AI Use Is Up and Training Is Down. Your Most AI-Savvy People Noticed First.

Only 51% of workers say they can get the training they need, down from 59% a year ago, and the decline came in the same 12 months that AI use at work climbed to 64%. PwC's Global Workforce Survey, published September 29, collected those answers from 49,364 workers across 48 countries and regions and 29 sectors. The technology spread faster than the teaching did. And the employees most fluent in that technology are the ones most prepared to leave.

Behind the 51% is someone you probably know: the colleague whose team received an AI assistant in the spring and a higher target by fall, with a training link somewhere in between. She uses the tool every day now. Nobody has shown her what excellent work with it looks like, and she has started to wonder whether she is getting better at her job or only faster at it.

What the PwC numbers show

Daily use of generative AI rose from 14% to 22% of workers in a single year, and overall AI use at work increased by 10 percentage points. Over that same period, the share of people who say they can reach the learning resources they need dropped by 8 points.

The finding that deserves the most attention sits outside the table. PwC identifies 14% of the workforce as "front-runners," people who combine scarce skills with strong AI capability, and 29% of that group say they are very or extremely likely to change employers in the next year. These are the employees who can examine a new AI workflow and tell you whether it is producing reliable work, and nearly 3 in 10 of them are considering their options.

Pete Brown, PwC's Global Workforce Leader, described the strain directly: "Workers are under growing pressure from rising workloads, rapid change and external uncertainty." He was equally direct about where that strain leads.

"There is a real risk that the global workforce is starting to move at different speeds."

The learning debt is compounding

I call what PwC measured the learning debt. Every time a company deploys a new tool and postpones teaching people to use it well, it borrows against its own future. The obligation appears on no balance sheet, yet every leader is carrying some of it, and it compounds as improvised workarounds harden into habits and as the people who could have repaid it take their expertise to an employer willing to invest in them.

This survey completes a pattern I described earlier this year, when the St. Louis Fed found that 95% of AI productivity talk on earnings calls is still in the future tense. Computing contracts run for years, while retraining budgets are approved one year at a time and are frequently the first line eliminated when a quarter disappoints. PwC now shows that pattern from the employee's chair: more technology on the desk every year, and a shrinking share of people who feel equipped to use it.

Economic history has recorded this sequence before. American factories had access to electric power from the 1880s, yet the major productivity gains from electrification materialized only in the 1920s. The economist Paul David traced much of that delay to the factory floor itself, where plants had to be redesigned around the new motors and the people operating them had to learn an entirely different way of working. The machinery was ready decades before the organization was.

Your evaluators are the flight risk

In nearly 20 years at Cisco, where I shaped a $1.1B innovation portfolio, the scarcest resource on any initiative was the person who could look at a new system and determine, with evidence, whether it was working. We could purchase a platform in a quarter. Developing someone capable of judging it took years, and when that person departed, the judgment departed with them.

That is the cost concealed inside the 29%. When front-runners leave, the company keeps its licenses and loses its capacity to evaluate them, so when the next vendor presentation arrives, nobody remaining has the standing to challenge it. I described the same decision when AT&T and Verizon placed opposite bets on their workforces, with one planning for fewer employees and the other committing $70 million to free AI training for job seekers and displaced workers.

Software arrives on a purchase order. Judgment arrives on a calendar.

Where to start paying it down

If your own numbers resemble PwC's, you have considerable company, because most organizations purchased the technology first and planned the education second. A practical first step for Monday morning is to identify your front-runners by name before a recruiter does, then ask what they wish they had been taught this year and what would persuade them to keep building their careers with you. Their answers will reveal more about your AI program than any adoption dashboard.

A second step is to place two figures side by side in the same leadership meeting: what you spent on AI this year, and what you spent teaching people to use it. In my experience, few executive teams ever see those numbers on a single page, and seeing them together tends to settle the budget conversation faster than any presentation about transformation.

And if you are among the many workers using AI every day with little training behind you, this question belongs to you as well. Ask your manager what the company intends to teach you this year, and present it as a question about the quality of the work, because that is exactly what it is.

These decisions add up beyond any individual company. A workforce moving at different speeds, in Brown's phrase, eventually shows up in careers that stall in midstream and in regional economies where capable people concentrate wherever the training is funded. It is also where employers and public workforce programs have the most to offer each other, since both are trying to reach the same workers before their skills expire. How quickly people can relearn, and who pays for it, is the central question of my forthcoming book, The Great Relearning.

So here is a question to take into your next leadership meeting. Who are your front-runners, and when did someone last ask them what they need to learn next? If the answer is slow to arrive, you have located your learning debt, and it can be paid down starting this week. If you want to compare notes, I'm easy to find.

Here is what makes Alex a credible voice on this topic: Alex Goryachev shaped Cisco's $1.1B innovation portfolio and now advises the California State University system on AI governance, so he has seen firsthand what happens when an organization buys new technology faster than it trains the people who have to judge it.

Bring Alex in to help your leadership team measure and pay down its learning debt →