AI fluency is fast becoming an organizational imperative. Yet too many companies emphasize the wrong objectives, view training as one-time events, or fail to tailor teachings to specific roles or departments.
Harvard Business School CIO Beth Clark sees no reason for most workers to understand how large language models work. They should, however, have enough knowledge of how models take in data, use that data, produce outputs, and transform workplace operations.
Clark acknowledges that imparting that knowledge to HBS employees is a unique and challenging task: Once IT taught workers how to use new software systems to do their jobs, now it has to train workers how to use AI to work in new ways.
“That,” Clark says, “is different than previous iterations of upskilling.”
That dynamic explains, at least in part, why so many organizations are failing to effectively train people for the AI revolution under way.
According to the 2026 Annual L&D Benchmark Report from TalentLMS, an employee training platform, only 64% of employees believe their companies actively support people in learning how to use AI. In terms of company-directed upskilling efforts, professional services firm Accenture has found that only 26% of workers report receiving training on how to collaborate with AI.
“We’re all guilty of saying we need upskilling or reskilling and stopping there. The question is now what, exactly, do we need to do and how do we do it,” says Sam Ransbotham, co-author of the recent report The Emerging Agentic Enterprise: How Leaders Must Navigate a New Age of AI from MIT Sloan Management Review and Boston Consulting Group.
Where upskilling goes wrong
There are multiple areas where organizations fall short when it comes to upskilling employees for AI, Ransbotham and others say.
Besides failing to train them at all (as some studies have found), organizations also often fail to train workers on how these models work, says Ransbotham, who is also a professor of analytics at Boston College’s Carroll School of Management and a guest editor for MIT Sloan Management Review’s Artificial Intelligence and Business Strategy Big Ideas initiative.
Granted, most employees won’t work in constructing models, he acknowledges, so they don’t need to become AI engineers. But they should have a solid understanding of how models work so they know when AI is not producing reliable results and when to question outputs.
“A shallow understanding of how AI works is appropriate, because you cannot be a good consumer of AI without knowing how this thing is made,” he adds.
Still, there’s no standard for the depth of instruction all workers need to have, Ransbotham says. “It would be nice if I could say, ‘Teach X,’ but the answer to how much knowledge is needed depends on the job, the context,” he adds.
That said, Ransbotham also sees organizations overemphasizing the technical components and infrastructure requirements when upskilling teams on AI. In such cases, they get too far into the technical weeds and fail to address how workers can innovate in using AI for their jobs — which is what upskilling should be doing.
“[Organizations should] get workers to rethink why they have a process at all and how to use AI to change the workflow, instead of teaching them to tweak around the edges. Teach them how to innovate, how to question things, to identify what’s not working well,” he explains.
Treating AI training as one-off events rather than a journey of continuous learning, refusing to allocate time and money to training, and taking a one-size-fits-all approach are other common upskilling errors.
“Common mistakes include making broad generalizations as to what all user groups need and being overly general in the training that’s provided, except when it comes to fluency,” says Katie Fitzgerald, a managing director at digital services consulting firm West Monroe.
Additionally, organizations often overlook the value of practice and exploration as a critical part of AI training, she says.
Furthermore, organizations that fail to “provide psychological safety” as workers feel overwhelmed by AI and fear they’ll be replaced by the technology also find their training doesn’t get the traction they’re seeking, says Peter Miscovich, executive managing director and global future of work leader at JLL, a global commercial real estate and investment management company. IT leaders must work toward achieving that safety as part of their upskilling plans.
A survey of workers conducted by Pew Research Center in October 2024 speaks to such sentiments, noting 52% of workers are worried about the future impact of AI use in the workplace, 32% believe AI will lead to fewer job opportunities for them in the long run, and 33% feel overwhelmed.
Training as a business differentiator
For Ransbotham, the investment in effective upskilling is worth it. Workers who know how to leverage AI can help transform the organization, he says. As such, organizations should aim to train workers how to excel in using AI. “The ability to use these tools better is what differentiates,” he adds.
CIOs are well positioned to lead such upskilling efforts throughout the enterprise, he and others say, because CIOs have a long history of educating and training workers on technology adoption and transformation.
Doug Gilbert, CIO and chief digital officer at Sutherland, which provides digital transformation and business processes services, agrees. CIOs also have strong motivation for taking the lead, he adds, as research has shown a high percentage of AI initiatives are failing. CIOs and organizations that invest more in upskilling their workers can turn those figures around, he contends.
