AI training is everywhere right now. Most enterprises are funding it, talking about it, and building it into internal learning programs. On the surface, that sounds like progress.
But the numbers tell a more uncomfortable story.
DataCamp’s 2026 research found that 82% of enterprises offer some form of AI training, yet 59% of enterprise leaders still say their organisation has an AI skills gap. At the same time, only 35% say they have a mature, organisation-wide AI upskilling program.
That is the paradox.
Enterprises are training. Employees are getting access. Budgets are being spent. And still, a large share of organisations do not feel genuinely AI-ready.
Key Points
- AI training is widespread, but the gap remains. DataCamp found that 82% of enterprises offer some kind of AI training, yet 59% still report an AI skills gap.
- Access is not the same as capability. The same research found that 68% provide access to AI learning resources and 46% provide basic AI literacy training, but that still does not guarantee real workplace confidence.
- The problem is often the design of training, not the existence of it. Leaders say passive formats, lack of hands-on work, and weak role relevance are major reasons training does not stick.
- Frontline adoption is still lagging. BCG found that regular generative AI use among frontline employees has stalled at 51%, and only one-third of employees say they have been properly trained.
- Better capability building improves business results. Among organisations with mature, workforce-wide AI literacy programs, reports of significant AI ROI nearly double to 42%, while reported lack of ROI drops to 11%.
- Training alone is not enough. Microsoft WorkLab found that nearly 80% of organisations say they cannot share data across teams in ways that make agentic AI work, and only 22% strongly agree they have documented key processes and data dependencies.
The real problem is not a lack of training
The easy explanation would be to say enterprises are simply not doing enough. But that is not quite true.
Many organisations have already moved past the “should we train people?” stage. DataCamp found that 68% say employees have access to AI learning resources, and 46% say they already provide basic AI literacy training. The issue is that access alone is not translating into practical, consistent workforce capability.
That difference matters.
A company can run a successful AI awareness campaign and still have teams that do not know how to use AI well in real work. People may understand the language of prompts, models, copilots, or automation, but still hesitate when it comes to applying AI to client work, internal reporting, analysis, compliance reviews, or operational tasks. That is where the gap lives. It is less about exposure and more about usable judgment.
Why training is not turning into capability
This is where the story gets more interesting.
DataCamp’s findings suggest the problem is not that enterprises are ignoring AI learning. It is that many are designing it badly for the way work actually happens. The research points to three recurring issues: passive learning, low role relevance, and lack of reinforcement over time.
The first issue is format. Video-based courses and blended online sessions are the most common training methods, but leaders say they often fall short. In DataCamp’s findings, 23% say video-based learning makes it difficult to apply skills in the real world, and 24% cite a lack of hands-on projects or labs. That creates awareness without confidence. People understand concepts, but they do not get enough practice using them.
The second issue is relevance. Roughly three in five leaders report challenges with third-party online AI training, including learning paths that are not tailored to specific roles and employees not knowing where to start. That means people may complete a course and still not know how AI should fit into their actual function.
The third issue is progression. Many organisations provide AI learning resources without structured pathways that build capability over time. DataCamp puts it plainly: AI literacy is not a one-off competency. It needs repetition, feedback, contextual reinforcement, and measurable development.
That is why so many learning programs feel busy but still underpowered. They inform people, but they do not always prepare them.
The gap is bigger than technical hiring
A lot of executives still hear “AI skills gap” and assume the issue is mainly about hiring specialists.
But that is only part of the picture, and often not the biggest part.
DataCamp’s 2026 analysis says the AI skills gap is not primarily about advanced engineering expertise. It shows up in more foundational capabilities: evaluating whether AI outputs are accurate or misleading, applying AI tools to specific workflows, translating AI-generated insights into decisions, and understanding governance, risk, and responsible AI use.
That is an important shift.
The gap is not just about whether you have enough machine learning engineers. It is about whether your broader workforce knows how to use AI sensibly, safely, and effectively in day-to-day work. In many organisations, that is the missing layer. The tools are present. The awareness is present. But the applied literacy is still uneven.
Frontline reality tells the truth
The leadership view is only one part of the story. The frontline view often tells you whether adoption is real.
BCG’s 2025 AI at Work research found that while more than three-quarters of leaders and managers say they use generative AI several times a week, regular use among frontline employees has stalled at 51%. It also found that only one-third of employees say they have been properly trained.
That gap matters because enterprise value is not created only in executive discussions or strategy decks. It is created in day-to-day execution.
If senior leaders are comfortable with AI but frontline teams are still unsure, inconsistent, or undertrained, then the organisation may look more mature than it really is. It may appear AI-enabled at the top while remaining fragile in the parts of the business where most work actually gets done. This is an inference from BCG’s finding that usage and training confidence are materially weaker among frontline employees.
Why more content will not solve this
This is the point many enterprises need to hear clearly.
The answer is not automatically “more training.”
