As AI changes who gets hired and how beginners learn their jobs, now is the moment for employers to rewire the career ladder around it, and for policymakers to make it easier for workers to move into the jobs AI creates, while hiring and training practices are still being decided. AI is reshaping work on two fronts at once. In offices, it is changing what entry-level jobs involve and whether employers fill them at all. On construction sites and factory floors, the AI buildout is creating more skilled trade jobs than America can fill, even as robots take on some tasks. Employers, schools, and governments are responding to these pressures with very different strategies, and their choices over the next few years will decide which workers and companies come out ahead.
In some white-collar office settings, AI helps beginners the most, because it hands them know-how that used to take years to learn, and yet employers are cutting the white-collar entry-level jobs where beginners would use it. Entry-level office jobs have long worked as informal apprenticeships, where beginners learn by doing routine work alongside experienced colleagues; we call this the “apprenticeship paradox.” The organizations that do rewire their entry-level jobs around AI will make people more valuable at the same time as machines become more capable, ensuring they have the managers and experts they need a decade from now. The rest will save money for a few years, but then may find too few people ready to step into more senior roles.
Work is where the energy and compute revolutions touch people’s everyday lives, and AI moved into the American workplace faster than most employers were prepared for. About half of American workers now use AI on the job at least occasionally, up from about one in five in 2023, according to Gallup, and a separate St. Louis Fed survey found that nearly four in ten had used it for work by mid-2026. Use runs higher in desk jobs, with about six in ten technology workers using AI at least a few times a week. Yet only about a quarter of workers say their employer has communicated a clear plan for integrating AI into how they work. Executives are being asked to rewire their organizations while running them at full speed.
AI can make a new employee as effective as a seasoned one in a fraction of the time it used to take to train them on the job. A 2025 study followed more than 5,000 customer service representatives who answered questions by online chat at a Fortune 500 software company. When the company provided them with an AI assistant that suggested replies in real time, new representatives with merely two months on the job performed as well as colleagues with more than six months on the job who worked without AI. The assistant had been trained on the conversations of the company’s best representatives, so it passed their know-how straight to the newest hires, who improved the most.
Yet statistics show that employers are currently hiring fewer entry-level staff, and the cuts fall hardest on young white-collar workers in roles where AI helps most. New-graduate hiring at the 12 largest U.S. tech companies is down about 65 percent from 2019, according to the venture firm SignalFire. AI company Anthropic reported in October 2026 that junior employees’ share of its new technical hires had fallen from 27 percent to 9 percent, though its total technical hiring kept growing. And Stanford researchers found that by June 2026, employment for 22- to 25-year-olds in the jobs most exposed to AI like software development and customer service was about 19 percent below where it would otherwise have been, mainly because employers hired fewer of them. By contrast, employment of older, more experienced workers in those same jobs has kept growing, and no “widespread, economy-wide job displacement associated with AI” has been observed. The Wall Street Journal recently described a broader “deep freeze,” a low-hire, low-fire labor market in which AI is one of several forces making it harder for young college graduates to find work, while the opposite is true for workers without college degrees.
AI has also complicated the hiring process itself, with software now working on both sides of the table in a kind of applications arms race: job seekers use AI to send out applications by the hundreds, and employers use AI to screen them. By May 2026, 63 percent of American job seekers say they have been interviewed by an AI, according to the hiring software company Greenhouse. But employers that let software make every early cut risk screening out strong candidates, and ones willing to put in the harder work of looking past AI filters could gain an edge by identifying talent their rivals miss.
For now, there is no settled playbook, and employers are making dramatically different bets on entry-level work. Some are cutting across the board. Employers cited AI for more than 120,000 announced job cuts in the first nine months of 2026, and as Andy Challenger, whose firm tracks layoffs, put it, “Regardless of whether individuals are being replaced by AI, the money for these roles is.” Some cut jobs and then reversed course. Klarna, the buy-now-pay-later company, shrank its workforce by nearly 40 percent while betting on AI, then started recruiting again after its CEO, Sebastian Siemiatkowski, admitted the AI-heavy approach had produced “lower quality” service. Others are hiring more entry-level talent. And many have yet to decide. “Almost none of them have an honest plan for what AI is doing to their people,” said Jackie Swanson of the research firm Gartner. Even spending on employee AI itself is in flux. In October 2026, Accenture capped the Claude usage it pays for at $29 per employee a month after some heavy users ran up bills of $500 to $1,000, and Uber used up its entire 2026 AI coding budget by April.
