope.

It will take the rest of the world some time to catch up and notice, especially since Altman's the best hype man the world's seen since PT Barnum. But it will happen, and all the AI stocks will crater.

I hope so. I wonder how many companies will fall because of it. Altman and Co. are selling a dream, a wish, a hope, to people who want so badly to believe.

 

[Kendall Motor Oil]

This is also completely correct. They aren't just unwilling to pay their replacements, they're unwilling to train them or raise them up in any way. They refuse to teach and then act condescending when the younger people don't know things.

This is a huge issue with the "just get into the trades" mantra "right wing" people repeat. A lot of trades are guild systems where you can't operate yourself without licenses that require working under someone who has one. The previous generation didn't train people and instead hired unlicensed, cheap labor to do the work and successfully shrunk the talent pool. Now old people are complaining there aren't enough electricians and plumbers.
General contractors are the same way. Why would they vouch for someone who will just be competition? A friend of mine has been trying to find someone to vouch for him for years and only found someone willing to do it because the guy is retiring. Turns out the general contractor meetings are just rich old people divvying up guaranteed work to their friends and one another. There's no competitive pressure and the available license holders is shrinking per capita. My friend ended up switching to another industry and makes more money with less stress.

The young tradesmen who want to run their own business are struggling but people have endless copes about it.

 

[Otterly]

A lot of trades are guild systems where you can't operate yourself without licenses that require working under someone who has one. The previous generation didn't train people and instead hired unlicensed, cheap labor to do the work and successfully shrunk the talent pool. Now old people are complaining there aren't enough electricians and plumbers.
General contractors are the same way.

So is medicine in this country. The BMA is effectively a guild.

 

[Cybertonia]

This is a huge issue with the "just get into the trades" mantra "right wing" people repeat. A lot of trades are guild systems where you can't operate yourself without licenses that require working under someone who has one. The previous generation didn't train people and instead hired unlicensed, cheap labor to do the work and successfully shrunk the talent pool. Now old people are complaining there aren't enough electricians and plumbers.

Plus, the oldheads are just straight up shitty people to work with. I've got an uncle who is a lifetime pipefitter for the Navy, and the way he describes his worksite makes it sound like it's a retirement home for guys who never emotionally aged out of high school. Every once in a while, they get like one new guy who is under 30, and every story is either about how they henpeck and shit on the new guy and nitpick every single little thing he does, or its about how they endlessly torment the new guy with juvenile hazing pranks literally every single fucking day for months!

Usually later then I end up hearing "oh man another new guy quit after only 6 months, what a fucking lazy generation!"

 

[Kendall Motor Oil]

Plus, the oldheads are just straight up shitty people to work with. I've got an uncle who is a lifetime pipefitter for the Navy, and the way he describes his worksite makes it sound like it's a retirement home for guys who never emotionally aged out of high school. Every once in a while, they get like one new guy who is under 30, and every story is either about how they henpeck and shit on the new guy and nitpick every single little thing he does, or its about how they endlessly torment the new guy with juvenile hazing pranks literally every single fucking day for months!

This happened to a relative of mine in a machine shop. They did their "hazing" which seemed like thinly veiled bullying to keep young people out. They were insanely hostile to him using the computer on the ancient CNC machine and kept setting him up for failure. The guy is a nice person and avoided confrontation which probably made it worse.

The electricians where my friend lives are all unpleasant to do business with. I don't understand why given only a few get all the business in a populated area. They only employ Mexicans or Cubans to do the work and are never in the field doing work themselves. They're also the ones crying to the police about handymen doing electrical work without a license..... consisting of changing a light switch or a ceiling fan. The police now have stings where they try to convince handymen to do basic electrical and plumbing for undercover female cops, then fine them if they do, and shame them on Facebook.

Usually later then I end up hearing "oh man another new guy quit after only 6 months, what a fucking lazy generation!"

The cope I've seen from guys in that generation is that's what they had to go through but I don't believe them.

 

[pikachudidnothingwrong]

Universal basic income here we come. If they don't need to pay people to work anymore, then they have a lot of free extra money to pay in taxes.
I didn't/don't want it to be this way either but the libertardians and market parasites have spoken. We should probably institute some kind of public service to give people something to do to. (for extra money)

 

[CEO of Gay]

I’m laughing because that exact scene was in my head last week. Like water? From the toilet?

You know what I find particularly funny about that scene? The phrase itself not necessarily completely wrong. It is simply wrongly applied: blindly repeated, without understanding.

Spoiler: Off topic Brando/Plant sperging

Plants do crave electrolytes. Every macro- and micro-nutrient that a plant needs, it gets in the form of cations and anions dissolved in the water held in the pores of the soil -- i.e. electrolytes. Nitrogen (NO₃⁻) (NH₄⁺), Phosphorous ([PO₄]³⁻), Potassium (K⁺), Sulfur (SO₄²⁻), Calcium, (Ca²⁺), Magnesium (Mg²⁺), Iron (Fe²⁺) (Fe³⁺), Molybdenum (MoO₄⁻), Boron (H₃BO₃) (BO₃⁻), Copper (Cu²⁺), Manganese (Mn²⁺), Sodium (Na⁺), Zinc (Zn²⁺), Nickel (Ni²⁺), and Chlorine (Cl⁻), they're all taken up in electrolyte form.

Where the idea of plants craving electrolytes breaks down regarding the application of Brawndo is twofold: the balance of what electrolytes are in Brawndo, and their overall concentration.

Brawndo was presumably originally formulated for human consumption. Thus, its electrolytes balance, ranked by decreasing concentration, is probably sodium, magnesium, chloride, and calcium. Whether any phosphorous is present is a coin toss (some real world sport drinks have phosphorous, some don't). That's hardly a match to the list of plant nutrients above, now is it? The most essential nutrients for a plant aren't even present, and even if Brawndo contains phosphorous, that's the one plant nutrient which soil is usually not lacking.

Second, humans need their electrolytes in greater volume than a plant does. Humans are bigger than most plants. Humans perspire and excrete more liquid in a day (and thus require greater electrolyte replenishment) than a plant does. That's before considering the losses from strenuous exercise, which presumably Brawndo was originally formulated to address. Using Brawndo, the electrolytes would utterly overdose the soil.

By the time of the movie itself, it is doubtful anyone still around actually understands the how and why of the formula, just that it has what plants crave: electrolytes. Combining the two factors above, by the time the Brawdo had added enough calcium to the soil, the attendant amount of sodium would probably have turned the soil so saline that plants couldn't take up any water. If the plants can't get the water they need from Brawndo, because its sodium content is enough to be poisonous to a plant, then the plants would be wholly dependent on rain just to dextoxify the soil. What little rain water remained after that might be enough for a slow death.

That's probably more autism than Mike Judge originally intended that joke to carry. But it reflects the behavior of LLM "AI"s perfectly, doesn't it? It's just facile repetition of a phrase without understanding, because that phrase is what's expected.

That's why these executives like LLMs so much. It is a mirror held up to them. It repeats the expected phrases, and so seems smart. It reminds them of themselves, and so like them, it must be smart, too.

I'll take my puzzle pieces now.

 

[Capsaicin Addict]

That's why these executives like LLMs so much. It is a mirror held up to them. It repeats the expected phrases, and so seems smart. It reminds them of themselves, and so like them, it must be smart, too.

I think you may be onto something. It doesn't help that in my own experiments with LLMs they do have a tendency to parrot back what you say to them. They're literally an echo chamber.