![](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2025\/11\/2f42f79a6a5b0603c225d4579da75861bf-Steven-Ahlgren-Lede.rhorizontal.w1100.jpg\")
<\/p>\n <\/p>\n
This article was featured in New York\u2019s One Great Story<\/a> newsletter.\u00a0Sign up here<\/a>.<\/p>\n Not long after graduating from the University of Texas at Austin in 2021, Donald King landed a job as an associate at the London-based consulting firm PricewaterhouseCoopers. King had always assumed he\u2019d work in business \u2014 he\u2019d started his own hedge fund while still an undergrad \u2014 but a few years into the job, he decided he was more interested in tech than finance. Early in 2024, after PwC announced a $1 billion investment in artificial intelligence, he switched roles and started working as a data scientist for the company\u2019s nascent Global AI Factory.<\/p>\n King worked with engineers at PwC and OpenAI to customize teams of autonomous AI systems, called agents, for Fortune 500 companies. Normally, multinational companies contract thousands of people to modernize their backend software. Home Depot, for example, might enlist an army of consultants to update inventory or its SAP accounts-payable processes. Recently, though, AI agents have gotten pretty good at that kind of work. Consultants are some of the most prolific AI users, and King thought of himself as a kind of pioneer in a New Age of automation, creating and then deploying agents for PwC\u2019s clients. \u201cP-dubs,\u201d as King calls it, expected a lot from its workers. King put in 80-hour weeks, which kept the 26-year-old from going out on weekends. But he made six figures and lived in a one-bedroom high above Hudson Yards in a building with a pretty nice gym, where he sometimes took camera-off meetings while doing pull-ups. \u201cI was a meat slave,\u201d he says, \u201cand it was kind of a dream job.\u201d<\/p>\n The goal was to help clients \u201cdo more with less,\u201d as King\u2019s bosses reminded him, by automating whatever task they threw at his team. Occasionally, when King lingered on the downstream effects of his work, he felt like Dr. Frankenstein looking at his monster. \u201cThere was a sense of awe and then it\u2019s kind of shock and fear and almost a disgust,\u201d he says. King knew consultants were called hatchet men for a reason, but it was becoming clear to him that the agents his teams built were capable of wiping out not just individual jobs but entire job categories. \u201cThere was a large telecommunications client, and we were doing some crazy stuff for them. Once, we created an agent that was literally, like, a Microsoft Teams agent that was pretending to be a real, human employee,\u201d King says. \u201cThat\u2019s when me and my other teammates were like, \u2018Whoa, we need to sit and just talk for a little bit. What are we even doing right now?\u2019 Because that\u2019s someone\u2019s job, and if we have 45 of these agents working together, how many human jobs is that going to take? Are we just automating away people\u2019s livelihoods?\u201d<\/p>\n One night, after preparing for a big presentation, King stayed up late into the night talking with a few of his co-workers about the implications of their work. One colleague, a senior manager, wondered whether his kids should even bother studying computer science. \u201cWe just kept asking things like, \u2018What exactly are we doing with this job?\u2019 \u2018Am I going to be protected from what I\u2019m building?\u2019 \u2018Should I be upskilling and trying to learn new things as well?\u2019\u201d<\/p>\n When the big four consulting firms started laying off employees last year, King wasn\u2019t concerned. It\u2019s not uncommon lately for young consultants to spend weeks at a time \u201con the bench,\u201d waiting to be drafted by a project manager, but three years into the job, King\u2019s utilization \u2014 his time assigned to projects \u2014 was 100 percent. He felt even more secure when, late last year, he entered a companywide AI hackathon and, out of thousands of entries, won first place. In October, he presented his winning product, a team of AI agents he\u2019d built in his free time, to some 70,000 PwC employees. \u201cI was thinking, Oh, this is going to unlock so many opportunities for me. I literally thought to myself, I\u2019m safe from the layoff.\u201d Then, two hours after King finished his presentation, PwC laid him off.<\/p>\n King recorded a video of his own firing. In it, a partner with a southern drawl matter-of-factly explained that PwC was reshuffling \u201cto align our workforce to accelerate our business strategy.