The differences between these wildly divergent corporate strategies come down partly to how much of a given job AI can do on its own and partly to leadership indecision, and both are creating winners and losers among companies. So far, the steepest losses for young workers have been in jobs where AI can do much of the work outright, such as answering routine customer questions. Companies that judge correctly which jobs AI can handle and which it can only assist will lower their costs and raise output per employee. Companies that cut too deeply can lose customers and end up paying twice, first in severance and then to rehire. In October 2025, Forrester reported that a majority of employers already regretted AI-related layoffs, and a February 2026 survey by the outplacement firm Careerminds found that two in three companies that made AI-driven cuts were already bringing staff back. Those that put off deciding one way or the other may find that they lose ground quietly, and then all at once.
Pausing entry-level hiring is a different bet from laying people off, and some employers are making it, counting on AI to get new hires up to speed quickly whenever they start hiring again. The customer service study referenced above suggests that bet could work for routine roles in which an AI assistant can get a new hire up to speed in weeks rather than months. But it’s a riskier bet for roles built on judgment and experience, where Stanford researchers found that the number of experienced workers employed has grown in jobs that depend on know-how learned through years of practice. A long pause leaves a gap that shows up years later, when too few mid-level employees are ready to run teams or replace retiring senior staff, and the company has to compete with its rivals for experienced hires at higher salaries.
In our view, the employers getting this right are rewiring entry-level jobs so that AI lets beginners accomplish more in the same amount of time, while they still learn judgment on the job. IBM, for example, said in February 2026 that it would triple its entry-level hiring and rewrite junior jobs for the AI era. IBM’s chief human resources officer, Nickle LaMoreaux, argues that companies that “doubled down on entry-level hiring” will be the most successful in three to five years, because cutting junior staff today leaves a hole in middle management tomorrow.
Consulting and accounting firms are rewiring junior roles too, training beginners to check and direct AI’s work. Since January 2026, McKinsey has asked finalists for business analyst jobs to use its internal AI platform on a case study, so interviewers can see whether candidates catch the AI’s weak or wrong answers and improve on them. At KPMG, the goal, in the words of global AI workforce lead Niale Cleobury, is for “juniors to become managers of agents,” meaning the AI software agents that now handle routine tasks.
Early evidence suggests that rewiring work around AI pays off for companies and for their junior staff. In a September 2026 study, Boston Consulting Group found that companies reporting that they rewired how work gets done around AI delivered annual shareholder returns that beat their peers’ by more than 11 percentage points. For example, at Reserv, an insurance technology company, employees now build AI tools that capture their own expertise. As a result, the company can create a new employee training program in about 95 percent less time than before, and junior staff take on higher-stakes work sooner. The economic rewards for workers are becoming clearer, too. Worldwide, PwC found that job postings requiring AI skills grew about eight times as fast as the overall job market and offered wages averaging 62 percent higher than postings without those skills.
Schools and colleges are the first rung of the career ladder, and depending on where you look in the world, they are making opposite choices about AI. South Korea rolled out AI-powered digital textbooks in 2025, then stripped them of official textbook status within months after parents and teachers objected that the rollout was rushed, and Sweden has spent about $157 million putting printed textbooks back in classrooms. But the United Arab Emirates bet the other way, making AI a required subject in every public school grade from kindergarten through 12th grade starting in fall 2025, and Beijing now requires at least eight hours of AI instruction a year for students in China’s capital.
American schools are split, and the country’s two largest public school districts have restricted student AI use. New York City’s public schools reactively banned ChatGPT very early, lifted the ban four months later, but then imposed a new one-year moratorium on student use of generative AI through eighth grade in September 2026. Los Angeles has gone further, barring students in every grade from using AI on school devices for the 2026–27 school year. Meanwhile, Ohio State University now requires every student to use AI in their coursework, and at Alpha School, a private network with about a dozen campuses, students spend about two hours a day on academics with AI tutors. In the end, graduates who know how to direct AI and check its work will be more prepared for the rewired workforce, and the rest will compete harder for a shrinking number of traditional jobs.
Blue-collar work is being rewired too, as AI models and platforms, robots, and self-driving trucks take on tasks that only people could do a few years ago. At BMW’s plant in Spartanburg, South Carolina, a humanoid robot from tech company Figure worked 10-hour shifts for 11 months and helped build more than 30,000 BMW X3s, and BMW is now bringing in Figure’s next robot model. Self-driving truck company Aurora’s driverless trucks now run a 1,000-mile route from Fort Worth to Phoenix nonstop, and California approved testing of driverless heavy trucks in 2026 over the objections of the Teamsters union. While in principle these innovations displace traditional human jobs, it’s important to point out that each of these advances, and others like them, also creates novel jobs such as those in robot maintenance, remote fleet supervision, and safety. Indeed, demand for robotics technicians has more than doubled over roughly three years.
