The AI Workforce Training Paradox
Traditional worker retraining programs have flopped, but we desperately need AI upskilling. The solution? Focus on foundational tech literacy instead of specific job skills.
Day and every day to follow will mark the lowest rate of AI adoption. The pace (how many people use AI) and depth (how extensively people use AI) of that adoption is very much an open question. This chart from Alexander Bick, Adam Blandin, and David J. Deming is packed full of information provides us with some initial answers as well as some key policy points.
First, it’s important to note that the VAST majority of ALL workers do not regularly use AI. Perhaps misleadingly, the chart omits a section on the workers who rarely use or have never used AI. If that had been shown, then it would be clear that AI adoption in the workplace (as of August 2024) was still fairly low. More than 65 percent of all workers under 30 do NOT regularly use AI at the office. That rate jumps to more than 80 percent for workers ages 50 and above. The low rate of adoption offers both positive and negative news.
On the positive front, it means there’s still an opportunity to help all workers learn how best to use AI. In other words, the low rate of adoption may suggest there’s an incentive for employers to offer company-wide AI training rather than to just a handful of early adopters of the technology. If offered, those broad trainings will diffuse AI knowledge and skills throughout the office. In turn, employees should see a productivity boost that will, of course, also benefit the company. Such training will also increase human capital across the workforce. Employees with AI knowledge will have better odds of landing their next job—one that aligns with their interests and their income needs/expectations.
On the negative side, the low rate of adoption makes clear that we’re missing out on realizing the full benefits of AI each day that adoption remains relatively sparse. Low pickup also indicates that the sort of mass training described above has yet to become common or, if provided, has yet to change worker habits. Either of those realities is troubling.
A lack of training will put a cap on the future employment prospects of employees. Future jobs will go to those with future skills. If employers do not invest in their current workers, they opt to simply replace them with younger, more tech-savvy people. Inadequate training poses a different problem. No one likes sitting through (or clicking through) some corporate-mandated video on a new tool or new procedure. If that’s the model for sharing AI knowledge then workers will not only fail to learn how to use AI but also become skeptical of AI as well as future training opportunities.
Second, disparate rates of AI adoption among different age cohorts shows that now’s the time to rapidly increase AI literacy and know-how among older workers. Millennials and Gen Z will have a comparatively easier time adjusting to new AI tools given their digital childhoods. Just as these researchers saw increased AI use among younger Americans, a University of Chicago survey determined that “[h]alf of adults aged 18-59 and 45 percent of teenagers (aged 13-17) have never used AI. Eighty percent of baby boomers (aged 60+) have not used AI.” The ease with which younger workers pick up new AI tools will give them a huge leg up in the Age of AI. Preventing mass layoffs and the resulting social malaise, its crucial that workers of all backgrounds get up to speed on AI. Traditional upskilling programs are not the answer. Don’t take it from me—here’s a great summary from The Atlantic:
In 2022, the U.S. Department of Labor published a comprehensive study of the Workforce Innovation and Opportunity Act (WIOA) and a host of similarly structured federal job-training initiatives. The programs did manage to put a lot of people through training, the researchers found. And many of those people were then hired in so-called in-demand jobs. But in the first three years after training, their wages increased only 6 percent compared with those of similar workers who didn’t receive training—from an average of about $16,300 to $17,300 a year—and the effect didn’t last. In the long term, their relative wages didn’t increase at all.
This poor track record is often attributed to ever-growing skill requirements for jobs in the fast-paced global economy. In fact, the programs fail because they’re designed with potential employers rather than employees in mind. In the case of the WIOA, the local workforce boards that decide which jobs qualify as “in-demand,” and therefore which are eligible for federal funding, are dominated by business interests—and what business wants is a steady stream of low-wage workers trained by someone else.
Herein lies the training paradox—we need to retrain workers (see point one) and worker retraining programs have traditionally flopped (see point two). My hunch—one I plan to write on in more detail in the future—is that we need worker retraining to focus less on specific jobs and more on foundational AI and tech literacy. We cannot predict what the jobs of the future will look like. Few folks in the 1980s would have guessed that software engineers would have been the most sought after employees just a few decades later. We will also struggle to predict what roles AI will create and what specific tasks they will involve. The best approach then is focusing on the fundamentals—a focus that will regular ongoing and substantive training that places the worker first.
Third, this may be (and should be) a key policy area for the Trump administration. Vice President J.D. Vance delivered the equivalent of a mic-drop address at the Paris AI Summit. He specified that this administration will prioritize a pro-worker AI agenda. That’s a tall order. As the adoption rates make clear, we have a long ways to go toward realizing that goal. That said, I’m thrilled the administration is aware of the need to pair AI advances with advances in worker capacity. AI progress does not have to be zero-sum.
The last word:
The data on AI workplace adoption reveals both challenges and opportunities ahead. While current adoption rates remain surprisingly low across all age groups, this presents a unique window for meaningful intervention. Companies that invest in comprehensive, worker-centered AI training programs now will position both their businesses and employees for future success. However, time is of the essence. The traditional approaches to worker retraining have proven inadequate, suggesting we need a fundamental shift in how we prepare our workforce for the AI era. By focusing on broad technological literacy rather than narrow skills, we can build a more resilient and adaptable workforce.
Big thanks to Alexander, Adam, and David for conducting this research! Read their full paper here.