
Workplace AI Is Outrunning Training, and the Learning Debt Is Growing
Less than 40% of workers in APA's 2026 survey reported knowing whether their employer had an AI-use policy, and LSE research with Protiviti found trained employees adopt AI at 93% against 50% for the untrained.
Key Takeways
- Less than 40% of workers for whom the question applied reported knowing whether their employer had an official policy about acceptable AI use, according to the American Psychological Association's 2026 Work in America Survey.
- In the AI at work study from The Inclusion Initiative at LSE and Protiviti, employees with AI training used AI at a 93% rate against 50% for untrained employees, and saved 22% of their working hours against 13%.
- Gallup's August 2026 survey found that 33% of frequent AI users feel pressure to use AI to keep up with expectations, compared with 17% of infrequent users.
- The learning debt is deferred retraining treated as a compounding liability, and every AI rollout that skips training and a clear acceptable-use policy adds to it.
Bring Alex to your organization: keynotes, workshops and advisory, in person or virtual. Check Alex's availability
Less than 40% of workers for whom the question applied reported knowing whether their employer had an official policy about acceptable AI use, according to the American Psychological Association's 2026 Work in America Survey. In a separate study from The Inclusion Initiative at the London School of Economics, 74% of professional employees now use AI at work, up from 29% in 2024. The tools have spread through the workforce faster than the training and the rules that make them pay, and every month of that delay adds to a liability I call the learning debt.
APA released that figure on November 16 in new guidance for employers, Deploying AI Into Workplaces: What Leaders Need to Know, which sets out 7 best practices. Two of those practices speak directly to the policy problem: leaders should introduce AI through a fair process that consults the people who will use it and the people affected by its outputs, and they should protect opportunities for people to build deep expertise, share knowledge, and learn from experience.
Dennis P. Stolle, JD, PhD, Executive Lead Psychologist for Applied Psychology at APA, described the pattern in a single sentence.
"Purchasing an AI tool and telling people to use it without further direction is not a sound strategy."
I have watched that pattern for years, first while building innovation programs inside a large company and now in the keynotes and advisory work I do with enterprise leaders. The license arrives first and the announcement email follows, while the training lands sometime after the next budget cycle, if it lands at all.
Training decides who picks up the tools
The LSE study, titled AI at work and run by The Inclusion Initiative at LSE in collaboration with Protiviti, surveyed 3,001 professional employees from March to May 2026. Protiviti is a consulting firm, and I found no independent coverage of the results, so I weigh them as one vendor-linked source until other research tests them. The pattern it reports on training still deserves a careful look.
| Measure | Employees with AI training | Employees without AI training |
|---|---|---|
| Use AI at work | 93% | 50% |
| Share of working hours saved | 22% | 13% |
Employees with training adopted AI at nearly twice the rate of those without it, and they saved 22% of their working hours against 13% for the untrained group. That difference makes the case for a training budget in the one currency every executive counts, which is time. Dr Grace Lordan, founding director of The Inclusion Initiative at LSE, added the caveat that finance teams should hear first: "Where those time savings go matters as much as the size of the savings." An hour saved and spent on more of the same work disappears into the week, while an hour saved and spent on learning starts to pay down what the organization owes.
The pressure runs toward more use
The same study found that 48% of employees worry about being judged for not using AI enough, while 20% worry about being judged for using it. Gallup's August 2026 survey of US employees, published November 16 in an analysis by Andy Kemp, points the same way. In that survey, 53% of US employees use AI in their role at least a few times a year, and 16% of all US employees use it daily. Among frequent users, 33% feel pressure to use AI to keep up with expectations, compared with 17% of infrequent users.
Watch any analyst handed a new AI assistant with no written guidance and a manager who keeps asking how much she uses it. She will use it, because the fear of being judged runs in that direction. She will also guess at the rules, because no one has shown her where they live, and each guess is a small risk the company has taken on without ever deciding to.
Pressure without training produces use without judgment, and the adoption numbers climb while leaders read the climb as progress.
Deferred training compounds into a learning debt
The learning debt is deferred retraining counted as what it is, a liability that compounds over time. Every leader carries it, whether the balance sheet records it or leaves it off. The principal is the training the workforce needed on the day the tools arrived. The interest is what accrues while that training waits: habits formed by guesswork, and junior employees who skip the slow work of building deep expertise because a tool hands them a passable draft in seconds.
Software budgets buy the tool. Training budgets buy the return.
The argument at the center of my forthcoming book, The Great Relearning, is that the capacity to relearn has become the most durable asset a company or a career can hold. Workplace AI puts that argument on a quarterly schedule, and each rollout that skips the training borrows against it.
If this describes your organization, you are in common company; your competitors are carrying the same debt, and by APA's count most workers in its survey cannot say whether their employer has a policy for tools they already use. A debt that everyone holds is a debt anyone can start paying down this quarter.
Repayment starts with the people using the tools
APA's guidance points to where repayment begins, with the people who will use the tools and the people affected by what those tools produce. These are the questions I am taking into my own meetings with executive teams, and you are welcome to borrow them. Can every employee with an AI license tell you, in one sentence, what your acceptable-use policy says? Who did you consult before the last rollout? Where did the saved hours go last quarter? What share of this year's AI budget pays for training? Those questions run through all of my writing on AI leadership.
Employees have a move of their own: if you use AI at work and have never seen the policy, ask your manager for it in writing, because a clear answer protects you both.
I hold the Innovator-in-Residence role at Tulane University and spent 20 years at Cisco, so I look at the learning debt from the side that teaches and the side that deploys. I'm easy to find if your leadership team is working out how to pay down its own.
Less than 40% of the workers in APA's survey reported knowing whether their employer had a policy for the AI tools on their screens. One clear policy, and the training behind it, will move that number faster than the next license will.
How should employers match AI safeguards to risk, according to the APA?
The American Psychological Association's guidance, Deploying AI Into Workplaces: What Leaders Need to Know, tells employers to match safeguards to the stakes of potential errors. In practice, a tool that drafts internal notes can run with light review, while a tool that shapes decisions about people warrants closer human checks. Dennis P. Stolle of APA said organizations need to invest in the people who will work with the technology alongside the technology itself.
How reliable is the LSE and Protiviti AI at work study?
The AI at work study was produced by The Inclusion Initiative at the London School of Economics in collaboration with Protiviti and released on November 17, 2026, through PR Newswire UK. The release reports 3,001 professional employees surveyed and leaves out which countries they worked in. With a consulting firm involved and no independent coverage found, its figures work best as directional evidence that leaders test against their own employee data.
How can a company measure its learning debt from AI?
A company's learning debt from AI can be estimated by comparing two counts: employees who use AI tools at work, and employees who have received training and can state the acceptable-use policy. The distance between those two counts is the outstanding balance. Running the same short employee survey each quarter shows whether the debt is shrinking, which happens when the trained share grows faster than the licensed share.
Here is what makes Alex a credible voice on this topic: Alex Goryachev is Innovator-in-Residence at Tulane University and spent 20 years at Cisco, where he shaped a $1.1B innovation portfolio, so he looks at workplace AI rollouts from the side that builds skills and the side that deploys the tools.
Invite Alex to help your leadership team pay down its learning debt →