\u201d King came to see AI \u2014\u00a0 and his work in particular \u2014 as an ouroboros. The AI agents he built were intended to reduce by 30 percent both the client\u2019s team and the team of PwC consultants working for that client. \u201cIt was actually like I was just feeding myself into the AI meat grinder.\u201d<\/p>\n In the months following his layoff, King\u00a0started his own marketing agency and\u00a0became a kind of layoff influencer, telling his story on TikTok and offering advice. He\u2019s got a captive audience \u2014 in July, \u201cnew entrants\u201d as a percentage of the total unemployed population hit its highest number since 1988. Over the past year, King has watched former clients go through their own cuts, often affecting entry-level jobs. He won\u2019t share their names, but plenty of Fortune 500 companies like Intel, Salesforce, Chevron, UPS, Microsoft, and Procter & Gamble have enacted or announced major layoffs this year.<\/p>\n The extent to which recent layoffs are the result of automation is contested, but plenty of executives have nakedly blamed AI. Klarna CEO Sebastian Siemiatkowski says AI helped him shrink his workforce by 40 percent. Marc Benioff suggested AI agents were responsible for 4,000 job cuts at Salesforce, and AI was an explicit part of Accenture\u2019s 11,000-person reduction. Lufthansa announced plans to eliminate 4,000 jobs in the coming years thanks to AI, and Goldman Sachs is piloting Devin, an AI agent that will automate software engineering, which doesn\u2019t bode well for the 12,000 human engineers currently doing that work.<\/p>\n In August, economists at Stanford\u2019s Digital Economy Lab found that AI is having a \u201csignificant and disproportionate impact on entry-level workers in the American labor market.\u201d In particular, it\u2019s currently affecting workers between the ages of 22 and 25 with jobs in AI-exposed fields like coding. That\u2019s because AI has allowed for \u201cprompt engineering\u201d and \u201cvibecoding,\u201d the practice of building software using plain English; it can also generate much of the code that junior developers once wrote. The Stanford paper found a nearly 20 percent decline in employment for young software developers\u00a0 \u2014 perhaps making entry-level coders the first statistically significant casualties of AI automation.<\/p>\n There is a sense of anticipation among everyone \u2014 employers, job hunters, economists, academics \u2014 about what kinds of work will go the way of coding. When Microsoft Copilot recently launched agents capable of carrying out complex requests in Excel, Word, and PowerPoint, it called their new features \u201cvibe working.\u201d Over the past few months, I\u2019ve heard the terms vibe advertising, vibe writing, vibe designing, and vibe physics. Talk of an economy dominated by AI output is making it difficult for Gen Z to plan for the future. \u201cI\u2019ve been having that dreaded conversation about law school with my parents,\u201d says Meghan Keefe, who spent the past seven months searching for an entry-level job in media. \u201cI just registered for the LSAT to appease them, but I\u2019d enter the workforce four years from now and AI is going to be more advanced. It seems to be taking paralegal jobs away from people who have already been working there and people who have gone to law school, so that\u2019s scary. That decades-old fail-safe is seemingly gone.\u201d<\/p>\n It\u2019s hard to fault her and other zoomers for their nihilism. Executives at Ford, JPMorgan Chase, and Walmart have publicly warned of major disruption to the labor market. Anthropic CEO Dario Amodei says AI tools \u2014 the very thing his company is building \u2014 could soon eliminate half of all entry-level white-collar jobs. IMF chief Kristalina Georgieva says a \u201ctsunami\u201d is coming for labor markets that could affect 60 percent of jobs in advanced economies. And Elon Musk predicts AI and robots will replace all jobs. \u201cIt does not matter if you are a programmer, designer, product manager, data scientist, lawyer, customer-support rep, salesperson, or a finance person \u2014 AI is coming for you,\u201d Micha Kaufman, the CEO of Fiverr, a freelance marketplace, said shortly before laying off 30 percent of his staff.