AI Leaders Walk Back Job Displacement Claims as Adoption Challenges Surface

Why AI leaders are hedging their bets

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience. A year ago, the leaders of the biggest artificial intelligence companies were saying AI would take over huge swaths of the work people do, and soon. Some put timelines on it. But recently, many have changed their tune. Nvidia's Jensen Huang took aim at others blaming layoffs on AI. He called the narrative "too lazy" -- AI's barely had time to become productive, let alone be responsible for cuts that started years ago. Sam Altman offered something closer to a mea culpa, admitting he expected far more entry-level white-collar jobs to be gone by now, and that his intuition must've been off. Their walk-backs aren't rooted in AI being less capable than they thought. It hasn't gotten worse at writing code or drafting emails or summarizing reports over the past year -- it's gotten better. What's changed is the leaders' read on what determines whether that capability translates into fewer jobs, or more output from the same people. PERCEIVED VS. DEMONSTRATED FLUENCY These leaders are catching up to something some companies already understood: Handing someone an AI tool doesn't automatically make them faster or better at their job. It simply changes what their mistakes look like and how fast they're noticed. A bad first draft used to take one hour to write and ten minutes to spot. Now a first draft can take 30 seconds using AI. If you're lucky, or smart about it, a manager reviews it. All too often, it goes straight into a client email or a board deck. In a lot of ways, the work has become faster, sure. But has it really improved? Most employees and managers will tell you AI's changed how they work. Look at what they're producing, though, and the story gets murkier: a summary nobody checked against the source, a prompt that took longer to fix than the task would've taken on its own, or an analysis that sounds sharp until someone asks where the numbers came from. This gap between perceived and demonstrated fluency is the part the AI leaders seemed to underestimate when they made their original predictions. It's the part still missing from many corporate AI strategies today. WHAT CLOSES THE GAP Leaders who framed this moment as machines stepping in for people were asking whether AI can do a task in place of a person. Some would still say that's the question. But this isn't the right question to organize a strategy around. Whether your people get more done with AI matters more than whether AI could, in theory, do the job without them. And that comes down to things most companies skip, like teaching people to use it well and checking what the work looks like once they do. They hand out tool access, run a generic onboarding module, and call it training -- then wonder why the skill never shows up in the work. What closes the gap is teaching through live, hands-on, role-specific application. People should learn to use AI on the same problems their job throws at them, not a sandbox exercise that disappears the moment training ends. A salesperson needs to practice AI-assisted prospecting on accounts they know, not a customer relationship management demo. A finance analyst needs to build a model with AI in the loop on next year's forecast instead of a tutorial dataset. The learning must live inside the work or it won't transfer to the work once the course is over. BUILDING CAPABILITY WITH INTENTION Organizations consistently overestimate where their workforce stands, and that overestimation is exactly what makes the gap so persistent. You can't close a problem you don't think you have. The AI leaders walking back their predictions are conceding the point: Machines were never going to make people optional. The companies that build workforce capability on purpose -- rather than assume it'll show up on its own -- are the ones who'll see the gains everyone keeps hearing about. A year ago, the prediction was that AI would replace the need for people. A year later, the people making that prediction are the ones admitting they were wrong. The only thing AI seems to have removed is the excuse not to invest in your workforce. Daniele Grassi is president and CEO of General Assembly.

