Artificial intelligence will change what more than half of American workers do on the job within the next three years, and could eliminate up to 15 percent of U.S. positions within five, according to a new analysis from Boston Consulting Group. The numbers land at a moment when millions of workers are already uneasy about the economy and wondering whether Washington or corporate America has their back.
BCG researchers examined the tasks tied to 1,500 jobs, tapping government labor data to gauge which roles AI could augment and which it could simply erase. Their conclusion: between 50 and 55 percent of U.S. jobs will be "reshaped" by AI in the near term. A smaller but still staggering slice, 10 to 15 percent, could vanish altogether over the next five years, CBS News reported.
That distinction between "reshaped" and "replaced" matters. It is the difference between a worker who uses AI to do her job faster and a worker who gets a pink slip because a chatbot does it cheaper.
Matthew Kropp, a BCG managing director and senior partner, told CBS News that the shift will be felt even in jobs that survive.
"What people do in these jobs will be different, even if the job is still there."
Kropp warned against the corporate instinct to treat AI as a headcount guillotine. Too many boardrooms, he said, reach for layoffs first and strategy second.
"There's almost a knee-jerk reaction, we'll cut jobs [and have layoffs]. It's indiscriminate, and that's harmful for society because we need people to have jobs, but also harmful for companies themselves."
The argument is straightforward: slash payroll without thinking, and you lose the institutional knowledge and human judgment that machines still can't replicate. That may sound reassuring coming from a consultant. Whether C-suites actually listen is another question entirely.
BCG's analysis paints a split picture. Some jobs will grow because AI drives costs down, which in turn drives demand up. Kropp pointed to software engineering as the clearest example.
"Software engineering is the poster child for this. There is a massive backlog of software engineering tasks that enterprises have. There is a huge amount of software that isn't built because it's too expensive."
If AI slashes the cost of writing code, companies will want more of it, and they will still need human engineers to direct, review, and maintain the output. In that world, the job doesn't disappear. It multiplies.
Call centers sit at the opposite end of the spectrum. BCG's own analysis noted that when AI handles routine customer inquiries, the volume of interactions "does not expand proportionally." Translation: companies won't need as many people answering phones. The consulting group said productivity gains in that sector are "more likely to reduce the number of representatives required." Those are polite words for mass layoffs in an industry that employs hundreds of thousands of Americans.
Hands-on trades like plumbing and face-to-face professions like therapy were cited as roles AI is unlikely to touch, at least for now. Nobody expects a robot to snake your drain or talk you through a divorce.
That should give working-class Americans in the trades some comfort. It also highlights a reality that conservative economic thinkers have stressed for years: practical skills and physical labor carry real value that no algorithm can easily replicate. The push to funnel every 18-year-old into a four-year university, and into the white-collar desk jobs now most exposed to AI, looks worse by the day, a point underscored by recent gains in manufacturing productivity and wages.
Kropp urged business leaders to invest in re-skilling rather than firing. He framed it as both a moral and strategic imperative.
"Yes, some will go away, but many jobs you'll be re-skilling, getting people to work in a different way, and you have to expend effort to do that."
He added: "We should focus on re-skilling, and making sure people doing it are moving to other areas in which jobs will be fine."
Easy to say from a corner office at BCG. Harder to execute when quarterly earnings calls reward cost cuts and punish long-term investment. The track record of corporate America on worker retraining is, to be generous, mixed. And Washington's record is worse. Federal job-training programs have been a punchline for decades, expensive, slow, and designed more to keep bureaucrats employed than to move displaced workers into new careers.
The question of whether federal labor data itself can be trusted adds another wrinkle. As recent revisions to Biden-era jobs numbers showed, the government's own employment statistics have a habit of looking rosier on first release than they do after the auditors arrive. If BCG built its model on the same datasets, the projections may undercount the disruption already underway.
Kropp offered one note of optimism that deserves attention, and skepticism in equal measure. He pointed out that technological revolutions tend to create jobs nobody saw coming.
"When social media came out, did anyone ever anticipate that social media influencer would be a job?"
Fair point. But "social media influencer" is not exactly a career path that puts food on the table for a 50-year-old call center worker in Ohio. The new jobs AI creates may be real. They may also skew young, coastal, and credentialed, leaving behind the very workers who can least afford another economic upheaval.
That is where policy choices matter. A government that keeps taxes low, regulation light, and markets free gives displaced workers the best shot at landing on their feet. A government that piles on mandates, subsidizes favored industries, and lets bureaucrats pick winners does the opposite. Growing optimism among economists about the U.S. economy under Trump suggests the current direction is right, but AI disruption will test whether that optimism holds.
BCG's numbers are projections, not prophecy. Consulting firms sell forecasts the way car dealers sell undercoating, with confidence and a markup. Still, the scale of the estimate is hard to dismiss. If even the low end proves accurate, tens of millions of American jobs will look different by 2028. Millions more may not exist at all.
Congress will eventually have to grapple with this. Whether lawmakers respond with market-friendly reforms or with the kind of heavy-handed regulation that Europe favors will shape the outcome. The decisions made on Capitol Hill, including who holds the gavel and sets the agenda, will determine whether AI policy protects workers or protects incumbents.
And the re-skilling conversation cannot be left to consultants and corporate HR departments alone. If government-funded training programs are involved, they need accountability and measurable results, not the kind of ideological compliance exercises that have infected higher education and workplace training in recent years. Workers need real skills, not seminars.
The American economy has survived every technological disruption it has faced, the cotton gin, the assembly line, the microchip. It will survive AI, too. But survival is not the same as thriving, and the gap between the two depends on whether leaders trust free people and free markets or reach for the same old playbook of spending, mandates, and excuses.
