This is about the same for many who push Gen AI in schools but offer no time for learning,  no professional development.  It’s the old, “The tech is so easy, why haven’t you spent time learning it on your own time?”

Human adoption is a top challenge for the “great uncertain terrain” of AI, Lumen CEO says Workforce solutions firm ManpowerGroup’s 2026 Global Talent Barometer found that worker relationships with AI have grown complicated. The study—based on interviews with nearly 14,000 workers across 19 countries—found that while regular AI usage jumped among workers by 13% in 2025, confidence in the technology’s use plummeted by 18%. 

“AI adoption is accelerating, but confidence is collapsing,” Mara Stefan, VP of global insights at ManpowerGroup, told Fortune. “Workers are being handed tools without training, context, or support.” (source)

It takes a tremendous amount of time to learn how to use AI in a way that ensures it will be reliable. That is, to make its predictions accurate to what you expect to see.

The main way is to use Retrieval-Augmented Generation (RAG) of the kind you see in NotebookLM or when you align Custom instructions with relevant source files. No one wants to cover that. Why? The initial workflow clarification process is time-consuming and difficult for folks who just do stuff but don’t think it through.

Once you understand this, learn to use your AI of choice, the work gets easier.  You know what to expect.

Confidence in AI use by workers plummet  because they are expecting a magic box to do their brain work. In time, if they stick with it, they get better.