AI Is Closing Some Career Doors

Young workers do not learn the real lessons of work from the glamorous assignments. The first messy draft and the customer question they are not ready to answer are teachers. They also learn in the meeting where they mostly listen, scribble notes, and start to understand what good judgment sounds like. For decades, that has been the first rung of the white-collar career ladder. AI is starting to crack it. The strongest early signal is showing up in tech. Software developers aged 22 to 25 have seen nearly a 20 percent employment decline from its 2022 peak. That is the labor market telling us that the work that used to help young people get in the door is changing quickly. This is spreading beyond software. AI is compressing junior work across knowledge industries: drafting, research, QA, analysis, customer support, and first-pass code. Companies gave these tasks to younger workers because they are lower-stakes, repeatable, and teachable. That work was where people learned. The apprenticeship problem Companies should use AI to move faster. I run companies, invest in founders, and understand the pressure to operate more efficiently. But there is a cost if nobody is thinking about where judgment comes from. Judgment gets built through reps. You make a recommendation and find out why it was wrong. You write something, get it redlined into oblivion, and learn what "good" means. AI can manage much of that work. It cannot give a 22-year-old the scar tissue that comes from getting something wrong in front of a customer. That is the apprenticeship problem. The optimistic case for AI is productivity, but who benefits from that productivity? If AI makes large companies more efficient while shrinking entry points for young workers, opportunity concentrates. The people already inside the system get more leverage. The people outside have fewer entryways. A different future That is not the only possible future. The same AI weakening some traditional entry-level paths is also giving individuals access to company-building infrastructure that used to require full teams and meaningful capital. A single founder can now use AI tools to write and review code, design interfaces, build landing pages, run customer research, create content, and automate support. What once required a small cross-functional team before product-market fit can increasingly be orchestrated by one highly capable founder. Depending on the tools, a solo founder's AI operating stack can run roughly $3,000 to $8,000 a year, or closer to $12,000 for a small team. The revenue math is changing, too. AI-native startups can generate many times more revenue per employee than traditional SaaS companies. Solo-founded companies now represent more than one-third of new U.S. startups. This does not mean everyone should become a founder. Being a founder is hard, gritty work. Most companies fail. With AI, there is still potential rejection pain, cash-flow pressure, and required discipline to build something people want. But AI does change who can credibly start a company. Some of the brightest young people entering the economy may prove themselves as builders instead of as junior employees. The new graduate mostly likely to become a founder will not always look obvious on paper. Pedigree and headcount matter less than motion and mindset. Investors will want to know: What have you built? How fast do you learn? Can you use tools across domains? 4 founder profiles I see four founder profiles emerging. The first is the permissionless builder: the person already shipping apps, GitHub repositories, or small online businesses before anyone grants them a job title. This person may not have the cleanest résumé, but they have proof of work. They have crossed the line from "I have an idea" to "I launched a thing." The second is the cross-functional orchestrator: the person who can talk to a customer, build the first version, figure out distribution, and know when the system is unraveling. AI rewards people who can connect functions and do more than execute one task. The best founders will translate across product, customer, operations, and go-to-market. The third is the domain-native problem spotter: the person close to a painful industry problem who knows why the obvious solution never worked. They may come from logistics, healthcare, finance, retail, or a family business. They know the problem from the inside. With AI, that proximity becomes more valuable because they can prototype around the problem without a full team. The fourth is the self-taught operator: the person who learns through tools, feedback loops, and failed attempts. They learn because they are already trying to make something work. In an AI-native world, the advantage goes to people who can teach themselves the next workflow before it becomes a job requirement. These are the people I would bet on. They have AI-native workflows, think in systems, experiment quickly, and have a history of building things before anyone asked them to. Build a different career ladder Founders have always done things that look unreasonable. They fly across the country for one customer. They build the ugly first version. They send the awkward email. They get told no, then find another way in. AI does not change that part of company-building, but the builder becomes more leveraged once they decide to move. The conversation about AI and young workers should include creation as well as displacement. If AI produces enormous productivity gains, will those gains sit inside a few large institutions or spread through a rising generation of AI-native builders? The first rung of the old career ladder is crumbling before our eyes. We can spend the next decade trying to preserve every piece of it, or we can build new ways for young people to learn, create, and own the value they produce. The next career ladder may be built by the people who never got a clean shot at the old one.

Want to Avoid Being Replaced By AI? A Former Microsoft Exec Says This Is How to AI-Proof Your Career.