<\/p>\n The speed with which AI has measurably displaced entry-level coders concerns Anton Korinek, head of the Economics of Transformative AI Initiative at the University of Virginia. \u201cOver the next five months, I would definitely expect it to bleed into additional white-collar areas. Even if companies aren\u2019t going through huge layoffs, maybe they\u2019re now running pilots with AI agents; maybe they are saying, \u2018Hmm, let\u2019s not commit to hiring that associate,\u2019\u201d he says. Korinek has been studying AI\u2019s potential impact on the economy for more than ten years, and while he\u2019s always believed it would roil the labor market, he never thought that moment would arrive so soon. \u201cYou can really feel this coming. We\u2019re at the point where it\u2019s important for people who are honestly concerned about rapid advances that would lead to rapid job-market destruction to raise alarm bells,\u201d he says. \u201cGen Z is first where all of us are going to have to go.\u201d<\/p>\n Anna graduated from a small liberal-arts school in 2023. A few months later, she accepted an offer to work as a copywriter at an ad agency specializing in digital-media ads for beauty brands. A history major in college, Anna, who asked to be identified by a pseudonym, hoped to be a writer, and she saw advertising as a good way to make money in the meantime. For the first six months of the job, AI was hardly ever mentioned. Then her bosses began encouraging her to \u201cexperiment\u201d with ChatGPT. Last spring, AI became a requirement of the job. Now, Anna and her co-workers have AI programs read voice-overs, which was once part of their job, and use LLMs to generate ideas. \u201cThere are these awful AI-generated images that are just scary mannequin people and have really dumb taglines. You\u2019re working on a night skin-care mask or something, and ChatGPT is going to say, \u2018Get your beauty sleep!\u2019 Just the dumbest stuff,\u201d she says. \u201cAnd the worst part is that the marketing teams at these companies love it. They\u2019re like, \u2018This is amazing!\u2019\u201d Anna would much rather not use AI, but she has no choice. \u201cMy boss told me, \u2018You should be talking loudly in the office about using AI\u2019 and \u2018Don\u2019t bash AI in front of the higher-ups because that will make you look bad,\u2019\u201d she says. \u201cIt\u2019s so obvious I\u2019m training something to replace me, and I think that\u2019s kind of why I\u2019ve been hesitant. Because if I can make it do my job successfully, they don\u2019t need me to do my job anymore.\u201d<\/p>\n Gen-Zers are now accustomed to the fact that AI permeates white-collar work starting with both sides of the hiring process. People looking for jobs use AI to tailor r\u00e9sum\u00e9s and cover letters in order to get the attention of AI programs vetting applications. Managers at Shopify must now justify hiring a human by first explaining why AI can\u2019t do the job. \u201cIn interviews, it has been, \u2018What\u2019s the measurable impact of what you can do versus what a machine can do?\u2019\u201d says a 27-year-old consultant who\u2019s been job hunting since Deloitte laid him off in September. \u201cI feel like what I went to school for is now just being automatically populated by an AI within a matter of seconds.\u201d<\/p>\n In the office, bosses like Anna\u2019s push AI as if its use is a patriotic duty, like tending to a 21st-century Victory garden. And each use case, no matter how well-intentioned or innovative, can feel like a step toward job annihilation. \u201cI automated my co-worker\u2019s job with AI right before rumors of incoming layoffs, and now I am not sure how I feel,\u201d posted Asm Goni, a 22-year-old employee at a nonprofit in Queens, on LinkedIn in October. Goni, who graduated from St. John\u2019s University with a computer-science degree in 2024, had heard that his co-worker was combing through a database that included some 4,000 client IDs \u2014 tedious work that would have taken a week or more to finish. Using open-source AI tools and vibecoding, Goni created a tool that automated the entire process. \u201cThe instructions to AI were literally like, \u2018Step one: Do this. Step two: Once you\u2019re here, go there. Step three: Once you\u2019re there and you\u2019ve copied this information, put it back into the spreadsheet.\u2019\u201d Ultimately, Goni\u2019s co-worker wasn\u2019t laid off, but the experience made him anxious. \u201cIf everyone understood how to use AI \u2014 how to prompt it and order thoughts and provide references \u2014 a lot of jobs would be eliminated tomorrow,\u201d he told me. \u201cThe reason you haven\u2019t seen mass layoffs is ignorance.