AI skills training gaps persist
- Deloitte’s 2026 enterprise AI report says the biggest blocker is now skills, not model access, as companies try to push pilots into production. - McKinsey says 88% of organizations use AI somewhere, but only about one-third have started scaling it, and just 39% see EBIT impact. - Washington is reacting too — the Labor Department issued an AI literacy framework in February, signaling the training gap is now a workforce issue.
AI adoption has reached the boring part — and that’s the real test. Most big companies already have access to strong models, copilots, and some kind of agent tooling. But the thing slowing them down is much less glamorous: workers often don’t know how to use the tools well, check the outputs, or redesign a workflow around them. That gap showed up clearly in a fresh wave of 2026 research, and it’s big enough that the U.S. government is now treating AI literacy as workforce infrastructure. ### Why is this suddenly a people story? Because the technology bottleneck eased. Deloitte’s 2026 enterprise survey says worker access to AI rose by 50% in 2025, and companies expect the share with at least 40% of experiments in production to double within six months. But the same report says the AI skills gap is the biggest barrier to integration, and education was the top talent response. Basically, companies bought the tools faster than they built the habits to use them. (deloitte.com) ### What skills are actually missing? Not mostly advanced machine-learning skills. The missing layer is applied AI literacy — writing useful prompts, testing whether an answer is wrong or misleading, knowing when a human needs to stay in the loop, and fitting AI into a real workflow without creating risk. DataCamp’s 2026 survey of 500-plus US and UK enterprise leaders says 59% still report an AI skills gap, and it shows up in output evaluation, workflow use, decision support, and governance more than in frontier-model engineering. (deloitte.com) ### If companies already train people, why isn’t it working? Because a lot of the training is passive. DataCamp says 82% of organizations offer some kind of AI training, but only 35% have a mature organization-wide upskilling program. Video courses are common, yet leaders say those formats are hard to apply in real work and often lack hands-on practice. McKinsey makes the same point from another angle: AI is changing jobs faster than classroom-style programs can keep up. (datacamp.com) ### How does that hurt ROI? It creates the classic pilot trap. McKinsey’s 2025 global AI survey says 88% of organizations now use AI in at least one business function, but nearly two-thirds still have not begun scaling across the enterprise. Only 39% report EBIT impact at the enterprise level. The pattern is pretty clear — teams can demo a chatbot, but they struggle to turn that into repeatable savings or growth because the surrounding workflow never changed. (datacamp.com) ### Why does workflow redesign matter so much? Because AI value usually comes from changing the job, not stapling a model onto the old process. Deloitte says 34% are truly reimagining the business, while many others are still using AI at the surface level. It also says companies have not broadly redesigned jobs around AI capabilities. That’s the catch: if workers treat AI like a smarter search box, returns stay small. If they rebuild review loops, escalation paths, and decision rights, returns get real. (mckinsey.com) ### Is government stepping in? Yes — and that matters. On February 13, 2026, the U.S. Department of Labor released an AI literacy framework with five foundational content areas and seven delivery principles. It’s meant to guide workforce and education systems, not just corporate L&D teams. That move tells you this is no longer a niche “prompt engineering” problem. It’s becoming basic employability infrastructure. (deloitte.com) ### So what does better training look like? Less one-off certification theater, more learning in the flow of work. McKinsey argues for continuous, peer-driven learning because models and features change too fast for static courses. The useful version looks practical: role-based examples, hands-on tasks, output checking, escalation rules, and managers modeling good behavior. In other words, teach people how to work with AI, not just how to admire it. (dol.gov) ### Bottom line? The AI story is shifting from access to execution. Companies no longer win just by buying better models. They win if ordinary teams can use those models safely, skeptically, and inside redesigned workflows that actually move the business. (deloitte.com) (mckinsey.com)
