Our Take
Employers are moving slower on AI governance than employees are moving on AI adoption, and that gap is now a data-loss and compliance liability, not a training problem.
Why it matters
Workers are logging nearly four hours a week with AI tools they taught themselves to use, entering unknown data into unvetted systems. When adoption outpaces governance by this margin, organizations have visibility into neither what data is flowing in nor how outputs are being used.
Do this week
Chief People Officer or VP of Learning: Audit how your workforce is actually using AI this week (shadow 10 random employees for an hour) so you can map unvetted tools before compliance discovers them.
Most workers are self-teaching AI while employers hold back
Nexthink, which monitors IT usage across 3.4 million employees, found that most U.S. workers using AI on the job are not getting formal training from their employer (company-reported). Instead, they are learning from social media, news articles, and conversations with friends. The same employees average 10 interactions a day and nearly four hours a week using these tools, saving roughly the same amount of time in return.
The training gap is severe. Research from Jobs for the Future found that 56% of workers say their employer has never consulted them on how AI tools are used in their work. Among workers who consider AI training important, nearly 6 in 10 are not being offered formal guidance.
Separately, Forrester research found that many organizations announcing AI-related layoffs do not have mature, vetted applications ready to replace the roles they are cutting, suggesting companies are moving faster on workforce decisions than on workforce preparation (analyst report).
The compliance and visibility problem is real
When employees learn AI practices from unvetted sources, organizations lose visibility into how tools are being used, what data is being entered into them, and whether outputs are being applied appropriately. This creates exposure across compliance, security, and performance management.
Liz Raymond, VP of Global Talent at Nexthink, framed it plainly: "When adoption outpaces training and governance by this margin, organizations have no clear path to AI value."
The survey data reinforces the tension. Both Gallup and Jobs for the Future found that although employees are experimenting with AI on their own, they want structured guidance, guardrails, and a say in how these tools reshape their jobs. Forrester projects that AI will augment 20% of jobs over the next five years, amplifying the stakes (analyst projection).
Build governance before tool adoption spreads further
HR leaders report that the most common misconception is that AI transformation begins and ends with tool adoption. In reality, successful transformation starts with understanding how employees are actually using AI, then layering in sustained investment in skills, governance, and employee experience.
Amy Mosher, chief people officer at isolved, defines AI fluency as employees knowing "how AI fits into their role, how it can enhance their work and how to use it responsibly" and practicing when to trust, question, and override AI outputs. That requires seeing the current state first.
The immediate task is not more training courses. It is visibility. Audit which tools are already in use, which data is flowing into them, and who needs to sign off before outputs are deployed. Only then can you design governance and training that workers will actually follow, rather than route around.