The bigger story in blue-collar work is a human shortage, because the AI buildout is creating more skilled trade jobs than America has workers to fill. “We haven’t seen this kind of growth for skilled tradespeople in generations,” said Don Slaiman of the International Brotherhood of Electrical Workers (IBEW) Local 26. The IBEW estimates the country will need about half a million electricians over the next decade, counting replacements for retirees, and one in three electricians is already between 50 and 70 years old. If the shortage persists, it is likely to slow construction of the very power plants and data centers the AI economy depends on, holding back growth.
For blue-collar jobs, policymakers can close the shortage fastest by expanding formal apprenticeships, paid jobs that combine on-the-job training with classroom instruction, along with short courses tied to specific openings. About 700,000 Americans were enrolled in registered apprenticeships, the ones that meet federal standards for pay, training, and supervision, as of March 2026, and the Trump administration has set a goal of more than 1 million. The federal money behind that goal is small. Every year since 2023, Congress has given the Labor Department about $285 million to grow apprenticeships nationwide, or roughly $400 for each current apprentice. President Trump’s 2027 budget would fold that money into a larger job-training grant to states that must spend at least 10 percent of it on apprenticeships, but for now Congress has kept funding at 2026 levels. The department currently spends most of that money on incentive payments to employers and industry groups that add apprentices, such as up to $3,500 for each new apprentice a manufacturer hires, and on grants to states, including about $85 million announced in April 2026.
Encouragingly, employers and community colleges are moving faster than Washington to add AI skills to that training. In April 2026, the Labor Department began looking for an outside contractor to write AI lessons and training standards that employers can add to apprenticeships for data center, telecommunications, and factory jobs, but the announcement came with no new money for employers or apprentices. Meta, by contrast, has tied training directly to its own construction. In July 2026, when Meta announced it would expand its AI data center in Richland Parish, Louisiana, to more than $50 billion of investment, Meta President and Vice Chair Dina Powell McCormick called the parish “a model for the rest of the country.” The company committed $5 million to nearby Louisiana Delta Community College for training and scholarships, the college built a four-week data center technician course with Meta, and more than 100 people have completed its new courses since January, all of them now working.
Some of the AI companies driving these huge workplace changes are now paying to help workers navigate them. In June 2026, Amazon, Anthropic, Microsoft, the OpenAI Foundation, IBM, and others backed RAISE US, a nonprofit led by former Democratic Commerce Secretary Gina Raimondo and former Republican Indiana Gov. Eric Holcomb that has secured more than $500 million of a $1 billion goal to help American workers retrain and move into new jobs as AI changes the economy. The companies have a direct stake in the outcome. More than 70 percent of Americans expect AI to mean fewer jobs, according to the Pew Research Center, and public opposition can slow the industry down, as local fights over data centers already show. Anthropic co-founder Jack Clark said RAISE US “is going to build some of the infrastructure we’ll need to navigate AI’s economic impacts.”
Executives, boards, and investors can judge any organization’s workforce strategy with three data-driven measures: how quickly new hires become fully productive, how many employees move into better jobs each year, and whether experienced employees are developing successors. Companies that track and share these numbers can make decisions about AI and jobs based on facts, and they will earn the trust of the employees whose work is changing.
In the end, energy and compute will determine how fast the U.S. economy grows in the near term, but the choices employers, educators, and policymakers make about work over the next few years will determine who exactly shares in that growth. The winners are likely to include the electricians wiring new data centers, the technicians who keep robot fleets running, young college graduates in fields like banking and law who join employers that rewire entry-level jobs and learn to direct AI, and the companies that train them. The people most at risk are recent graduates trying to start careers in fields where employers have stopped hiring beginners, and the companies most at risk are those that wait too long to decide how AI fits into their work.
Schools and colleges will see winners and losers too. With the number of U.S. high school graduates now past its 2025 peak, colleges are entering years of competition for a shrinking pool of students, and those that can show their graduates know how to use AI on the job should have a stronger case to make to students and parents. Business leaders should hire and train beginners with AI built into their jobs, educators should graduate students who know how to use it, and policymakers should fund the apprenticeships that fill the trades the buildout needs. If they act now, AI can grow the American economy faster and make its workers more valuable. If they wait, a generation of young workers will find the first rungs of the ladder missing.