\u201d<\/p>\n Ignorance isn\u2019t really what\u2019s holding the line \u2014 executives at the world\u2019s largest employers are certainly aware of AI\u2019s capabilities and its limitations. But layoffs aren\u2019t the only measure of AI\u2019s effect on the labor market, especially when it comes to creative industries that run on freelance work. Several graphic designers told me that business was slow this year and they suspected AI had something to do with it. An illustrator said she was happy to hear from a prominent Substack writer about a potential commission, but when the writer heard the cost of original artwork, they went with AI art instead. The director of the Museum of the City of New York, Stephanie Hill Wilchfort, said she no longer needs to hire copy editors to work on annual reports. It hasn\u2019t saved the museum much money, but for an arts nonprofit, every penny counts.<\/p>\n Joshua Neckes co-founded Bobsled, a data-sharing platform, in 2021 and quickly grew it to around 50 employees. Over the past year, he\u2019s stopped hiring junior coders. AI has also slashed his legal overhead. \u201cI just throw a contract into ChatGPT and say, \u2018Call out the passages that are unusual.\u2019 Then I read it and I\u2019m smart enough to understand what\u2019s going on here. Then I bring it to my lawyer with my notes all ready. An associate doesn\u2019t need to look at the thing, and I just cut five hours off my legal bill.\u201d<\/p>\n Neckes sees AI as a total workforce reset. \u201cOur economy has a decent percentage of fake jobs that barely had a reason to exist in the first place. There\u2019s just no reason to deal with the headache of having young employees who frequently do the wrong thing, who frequently, you know, take up time and space.\u201d I asked him what Gen Z \u2014 the entry-level coders and associate attorneys in those fake jobs \u2014 were supposed to do without opportunities. \u201cBe smarter around the new shit and be cheaper,\u201d he says.<\/p>\n This attitude is, unsurprisingly, popular among CEOs. At a Beverly Hills conference last spring, Nvidia CEO Jensen Huang said that \u201cyou\u2019re not going to lose your job to an AI, but you\u2019re going to lose your job to someone who uses AI.\u201d Universities, HR types, and clickbait career sites are onboard, too, urging students and workers to futureproof their careers by \u201creskilling\u201d and beefing up their soft skills. Ravin Jesuthasan, a self-described \u201cfuture-of-work futurist,\u201d told me, \u201cThe quantum of upskilling and reskilling that each of us will need to do is going to increase exponentially.\u201d AI fluency will be central to what Bryan Ackermann, lead AI strategist for the global headhunter Korn Ferry, calls \u201ca new set of job families.\u201d What kinds of jobs are these, exactly? \u201cPrompt engineering, but I\u2019m not sure how long that is going to be a thing,\u201d he admits. \u201cI\u2019ve heard \u2018AI ethicist,\u2019 \u2018AI bias tester,\u2019 or I\u2019ve heard \u2018AI monitor.\u2019\u201d The day-to-day work of these roles is TBD \u2014 but so is the day-to-day work of most white-collar jobs in five years. \u201cThere\u2019s going to be a rapid evolution of what it means to be a manager,\u201d Ackermann continues. \u201cFor example, how do you have a difficult conversation with an employee when the employee that\u2019s the subject of the difficult conversation is not human?\u201d<\/p>\n \u201cJunior-level people are going to have to have a little bit more ambition,\u201d says Elisa Silverglade, director of automation for Techromatic, a small information-technology consultancy in Westchester that works with small and medium-size law firms. Before Techromatic, Silverglade spent 15 years running her ex-husband\u2019s law office, so she understands the interminable busywork that firms run on. \u201cThe copying and pasting, the uploading and downloading, the renaming, the saving, the attaching, the filing and the filling. Humans don\u2019t need to do that part anymore,\u201d she tells me. Silverglade doesn\u2019t want paralegals and associates to see technology as a threat to their jobs \u2014 her LinkedIn describes her ethos as \u201cautomating with empathy.\u201d \u201cThis is not replacing a person. This is allowing people to have time to do higher-value activities that not only serve the company better but are more fulfilling,\u201d she says. But how many ambitious employees do we expect employers to retain? And if AI is only getting better, what\u2019s the point of upskilling?<\/p>\n When Bryce Harris was in the eighth grade, he built his mom a computer using only YouTube and parts he bought at a local computer store. In 2024, he was among the first students at the University of Texas at Austin to graduate with a degree in informatics, the study of computational systems. Harris went straight to work for Microsoft, where he was an AI product manager, building and implementing AI agents for small businesses. Last spring, less than a year into the job, Microsoft laid off Harris, one of roughly 15,000 employees the company has axed this year. Harris didn\u2019t blame AI for the layoff, but I asked him whether, as a Gen-Zer with frontline experience of AI\u2019s integration into the labor market, he had ever experienced automation anxiety. \u201cI\u2019m an optimistic person,\u201d he told me. \u201cI try to stay away from that anxiety stuff. And as a builder, as an automator, I built the agents and I built those systems that can help, you know, redistribute those jobs.