One year ago, on Labor Day 2025, Colorado AI News published something close to a warning.
Artificial intelligence, we wrote, was no longer merely threatening to disrupt the American workplace someday. Companies were already cutting jobs, changing how they hired and assigning work once done by people to machines. Young and inexperienced workers appeared particularly vulnerable.
A year later – and with ChatGPT's fourth birthday less than three months away – it seems worth asking the obvious question: Was that correct? Is AI actually taking our jobs? In looking at the latest numbers, the answer appears surprisingly reassuring.
The U.S. economy added 162,000 jobs in August, according to the Bureau of Labor Statistics, while unemployment held steady at 4.1%. There is softness underneath those numbers, and America's labor market is nowhere near as overheated as it was several years ago.
But there is also no sign in Friday's report of an economy being hollowed out by artificial intelligence – and that's notable after several years of warnings from tech leaders that AI could eliminate jobs on an extraordinary scale. Anthropic CEO Dario Amodei, for example, warned last year that AI could wipe out half of entry-level white-collar jobs and push unemployment as high as 20% within several years. Yet, at least today, on Labor Day 2026, unemployment is at a relatively low 4.1%. And that's actually down 0.2 percentage points from a year ago.
That doesn't mean the disruption hasn't arrived. Increasingly, however, it looks as though we may have been looking for it in the wrong place.
And even if artificial intelligence never eliminates millions of jobs, it raises another distinctly labor-focused question: If AI allows millions of Americans to produce substantially more during every hour they work, who gets the benefit?
The job survives. The work changes.
One of the most closely watched studies of AI and employment suggests that an economy-wide employment shock has yet to materialize.
In an August update, Stanford Digital Economy Lab researchers Erik Brynjolfsson, Bharat Chandar and Ruyu Chen examined payroll records covering millions of American workers through June. Their first finding was striking: they found no evidence of widespread, economy-wide job displacement associated with AI.
There is an important exception. Employment among 22- to 25-year-olds in occupations highly exposed to AI is significantly weaker than it is among young people in less-exposed occupations, with the difference appearing to come primarily through reduced hiring rather than layoffs. That's concerning, and it reinforces what Colorado AI News has previously reported about the increasingly difficult environment facing recent graduates and the "seniorization" of entry-level work — but it isn't mass unemployment.
Other research suggests something subtler may be underway. Workforce-data company Revelio Labs has found slower employment growth in occupations containing the greatest amount of AI-exposed work. At the same time, companies that have adopted AI have increased headcount faster than companies that have not, complicating any simple equation between AI adoption and job cuts.
Perhaps more revealing is Revelio's finding that about 87% of the recent change in the composition of American work is occurring within occupations, rather than because workers are moving from one occupation to another. The takeaway is that the job title may remain the same even as the work underneath it changes.
AI escapes the tech department
That change is increasingly visible in job postings. According to the Indeed Hiring Lab, 6.3% of U.S. postings now mention AI, nearly twice the previous peak of 3.3% in 2022. (Am I the only one surprised that it's not higher than 6.3%?) Additionally, an impressive 63% of those AI-touched titles were outside traditional technology occupations, according to Indeed's July analysis.
AI is turning up in sales, human resources, education, legal services, marketing, customer service, and administrative work.
Increasingly, the accountant remains an accountant, the attorney remains an attorney, and the software developer remains a software developer. But some portion of the research, drafting, coding, analysis, document review, or administrative work attached to those jobs can now be handed to a machine.
Colorado makes an interesting laboratory
Colorado is a particularly interesting place to watch this unfold because the state's labor market is hardly booming.
Colorado had about 2.974 million payroll jobs in July, according to BLS data, just 12,100 more than a year earlier, which indicates growth of only 0.4%. CU Boulder's Leeds School of Business recently cut its forecast for 2026 Colorado employment growth to just 0.2%, or approximately 6,300 jobs. Along the Front Range, the picture is similarly subdued: Boulder County had about 201,500 nonfarm jobs in July, down 0.3% from a year earlier, according to the BLS.
Those numbers do not demonstrate that AI is eliminating Colorado jobs. The Centennial State faces challenges having little or nothing to do with artificial intelligence, including housing affordability, slowing population growth, interest rates, and changing migration patterns. Employers nationally also have plenty of reasons besides AI to be cautious about hiring.
But Colorado has a highly educated workforce and a large concentration of people doing exactly the kinds of cognitive work generative AI is increasingly capable of assisting with – or performing itself.
Combined with sluggish hiring, that makes the Centennial State an especially interesting place to watch for a different kind of AI disruption – one measured less in mass layoffs than in slower hiring, attrition, rising expectations, and the increasing productivity of the workers who remain.
Which raises a harder-to-measure question: What happens when workers become much more productive even as employment holds steady?
The productivity bargain
That may be the more important Labor Day question: If artificial intelligence makes workers considerably more productive, who gets the benefit?
The optimistic version is easy to imagine. AI strips away drudgery and workers spend less time formatting PowerPoints, summarizing meetings, searching through documents or writing repetitive emails. Employees move toward higher-value, more interesting work. Companies become more productive and profitable. Maybe wages rise. Perhaps, at some point, workers even get some of the ultimate productivity dividend: more time. It's an appealing vision – but does anyone really expect that last part to happen on any meaningful scale?
We've seen enormous leaps in office productivity before. Think about what happened when secretaries and administrative workers moved from typewriters to word processors and personal computers during the 1980s and 1990s. Letters and reports no longer needed to be retyped from scratch because someone wanted to change two sentences. Documents could be revised, copied, stored, formatted and eventually distributed with dramatically less effort. Office workers became vastly more productive.
