If artificial intelligence has left HR leaders feeling both excited and overwhelmed, they’re far from alone. Companies see enormous potential in the technology and are racing to prepare their workforces for what comes next. The challenge is that no one agrees on the best way to get there.

Dr. Ken Matos, the director of market insights at HiBob, spoke with PBS News correspondent Megan Thompson during a From Day One webinar on the matter. As Matos sees it, employers are still in the early stages of adopting AI.  Expectations are high, but plans are still vague, as businesses experiment with a wide range of approaches to AI use and employee training. “What’s problematic is that hole in the conversation where it’s a very confident ‘yes, we’re going to do this,’ and a very panicked, ‘I don’t know what doing this really means.’”

The hole shows up clearly in HiBob’s survey of 1,200 corporate decision makers, says Matos. Question after question, what emerges is a consensus that AI  is important and that workers must be prepared to use AI tools, but there is rarely any agreement about best approaches for use or training.  The survey shows that 75% of decision makers agree that even non-technical jobs will require at least moderate AI proficiency,  he says. 

But when it came to what that training actually looked like, there was little consensus. HiBob’s survey listed nine different approaches to upskilling employees, from external seminars to hands-on practice labs and sandbox experimentation, and adoption was remarkably even across them, ranging from 23% to 30%. In other words, most employers were trying two or three methods, but there was no clear playbook for which ones to use

There’s also a conundrum over just who should be in charge of that training.  Again, respondents came back with a wide range of answers, from managers to technology vendors to unions to public schools.  The leading answer, by a narrow margin, was direct supervisors. But only 36% of respondents thought that those same supervisors were qualified to do the teaching. “Managers of today were not hired because of AI skill,” Matos said, “and those who are promoted are not necessarily good managers who are able to teach.”

The AI Skills That Matter Most

So what skills should employers be developing? Matos says that adaptability tops the list. Until now, workers could count on computers to produce the same output from the same input. If the result was different, something had likely gone wrong. AI changes that equation. Even the engineers who build these models can't always predict how they'll respond. As models evolve behind the scenes, the same prompt can produce a different answer than it did yesterday. Employees need to be comfortable working with that uncertainty, adjusting to shifting systems and unexpected outcomes.

The second is understanding how AI itself works. This is more than just writing good prompts. It starts with knowing what tasks AI is good at and where humans need to take control. “AI is absorbing a lot of the explicit knowledge” Matos observes, referring to discrete information you might find in a book or a spreadsheet. But there is also tacit knowledge which involves experience and understanding people. AI will lack this sort of knowledge, he says. 

The third piece will be safety and ethics. “So much of AI usage is envisioned as individual AI people going forth and just doing things and spinning up vibe-coded apps and other applications,” Matos said, but they need to think about what data is in the system. Employees who use AI should pay attention to what data an application really needs, what is and isn’t safe, and what the consequences will be if there’s a breach.

Coaching For AI: Learning Together

If direct managers are going to be doing the bulk of the training, then companies should rethink just what training means. Matos says that they should adapt a coaching mindset instead. He encourages managers to guide and learn with their subordinates. “You’re not supposed to go and tell them what to do or how to do it because you know it better. You’re there as a critical thinking sounding board to reflect back what they’re saying, to ask them questions.”

With that in mind, employees may use AI to generate a report, but they shouldn’t send it to a supervisor until they understand what it’s saying. And rather than acting on every insight AI produces, teams should identify one or two key findings, verify them, and build from there.

AI will change workflows and processes, but Matos recommends that this process not be rushed into.  His first step?  “Map your workflows, don’t even worry about AI yet,” he said. Once you have your actual processes documented, “then you can start saying, ‘Where does AI actually help make this faster, smoother, more integrated?’”

“One of the challenges with AI is that no one really understands AI,” Matos said. The uncomfortable truth is that a lot is still unknown about the functions and capabilities of this technology. Accepting that will mean proceeding a bit more slowly with adopting artificial intelligence, but it will free managers to explore the technology along with the rest of their teams.  The companies that commit themselves to really learning this technology, balancing bold experimentation and methodical application, will be best positioned to thrive.

Editor’s note: From Day One thanks our partner, HiBob, for sponsoring this webinar. 

Paul Kersey is a former attorney and freelance writer who has covered events for Bloomberg News and other outlets.  Paul is based in Chicago, IL.

(Photo by Kindamorphic/iStock)