TL;DR: Your people are adopting AI faster than your organization is preparing them for it. The gap between the two is where risk lives.
A number in this week’s workforce data is worth sitting with.
Fifty-five percent of workers now use generative AI or AI agents daily or weekly. Only 33% have used any organization-provided AI training in the past six months. More than 28% say their organization provides no AI training at all. Those figures come from The Conference Board’s Skilling for AI study, published July 28, 2026, based on interviews with 35 enterprise leaders and a global survey of nearly 1,300 workers.
Read it again with your own org in mind. More than half your people are already using AI to do their jobs. Most of them taught themselves.
The gap is not really about training
There is a second number that makes the first one urgent.
A separate survey from INTOO and The Harris Poll, fielded in June 2026 among 1,085 employed U.S. adults, found that 72% say AI is being used for business purposes somewhere in their organization. Only 12% report company-wide use through formal, approved tools. And 22% say the use they see is informal and employee-driven, with no company-approved tool or process behind it.
So AI is nearly everywhere, and it is governed almost nowhere. That is the actual finding. Adoption has already happened. Enablement and governance have not caught up.
Put those two datasets side by side and the picture is clear. Employees are making individual judgment calls, dozens of times a day, about which tool to use, what information to paste into it, how much to trust what comes back, and whether to tell anyone they used it. Those judgment calls are being made without shared standards, without guardrails, and often without anyone in a position to check.
This is what we call Shadow AI, and that 22% is worth pausing on, because it is not an inference drawn from a gap in the data. It is employees describing shadow AI inside their own organizations, unprompted. It is also not a discipline problem. It is a vacuum problem. People fill the space where guidance should be.
The gap between AI use and AI training is a risk problem wearing a training problem’s clothes.
The Evaila Lens: We do not treat enablement as the soft part of an AI program. It is the Ready phase of our CHART Framework, and it is the step most AI consultancies skip. Tools get deployed, adoption gets assumed, and the organization discovers six months later that half the workforce is improvising.
”We offer training” is not the same as “our people are prepared”
The instinct here is to buy a training module and check the box. The same Conference Board data suggests that is not enough.
Only 48% of workers agreed their organization gives them sufficient time during work hours to develop AI skills. Just under 48% agreed they had the tools, access, and resources to build AI capability. Enterprise leaders in the study reported that developing real AI skill requires more than access to training. It requires dedicated time, hands-on practice, and managerial support.
That tracks with what we see in the field. A recorded course watched at 1.5x speed between meetings does not build confidence. What builds confidence is working through a task the person actually owns, with a tool they will actually use, and finding out where it helps and where it fails.
The INTOO data shows what happens when that support is missing. Twenty percent of employees said they want to advance their use of AI but do not know where to start. Another 20% said they are worried about making mistakes. These are motivated people waiting for permission and a path. That is one of the cheapest problems a leadership team can solve.
What good enablement actually looks like
Three things separate organizations that close this gap from those that keep talking about it.
Shared language before shared tools. Everyone needs the same working understanding of what AI is good at, where it fails, what a hallucination is, and which decisions still require a human. Without that baseline, policy is just paperwork, because people cannot apply a rule they do not understand.
Guardrails that name the real situations. “Do not input confidential data” is too abstract to change behavior. “Here is where client information can and cannot go, here is the approved tool for this workflow, and here is who to ask when you are unsure” is specific enough to follow. Governance belongs inside the workflow, not in a PDF nobody opens.
Function-specific practice. Finance, HR, operations, and sales do different work and face different risks. Training built around each function’s real tasks lands. Generic training does not. This is the difference between people knowing AI exists and people knowing what to do with it on a Tuesday.
The Evaila Lens: We run this on two tracks at once. Going Broad means raising baseline fluency and safe-use standards across the organization so the 55% who are already using AI are doing it well. Going Deep means picking one or two high-value workflows and redesigning them properly. Broad enablement reduces risk. Deep transformation creates advantage. Most companies need both, and they need them in that order.
The cost of waiting is not neutral
Every month this gap stays open, two things compound. Habits form, and unsupervised habits are harder to correct than absent ones. And exposure accumulates quietly, in ways that usually surface at the worst possible moment.
There is also an opportunity cost. When only 12% of organizations have AI running through approved tools and processes, the ones that get there first are not winning on model quality. They are winning because their people know what to do.
AI is already in your business. The real question is whether your people are using it safely and well, or just quietly.
If you are not sure where your workforce actually stands, that is the place to start. Our AI Readiness Assessment looks at people alongside data and systems, because readiness is never just a technology question. And AI Foundations Training is built to take teams from AI-curious to AI-capable with real examples tied to the work they already do.
Demystifying AI. Delivering Results.