\u201d Part of Harris\u2019s job, he explained, was to provide companies with a plan to reabsorb employees whose jobs had been automated. I asked Harris whether the companies have adopted those plans. \u201cI wouldn\u2019t say I\u2019ve seen it, but I\u2019ve created programs that would have helped do that. Did the companies listen to what I advised? Well, that\u2019s another thing.\u201d<\/p>\n If the number of entry-level jobs dwindles across industries, naturally, the supply of experienced employees will wane. Without entry-level positions, how will the next generation of middle managers and executives get training? Millennials would be the last keepers of unwritten workplace knowledge, the intangible lessons and institutional know-how that haven\u2019t been, or can\u2019t be, fed into LLMs. What would the economy look like then? For those hoping to prevent the worst-case scenario, automation concerns have inspired a new wave of interest in ideas like universal basic income, profit sharing, and employee-owned enterprises (see Bernie Sanders\u2019s recent video<\/a>, \u201cAI Could Wipe Out the Working Class\u201d). But policy moves a bit slower than AI and, if income is abruptly severed from labor, there is a genuine fear that the chasm of inequality may quickly solidify. That fear has launched a meme warning that the deadline for prosperity is fast approaching. As @creatine_cycle recently posted on X, \u201cYou have 2 years to create a podcast in order to escape the permanent underclass.\u201d<\/p>\n You, a white-collar worker with a decade of experience, probably have a hard time imagining AI automating your job. And it\u2019s true that, at this very moment, AI probably can\u2019t navigate your company\u2019s ancient, proprietary software; it doesn\u2019t have the relationships with clients that you\u2019ve developed over the years; it can\u2019t pick up a report from a printer tray and walk it down the hall to your computer-illiterate boss; and it certainly can\u2019t walk the fine line between team-building gossip and lawsuit-waiting-to-happen at happy hour. It\u2019s probably comforting to imagine that automation really poses a threat only to young people who haven\u2019t yet developed your skills, who know only what they learned in a classroom. But what happens when AI becomes proficient at the core functions of your job? When management begins to realize your company can, as they say at P-dubs, do more with less? In September, OpenAI released the results of a study that tested AI\u2019s ability to accomplish 1,320 \u201creal-world economically valuable tasks\u201d across 44 occupations. The assignments were created by people with experience at Aetna, Bank of America, BBC News, Douglas Elliman, HBO, Boeing, the Justice Department, and others. To judge ChatGPT\u2019s performance, researchers had professionals with plenty of real-world experience complete the tasks so that the results could be compared blindly. A real-estate agent designed a sales brochure for a Florida home; a registered nurse assessed skin-lesion images and drafted a consultation report; lawyers were asked to draft wills or memos on complex mergers.<\/p>\n On average, the humans won \u2014 but barely. AI\u2019s biggest stumbling blocks weren\u2019t hallucinations or factual errors but issues like formatting and following instructions, areas where AI is fast improving. \u201cAIs have quietly crossed a threshold,\u201d wrote Wharton professor Ethan Mollick on Substack. \u201cThey can now perform real, economically relevant work.\u201d There is no guarantee AI will improve from here, and many economists believe that, like past technologies, AI will ultimately create more jobs than it kills. But Korinek, the Virginia economist, believes advances in agentic AI over the coming months will rapidly expand what these systems can handle autonomously, from research and analysis to creative problem-solving. \u201cWithin three years, I expect AI systems will write much better economics papers than I could,\u201d he tells me. When I suggest that in the future, we\u2019ll all work as plumbers, carpenters, and artisans, Korinek cuts me off. \u201cHave you seen the robots lately?\u201d<\/p>\n