But the usual response was not, "Wonderful! You finished today's work by 2:00, so go home. And of course, we'll pay you for the full day." Instead, the definition of a reasonable day's output changed. Administrative staffs eventually became leaner – and many dedicated secretarial positions disappeared altogether. Individual workers took on broader responsibilities. Turnaround times accelerated. Businesses produced more.
There is every reason to think AI will follow a similar pattern, only much faster and across a far wider swath of the economy. A marketer who can produce in one hour what once took five may not receive four hours back – he'll likely be expected to create another campaign. A software developer whose AI coding assistant makes her significantly faster may not enjoy an emptier afternoon – she may inherit a longer backlog.
And a company that once believed it needed 12 people to perform a function may eventually decide that eight AI-equipped employees can do it just fine. None of that requires a mass layoff – it can happen one unfilled vacancy at a time.
Productivity is good. The distribution matters.
None of this is an argument against productivity. Productivity growth is one of the fundamental ways societies become richer. Over long periods, technological advances have helped raise living standards, increase wages, lower the cost of goods and services, and eliminate enormous amounts of tedious and dangerous labor – and, importantly, workers have shared in those benefits. The question is not whether productivity is good, or whether workers ever benefit from it. It's how the gains get divided, and how quickly.
If AI enables a company to produce 50% more with the same workforce, what happens to that additional value? Does some of it show up in higher wages, or does the company hire more people and expand? Do employees get more generous benefits, better training, greater job security – or perhaps even a shorter workweek? Or does the organization simply reset the baseline and decide that what used to qualify as exceptional productivity is now the minimum expected?
There is nothing inevitable about the answer. Markets will play a role. Workers with scarce skills can demand more compensation. Companies competing aggressively for talent may voluntarily share the gains. Organized labor can bargain over them. Public policy can influence them. Some employers will undoubtedly use AI to create genuinely better jobs, while others will use it primarily to get more output from fewer people. That makes the AI workplace not merely a technological story, but a labor story.
Workers still have agency
None of this means workers are simply passive passengers on the AI train, waiting to see where employers, markets, or policymakers decide to take them.
The lesson from the past year may actually reinforce one of the central arguments we made last Labor Day: The workers who learn to use AI most effectively are going to be in a much stronger position than those who don't.
But the stakes are no longer simply about keeping a job. If AI is going to make workers dramatically more productive, then those who can demonstrate that productivity – who can redesign workflows, automate routine tasks, improve quality, and show measurable results – should also be better positioned to argue for a share of the value they create.
That doesn't guarantee a raise, a promotion, or a four-day workweek. Bargaining power depends on many things, and an employer may decide that the reward for producing twice as much is simply being expected to produce twice as much.
Still, there is a profound difference between having AI imposed on your job and becoming the person who understands how AI can make that job better.
For workers, that means experimentation cannot remain optional. Learn the tools relevant to your profession. Figure out where they actually save time or improve results. Document those gains. Develop the judgment needed to know when AI is wrong. And, perhaps most importantly, become someone who can help decide how the technology is used rather than merely adapting after those decisions have been made.
A year ago, we argued that AI literacy was becoming a baseline skill of the 21st-century workplace. On Labor Day 2026, I would go one step further: AI fluency will become a source of worker leverage. And workers are going to need all the leverage they can get.
What does an AI-powered, 40-hour week look like?
For all the astonishing technological change of the past century, the standard American full-time workweek has proven remarkably durable. John Maynard Keynes, one of the most influential economists of the 20th century, famously predicted in 1930 that technological progress might eventually reduce the working week to around 15 hours. America got unimaginably more productive – but we didn't get the 15-hour workweek.
There are plenty of reasons for that. As societies grow wealthier, people consume more rather than converting every productivity increase into leisure. New industries and occupations emerge. Expectations increase. And the relationship between productivity and working time is far more complicated than one technological advancement producing one corresponding reduction in hours.
Still, AI brings the question back with unusual force. Generative AI isn't simply a faster typewriter. Its ambitions extend into cognitive tasks that occupy enormous portions of the modern workday: writing, research, analysis, planning, coding, design, and communication. Imagine that the more bullish productivity forecasts are right and a knowledge worker eventually becomes 50% or even 100% more productive with AI. What exactly happens at 2:00 p.m. when that employee has completed what used to constitute a full day's work?
That may sound like a theoretical question, but it won't remain one for long. Companies will answer it through staffing decisions, compensation systems, performance expectations, and corporate culture. Workers will answer it through the jobs they choose, the skills they develop, the wages they demand, and sometimes the bargaining power they exercise collectively.
And society may eventually answer it through debates about shorter workweeks, overtime, benefits, taxation, and how the extraordinary wealth created by AI should be distributed. Those arguments aren't separate from the AI revolution – they are part of it.
Labor Day after the first wave
Labor Day was created to recognize workers, not jobs, and that distinction feels particularly meaningful in 2026.
For most of the generative-AI era, the public conversation about employment has revolved around a frighteningly simple question: Will AI take my job? Nearly four years after ChatGPT's launch, the evidence doesn't support a simple answer. Some workers appear to be getting hurt already, particularly younger people trying to enter highly AI-exposed occupations. Some jobs are disappearing, and some companies openly say AI is part of the reason. Other AI-heavy companies are expanding. Overall unemployment remains relatively low.
Meanwhile, something less dramatic – and potentially more consequential – is happening every day. Millions of people are going to jobs that look a little different than they did a year ago. They're drafting with AI, coding with it, researching with it, analyzing with it, screening with it, creating with it, and increasingly managing it. Employers are changing what they expect people to know, and how much they expect them to accomplish.
The question for Labor Day 2026, then, may not simply be whether workers can hold onto their jobs. It's whether workers will share meaningfully in the extraordinary productivity gains those jobs may soon produce.
The AI jobs apocalypse hasn't arrived. The AI workplace has.
Now comes the argument over who it works for.
