Most employees use AI at work, but organizational value still lags. In this article, we look at what the latest research says about AI adoption and the related skills gap, and discuss L&D's role in leading meaningful change.

If you've been in the workforce for a little while, you may have seen this pattern before: a new technology, with promises of limitless change for the better. The folks who make budget decisions are excited. The direction is set. The goal: better performance. The promise of wins. And yet, the reality of the implementation (which we often forget through the rose-colored lenses of hindsight) is a bit harder, slower, and yes, costlier, than originally promised. Yet we navigate and embrace it, and the organizations that do it well are rewarded.
AI is following very much the same script. The Human-Centered AI unit at Stanford says generative AI reached adoption by more than half the population within three years, which is faster than the personal computer or the internet (Stanford HAI 2026 AI Index). And, particularly in the U.S., those adoption trends seem to be stronger within organizations: about 3 in 4 frontline workers now say they're regular users of AI (BCG AI at Work 2026).
But in short: we've got the tools. People are using them (though how and when remain issues to address). And in most cases, we're not yet seeing the payoffs that were promised. Budgets, meanwhile, are still following the tools: Training Industry's Trends 2027 report finds that adopting new technology or tools is the top reason learning budgets are rising next year, even as investment in course development shrinks.
Consider these two statistics from the latest McKinsey AI report. 44% of companies say they're scaling AI across the enterprise. And 37% of companies say it's affecting the business, with an actual effect on earnings. In other words, more than a third of respondents are seeing AI working for them in a measurable way.
Buried in the overall usage numbers - how many companies are using it or aren't, or how many employees are using it or aren't - is the question of how deeply AI has been embedded in workflows. At present, where workers are using AI, about two-thirds "restrict use to three or fewer tasks," generally for task support (NBER, 2026).
It's a pattern we've been tracking for at least the past year: broad usage for task support, but in a way that hasn't yet added up to enterprise performance. (As AXIOM's Jason Ruff said when we launched our AI Center of Excellence: "People everywhere are using AI, but use of the tools doesn't necessarily create value.")
And that's the gap L&D leaders should be addressing in planning, strategy, and budget discussions. When 88% of respondents say they will need "major" upskilling in the next five years, and only 36% feel properly trained, the challenge is clearly right ahead of us, if not here already. Nearly 3 in 4 say the skills expected in their current role have already changed.
Deloitte’s 2026 Global Human Capital Trends survey describes the same gap from the organization’s side: just 8% say their organizations are highly effective at meeting the continuous, “always-on” learning needs of their workforce.
That gap shows up inside L&D, too. In our own Pulse of the L&D Practice 2026 survey at the start of the year, investment in AI was the strongest single signal in the data, yet L&D teams are not necessarily receiving their own skills development around AI. Learning teams are being asked to lead workforce readiness with an expectation AI will help them do more with less, often without the necessary support in managing their own skills and workload capacity.
The first AI training tended to focus on prompt writing or "prompt engineering." Those skills are still important, but as a foundation for building another level of judgment-based and decision-making skills. Rather than AI doing jobs for people, it's about managing and directing the AI. The core skills that are needed include "contextual judgment, framing problems, and interpreting results." (BCG).
It's an idea that's increasingly setting the expectation for how organizations train their teams to use AI tools. In our recent podcast discussion, Mateland Mayes put it this way: "do not hand your thinking over to it."
The benefit from AI tends to be seen by organizations which are putting in the effort to change how work gets done. Almost three-quarters of AI high performers have fundamentally redesigned workflows with AI use in mind (McKinsey). Training Industry's Trends 2027 report calls this an "AI-native" approach: rather than adding AI to an existing process, which may make it faster without making it better, ask how you'd design that process today knowing what AI can do. L&D can ask the same question of its own work. Some of the most impactful AI learning has come from immersive experiences, where teams are brought together to explore how work processes and decision making can be rebuilt: this can be built into an "escape-room-like" environment where every input and decision can be evaluated to surface learnings and changes which even the team involved might not see in real time. It's using AI to transform how AI is used. And that type of intentional learning is important to overall success, according to two-thirds of organizational leaders surveyed by Deloitte.
Work redesign isn’t solely a learning task, but it creates learning needs at every step: new roles, new handoffs, new decision rights, and new expectations of managers.
Among BCG respondents who regularly use AI and have heard of AI agents, half (50%) say their company has not put clear guidance in place for managing teams of humans and AI. And, as we discussed recently on the podcast, controls on the use of AI may not be created by L&D, but do need behavioral support and reinforcement throughout the enterprise, and that's squarely L&D's role.
Taken together, the state of enterprise AI adoption outlines a core agenda for L&D:
• Start with the current state. Understand where AI is being used, by whom, and with what skills.
• Prioritize by business value. Rather than boiling the ocean, focus on where opportunities or risks are most pronounced. Both will be where learning interventions that support behavior change will make a measurable difference.
• Enable in the flow of work. Integrating new skills in the flow of work is not a new concept for L&D, and its relevance in the AI skills era is reiterated by the latest BCG survey.
• Measure outcomes. When you've developed clarity about the business value expected from AI implementation and the skills to enable it, these become the meaningful metrics to use. AI adoption is the attendance log of the era: it's a "butts in seats" measure, easily obtained but not meaningful. Training Industry's Trends 2027 report notes that more organizations are prioritizing better capability metrics, because completion rates and satisfaction scores show who took part, not whether people can do the job. Defining what good performance looks like is work to do with business stakeholders before training begins. Measuring outcomes begins to help you and your organization understand what's paying off, and what isn't.
Many of these ideas fed into the creation of the AXIOM AI Center of Excellence, with the recognition that every company's AI journey and needs are different. We developed not a solution, but a broad menu of services, to mobilize the right combination of assessment, strategy, learning content, training delivery, and embedded expertise to help organizations make focused, successful AI adoption a reality.
Whether you need custom content, flexible L&D staff, or trainers and facilitators, we can help you plan the right approach.
Contact Us