If the model is weak, scaling it just spreads the weakness further. More webinars, more videos, more generic learning modules, and more platform access can all create the appearance of momentum without solving the real problem. DataCamp’s findings suggest that what matters is not training volume, but learning design.
There is a strong business signal here too. DataCamp found that only 21% of leaders overall report significant positive ROI from AI investments. But among organisations with a mature, workforce-wide AI literacy upskilling program, that figure rises to 42%, while reports of no positive ROI fall to 11%.
That tells a bigger story than training alone.
Better capability building is not just a people-development issue. It is directly connected to whether AI investments produce results.
The skills gap is often a readiness gap in disguise
Training does not happen in a vacuum.
Even a strong learning program will struggle if the rest of the organisation is not ready to support AI-enabled work. Microsoft WorkLab’s reporting on agent readiness makes that clear. It found that nearly 80% of organisations say they cannot share data across teams in ways that make agentic AI work, and two-thirds lack executive champions willing to clear the path. It also found that only 22% strongly agree that their organisation has documented key processes and data dependencies.
That changes how we should think about the problem.
In many enterprises, the so-called skills gap is mixed with a workflow gap, a governance gap, and a readiness gap. Employees may not be underperforming because they failed a course. They may be underperforming because the data is fragmented, the processes are unclear, the ownership is vague, and the use cases are still disconnected from how work is actually organised.
Training matters. But without clarity, support, and usable systems around it, training cannot carry the full weight of transformation.
What better looks like
The organisations making genuine progress tend to shift the question.
Instead of asking, “How do we train more people on AI?” they ask, “How do we make AI usable in real work?”
That leads to better decisions.
According to DataCamp, the most effective AI upskilling programs are scalable, role-relevant, hands-on, reinforced over time, and measurable against performance outcomes. That is a very different model from one-off awareness sessions or passive content libraries.
BCG reinforces this from another direction. Its research found that regular AI use is much higher when employees receive at least five hours of training and have access to in-person training and coaching.
Put simply, the best programs do not just explain AI. They help people practise with it, apply it, and build confidence using it in the context of real work.
That is what closes the gap.
What leaders should do now
If your organisation is already investing in AI learning but still feels short on real capability, this is where to look first.
Ask whether your current training is tied to actual roles, actual workflows, and actual outcomes. Ask whether employees are getting hands-on practice instead of just passive exposure. Ask whether managers know how to translate AI learning into changes in daily work. And ask whether your teams have the data access, governance support, and process clarity needed to use AI well once the training ends. These recommendations are grounded in the patterns reported by DataCamp, BCG, and Microsoft WorkLab.
That is usually where the truth sits.
The skills gap is rarely just a learning problem. More often, it is a sign that the enterprise has not yet aligned learning, leadership, workflows, and governance around the reality of AI-enabled work.
Final thought
The headline is powerful for a reason: 82% train, yet 59% still have a gap.
But the deeper point is even more important.
Most enterprises do not have an AI motivation problem. They have an AI translation problem. They are trying to convert access into capability, and content into confidence, without fully reworking how learning connects to the actual flow of work.
The organisations that solve this will not be the ones that simply launch more AI courses.
They will be the ones that build a workforce that knows how to use AI well when the course is over.
FAQ
What is the AI skills paradox?
The AI skills paradox is the gap between investment and real capability. DataCamp’s 2026 enterprise research found that 82% of organisations offer some form of AI training, yet 59% still say they have an AI skills gap.
Why do enterprises still have an AI skills gap after training?
Because access to training does not automatically create practical capability. Leaders report problems with passive learning formats, lack of hands-on work, weak role relevance, and poor reinforcement over time.
Is the AI skills gap mainly a technical hiring problem?
No. DataCamp’s research says the gap often shows up in applied areas such as judging AI outputs, applying AI to workflows, making decisions with AI support, and understanding governance and responsible use.
Why is frontline adoption still lagging?
BCG found that regular generative AI use among frontline employees remains at 51%, and only one-third say they have been properly trained. That suggests many organisations have not yet translated AI learning into confident, everyday use across the broader workforce.
Does stronger AI upskilling improve ROI?
Yes. DataCamp found that organisations with mature, workforce-wide AI literacy programs are much more likely to report significant positive ROI from AI investments, with that figure rising to 42% in those more mature organisations.
Why is this also a governance and readiness issue?
Because training alone is not enough if the environment around it is weak. Microsoft WorkLab found that many organisations still struggle with cross-team data access, executive sponsorship, and documented processes, which makes it harder for employees to apply AI effectively even when training exists.
Closing the AI skills gap takes more than another training rollout. It takes a practical readiness model that connects learning, adoption, governance, and real business use.
If your organisation is investing in AI but still struggling to turn training into capability, visit GIOFAI to explore how a stronger AI governance and enterprise readiness approach can help you move from awareness to real workforce confidence.
Visit GIOFAI