Such findings have indeed motivated Gilbert and his Sutherland colleagues, who earlier this year recognized the need to ramp up their approach to upskilling employees for the AI era.
He acknowledges that some of the company’s past training approaches, such as using training videos that were for many workers just a check-the-box activity, weren’t delivering the desired results. So he worked with his company’s HR and training leaders to relaunch an upskilling program.
Now workers at all levels must learn foundational AI skills and demonstrate they’ve mastered AI fluency. Those who do earn a certification. Sutherland employees, including leaders, then take training courses tailored to their roles and departmental needs — earning more certifications and satisfying personnel performance goals as they go.
The company still uses videos, but it also has real-life training and hands-on activities within a sandbox environment to give workers opportunities to practice their new skills, Gilbert says. Training is ongoing.
Gilbert believes this combination ensures workers learn what they need to know about AI to work better, continue to learn as the technology evolves, and use AI in ways that deliver gains for the organization.
Anatomy of an effective upskilling approach
Others endorse the techniques Gilbert and his company are using to ready the workforce for the AI era.
For example, Ransbotham also promotes a tailored approach. He advises CIOs and their executive colleagues to take an inventory of skills, identify what skills are weak or lacking, and then train for those missing skills. He also recommends training workers on AI literacy and coding, too, as coding will help people make more use of AI tools. And he recommends enterprise leaders foster a culture where workers can try new ways of working using AI, even if they don’t always succeed, as experimentation is a good way to learn.
Miscovich suggests a similar approach to upskilling. He endorses training tailored to specific roles, hands-on sessions, and continuous training so that workers can keep pace with AI advancements. He also stresses the need for upskilling executives in parallel with the rest of the workforce so that the organizational leadership is equally engaged in preparing for what’s ahead.
“It’s a compete reeducation of everyone continually. That’s new for most organizations,” Miscovich says, adding that organizations should track upskilling progress, recognize successes in learning, and tie successful upskilling to personnel performance objectives.
All this can create a culture of learning, he and others say.
“The learning that I have seen be the most effective is at organizations where there’s a culture of curiosity, exploration, use case sharing, and formal learning programs,” Fitzgerald adds.
CIOs delivering upskilling wins
Clark, the Harvard Business School CIO, is one IT leader who has seen the benefits of upgraded upskilling.
She cites her side-by-side work with an HBS director to create a gen AI-fueled search tool. As part of the hands-on lessons, the director came to understand the technology’s flexibility and usefulness and learned how she could use it to work differently. Clark says it shows how teaching-by-doing inspires workers.
“They come to an appreciation that this technology is vastly different, and they see that they can use it to think about doing their jobs in lots of different ways,” she adds.
With such insight in mind, HBS in 2025 launched an immersive, case-based, experiential AI Academy that is a mandatory program for faculty and staff on campus.
Others are taking a similar multimodal approach.
Ravi Pendse, vice president for IT and CIO at the University of Michigan, believes gen AI will be the most impactful technology of this century and, if used responsibility, ethically, and thoughtfully, will be a “force for positive.”
Pendse is aware of the disruptions AI will bring to the workforce, citing the World Economic Forum’s 2025 Future of Jobs Report that estimates that AI will displace 92 million jobs but create 170 million more by 2030. Moreover, he says, AI will touch all types of jobs and workflows, even if the roles themselves don’t go away.
“So even though we’re looking at net positive for jobs, there will still be displacement,” Pendse says, which is a big reason why “we feel upskilling is needed.”
It’s also part of the reason why Pendse focuses “not on job security but career security.”
Pendse, who as CIO helps lead the university’s upskilling efforts, says he conveys that message to individuals throughout the workforce, stressing that they should commit to upskilling and that the university will provide training, too. “It’s about teams coming together and learning together,” he says.
To bolster AI skills among its employees, the university has weekly challenges, regular weekly innovation times, learning forums for professional community clusters, and article-sharing. It also offers in-person training courses Monday through Thursday as well as drop-in office hours with emerging technology team members.
To further ensure that its AI training is effective, Pendse says the university uses multimodal lessons and creates them quickly to prevent the information from being outdated.
Pendse also seeks training that builds the skills workers will need tomorrow, not just what they need today. And, like others, he wants programs that deliver “just-in-time continuous training.”
The goal, he says, is for every individual to understand the impact on AI on their work and know how to use AI tools effectively to do everything, from reimaging workflows to solving what were once intractable problems.
SUBSCRIBE TO OUR NEWSLETTER
From our editors straight to your inbox
Get started by entering your email address below.
