PwC surveyed nearly 50,000 workers across 48 countries and sorted them into four groups based on how comfortable they are with AI and how rare their skills are. At the top sit the "front-runners," who make up 14% of the workforce. At the bottom sits a group PwC calls the "engine room": more than half of everyone surveyed.
The engine room isn't full of low performers. These are the people who keep daily operations running: processing orders, answering calls, filling out paperwork, doing the steady work every company depends on. They simply haven't been given the time or training to pick up AI skills.
That gap now shows up in how people feel about their jobs, not just in what they can do. Front-runners are far more confident about job security and far more likely to push for a raise or promotion; PwC found this group is 33 percentage points more likely to be seeking a promotion than other workers. Engine room workers, by contrast, are the least likely of the four groups to feel ready for change, and only about half feel confident they will keep their job.
This isn't really about who tried harder to learn AI. PwC found that trust in managers has dropped 7 percentage points across the whole survey in a single year, and access to training has gotten worse at the same time.
Two years ago, 59% of workers said they could get the training they needed at work. Now only about half say the same, even as AI use keeps climbing. Companies have been spending their attention and budget on the smaller group already comfortable with AI, while the bigger group that keeps the business running has been left to catch up alone.
The same split shows up one level up, between companies rather than workers. A separate PwC study covering more than a billion job listings worldwide found that firms using AI heavily are growing headcount and raising wages faster than firms that use it lightly. The best performers inside that group are pulling far ahead of everyone else, including other AI-adopting companies.
The advantage isn't going to companies that simply bought AI software. It's going to companies using AI to make their most skilled people better at judgment, creativity and leadership, while other companies fall further behind on both fronts.
This runs against the message coming out of Washington, where the White House's top economic advisor has said there is no sign that AI is costing people their jobs right now. That may be true in the aggregate.
But PwC's numbers point to a different risk. It's not a shrinking headcount, it's a growing gap between the few employees becoming indispensable and the majority who keep the lights on without ever getting a chance to learn the tools everyone assumes they already know. Fixing that gap, not buying more software, is the actual task sitting in front of most managers this year.