85% of workers can’t connect AI training to their job

A survey of 2,000 people finds AI training often fails at the moment it should matter: 85% of workers say they can’t link what they learned to their actual role, while 78% report training sits in systems disconnected from where they work. The result is an AI r

By the time the training session ends, the real work is still waiting.

For many employees, that’s where the promise of AI adoption breaks. Docebo points to an inflection moment for artificial intelligence—one where the question is no longer whether organizations are adopting it. but whether their people can actually use it. The uncomfortable answer, based on a survey of 2,000 people, is that most aren’t.

The core issue is not described as a technology failure. The tools work. The breakdown is simpler to name and harder to fix: companies moved fast on AI tools and initiatives, but not on building the human capability to use them.

The survey lays out the failure in three walls that stack on top of each other.

First, 56% of workers say they’re buried in manual, pre-AI tasks, leaving no time to learn the tools meant to free them from those tasks.

Second, even when people find time, 85% can’t connect what they learned to their actual role.

Third, 78% say training happens in systems completely disconnected from where they work.

Taken together, the picture is stark: a workforce can have access to AI, but not the real ability to use it.

Docebo argues that organizations are still running the same playbook—deploy the tool. schedule the training. check the box. move on. The consequence, the company says, isn’t transformation. It produces completion rates. And the metrics being tracked often miss the difference between having done training and being able to apply it.

Seats purchased and licenses deployed don’t prove readiness. The same goes for modules completed. Those are described as procurement metrics dressed up as readiness metrics—useful for telling what an organization spent, but not what people can actually do.

The standard that Docebo says matters instead is evidence: whether a person can demonstrate capability in the work they perform, on the systems where work happens, in a way that can withstand a regulator or a CFO.

Real readiness, the argument goes, is a skills question. Can a person apply what they learned to their specific work, their specific goals, right now? If an organization can’t answer that, Docebo says it doesn’t have an AI strategy—only an AI expense.

The survey also frames the problem as one that doesn’t get easier with time. The infrastructure employees rely on to upskill wasn’t built for the rate of change AI introduces. It was built for slower, more predictable skill development—an era that, in Docebo’s view, is now over.

The change, Docebo says, isn’t just better content or more training hours. It requires a fundamentally different philosophy for how learning works inside an organization—one that connects learning to an individual’s specific role and goals. and embeds learning AI into the culture of how work gets done.

That shift starts with building pathways that move with the employee. Docebo cautions that AI literacy isn’t one skill; it’s dozens of skills that show up differently by role. A sales rep using AI for proposal research is dealing with different gaps than a manager interpreting AI-generated performance insights. or an analyst pressure-testing AI outputs before taking action. The training window. Docebo says. is when someone uses a new AI feature for the first time—tied to real work. not six weeks later in a scheduled session.

Next comes the decision to treat expertise as infrastructure. Docebo notes that in many organizations. a small number of people have already figured out how to use AI tools effectively. building workarounds and instincts no formal curriculum has captured. The proposal is to find them. assign them as mentors during tool rollouts. and run short peer-led sessions around how the tools actually work.

Then there is the question of where learning lives. Docebo argues that employees shouldn’t have to leave their workflow to build an AI skill. Instead. AI learning should be surfaced inside the tools where work happens—for example. a prompt in a CRM when a rep uses an AI-generated summary for the first time. a triggered module when someone accesses a new feature. or a nudge when a team adopts an AI-assisted process. Friction, the message warns, is where learning dies.

Docebo also says organizations need a clear starting point for ownership. Ownership without direction becomes abandonment. The recommendation is a skills assessment tied to how AI is changing work specifically—capabilities the role requires now and will require in 18 months—not generic digital literacy. Visibility into a gap, Docebo says, is what turns self-directed learning into a habit.

But this training philosophy, Docebo insists, requires infrastructure many companies still aren’t building. That includes a persistent learner profile that follows employees from role to role; skills intelligence that infers capability continuously rather than on an annual review cycle; and learning that lives inside the tools where work actually happens.

The article’s final warning is that capability doesn’t compound on its own. Docebo argues readiness compounds only when it sits on top of data no AI agent can fabricate: compliance records tied to specific people on specific dates; skills graphs that reflect what a workforce can actually do; learning history across years and roles; and external training data tied to revenue. certification. and customer outcomes.

In that framing, the advantage belongs to organizations that build both the data foundation and the human capability to use AI together—not the ones that simply moved fastest on tools. Every day the gap between tools and readiness widens, Docebo says, it gets harder to close.

AI training workforce readiness skills assessment Docebo AI adoption learning in the flow of work skills intelligence compliance data