His simple test could tell you if your job might be vulnerable to AI. Rob Collie spent 13 years at Microsoft helping lead business intelligence work in Excel. Even with that technical background, he says that AI initially scared him. "I wasn't immune to that," Collie says in a new interview with Entrepreneur about the technology's potential effect on his business and career. He says his "fear turned into excitement" after he began studying how companies can actually put AI to work. Now, as CEO of the 50-person consulting firm P3 Adaptive, Collie is betting that AI will not simply eliminate work. It will change who gets to create software, solve operational problems and become indispensable inside a company. Collie recently published a book about organizational AI strategy called Fair Game. His central advice to workers is straightforward: Do not wait to see what AI does to your job. Help your employer use it. "You want to be on that train," Collie says. "The closer you can get to AI, and the more you can embrace it, the more you can give it a hug, that gives you the opportunity to be nimble." That matters because, in Collie's view, no one can reliably predict which roles will change most or how quickly. However, workers can develop practical AI skills that make them agents of change rather than passive observers. How to have an AI-proof career Collie says that the best way someone can AI-proof their career is to become an active adopter of AI at work, not waiting to see if the technology replaces them. He advises workers to learn to apply AI in a real business context beyond using a general chatbot for one-time tasks. He also tells workers to become part of the company's "adoption team" by identifying practical workflows AI can improve. "You can blow people's minds with this stuff," Collie says. "You can be the AI expert at your company in relatively short order, and that's the approach that I've been taking with my company." Collie says that workers can build or help create specialized AI assistants that use company-specific instructions, documents and data. He notes that no one can reliably predict where AI is headed, or how their job will change, so they should focus on developing agility rather than betting on any one job in particular. Collie urges workers to use AI to strengthen their judgment and productivity, not replace thinking. His preferred employees are "AI-first," meaning that their first question when approaching a task is what kind of AI system could help them do it better. He also encourages workers to leverage business knowledge when using AI. Collie says that people who understand their company's processes are more likely to build helpful tools with AI. The jobs most at-risk from AI Collie does not deny that AI will displace some workers. But he believes employers make a mistake when they view the technology only as a headcount-reduction tool. "You're missing some of the things that it can do for you," he says. AI can provide capabilities a company did not previously have, rather than merely replace existing tasks, he adds. Still, he sees particular risk for entry-level roles where employees can become productive quickly. His test is simple: How much training and organizational knowledge does a worker need before doing the job competently? "The entry-level jobs that are the most replaceable are the ones where a new hire can be truly productive in the first two weeks," Collie says. Call-center work is one clear example. These jobs often involve structured policies, decision trees and predictable response flows, all information that can be documented and supplied to an AI system. In contrast, roles that require years of accumulated judgment, nuanced relationships and substantial company-specific experience may be harder to fully automate. Collie uses the technical concept of an AI's "context window" to explain the difference. An AI can absorb only a limited amount of information at once. If the knowledge required to perform a job fits into that practical limit, the role may be more exposed. If it takes a long time and substantial experience to become effective, AI may have more difficulty replacing the worker outright. For young professionals, however, the threat extends beyond individual jobs. Collie said many companies appear to have become hesitant to hire inexperienced candidates while they try to understand AI's impact. "The zero-years-experience job is now turning into, like, a three-years-experience job," he says. He believes some of that pullback may reverse as employers gain confidence about where human talent still adds the most value. But job-seekers should not count on a return to the old entry-level ladder. AI is a new form of computing Collie's career has long centered on making sophisticated technology accessible to people outside large enterprises. At Microsoft, he led business-intelligence features in Excel 2007 before joining the early Power Pivot effort. That eventually became Power BI, a business analytics platform that turns raw data into interactive charts and dashboards. He later founded P3 Adaptive to help smaller companies use tools that had once been largely out of reach. That experience shapes how he sees the AI moment. He rejects the idea that businesses need to build their own large language models or hire elite AI researchers before getting started. "AI is just a new form of computing that we haven't had before," Collie says. He adds that companies need to learn its strengths, limitations and practical uses, much as they learned to use traditional software. Off-the-shelf chatbots can boost individual productivity, Collie says, but they cannot automatically understand a company's specific context. For example, ask a general-purpose AI tool to write a proposal, and it may produce generic, lowest-common-denominator copy. The difference comes when a business gives the system clear instructions and relevant internal knowledge. Collie compares that process to onboarding a highly educated new hire. An AI may know a vast amount about the world, but it knows nothing about an individual organization. "If you hired someone who had a Ph.D. in everything but had never met your business, they wouldn't be useful on day one," he said. "You would need to give them training." Build an AI handbook Collie calls those written instructions "handbooks." They explain what the AI's role is, which information it should use, what standards it should follow and how it should communicate. A marketing assistant could be supplied with a handbook that captures a company's brand voice, audience, products and approval rules. Instead of opening a chatbot that must be re-briefed every time, an employee could work with a specialized assistant that already has the relevant context. "Every time it wakes up, it needs to be handed that manual before it starts working with you," Collie says. "That is super easy to do." He argues that business knowledge, not advanced technical training, is the starting point. A manager who knows how a team handles customer issues, or a marketer who knows the company's tone, can help create the handbook. Technical employees may still be needed to connect systems and data securely, but they should not be isolated from the people closest to the work. "Don't treat AI as a technology problem," Collie says. "It's a business problem. It's about teaching it about your business." Collie is applying that philosophy inside P3 Adaptive. The firm has frozen hiring while it retrains its current staff and rethinks roles built around dashboard development. Going forward, he said, workers will need to be "AI-first" in how they approach their work. That does not mean allowing AI to think for them. It means using it to handle routine work, accelerate drafting and coding, organize context and extend what skilled professionals can accomplish. "It's not using AI to replace their brain," Collie says. "It's AI to take on lots and lots of stuff for them to make them more effective and more productive."