Your Boss Thinks 4% of Staff Use AI Daily. The Real Number Is Three Times Higher.

The 4% vs 13% gap, explained

McKinsey surveyed over 3,600 employees and nearly 240 C-level executives across six countries for their Superagency in the Workplace report. The numbers that came back were striking. [1]

When asked what percentage of employees use generative AI for at least 30% of their daily work, C-suite leaders said 4%. When employees answered the same question about themselves, 13% said they did. That’s a threefold gap. Leaders aren’t just slightly out of touch; they’re working with a picture of AI adoption that’s fundamentally wrong.

And it gets more telling. Only 16% of C-suite leaders expect employees to be using AI for 30%+ of their work within a year. Among employees themselves, 34% expect to be doing so. Employees are not just using AI more than leaders think; they’re planning to use it much more, much sooner, than leadership anticipates.

Why employees go quiet about their AI use

This is the part that tends to surprise people. Employees aren’t hiding their AI use because they’re doing something wrong. They’re often hiding it because they’re not sure if it’s allowed, or because they don’t want to be told it isn’t.

Think about it from an employee’s perspective. You’ve started using Claude or ChatGPT to draft your weekly report. It takes you 40 minutes instead of 2 hours. Your output is better and your manager is happy. But your company has no AI policy, IT hasn’t blessed the tool, and your colleague got a vague warning about “using AI responsibly” last month. Do you mention it in your 1:1? Probably not.

This quiet adoption pattern is consistent across industries. Most employees are using AI through personal accounts and free tools, not through company-approved software. It doesn’t show up in procurement data or IT logs. It’s essentially invisible to leadership. [1]

There’s also a competence gap working in reverse. Leaders who aren’t using AI themselves often don’t understand how integrated it has become into everyday cognitive work tasks. Writing, research, summarising, planning, responding to emails. If you haven’t experienced how deeply AI can embed into a typical work day, you won’t expect your employees to have done so either.

What employees are actually doing with AI at work

The McKinsey research also found that 94% of employees report familiarity with generative AI tools. [1] That number is almost certainly accurate. By 2026, everyone has at least opened ChatGPT. The question is what they’re doing with it.

From what we see working with corporate teams at Future Factors AI, the most common patterns are:

• First drafts of anything written: Reports, emails, proposals, summaries. The AI writes version one; the human edits and improves.
• Research and synthesis: Summarising lengthy documents, pulling key points from reports, comparing options across categories.
• Meeting preparation: Generating agendas, briefing documents, potential questions to ask, talking points for difficult conversations.
• Data interpretation: Pasting tables or numbers into a chat interface and asking for plain-English explanations of what they mean.
• Learning on the job: Using AI as a real-time explainer for unfamiliar topics, jargon, or concepts that come up in meetings.

None of this is exotic. It’s the everyday work that professionals do. AI is being woven into it quietly and steadily, regardless of whether there’s an official company AI strategy in place.

The AI investment paradox

Here’s where the McKinsey data gets uncomfortable. 92% of companies say they plan to increase AI investment over the next three years. But only 1% say they’ve reached maturity in AI deployment. [1]

And the revenue results are mixed at best. Among US C-suite respondents, only 19% said revenues had increased by more than 5% from generative AI. 36% reported no revenue change at all.

What’s happening? Companies are spending on AI, but mostly on infrastructure and tools rather than on the people who use them. You can have the best AI platform in the world; if your team doesn’t know how to use it effectively, the ROI won’t materialise. The investment gap isn’t in technology. It’s in capability building.

The companies that are seeing real returns from AI share a common characteristic: they’ve invested in structured, practical training. Not one-off lunch-and-learns, not hour-long webinars. Sustained, role-specific programmes that teach people how to apply AI to their actual work tasks. Building individual AI workflows is part of it, but the organisational layer matters just as much.

The training problem no one is solving

48% of employees say training is the most important factor for AI adoption. Nearly half report receiving minimal or no AI training from their employer. [2]

That gap is astonishing given how much is being invested at the platform level. The pattern we see repeatedly is: a company buys Microsoft Copilot or a ChatGPT Enterprise licence, rolls it out to the team, and sends a “here’s your login” email. Then wonders why adoption is low or uneven.

The problem isn’t that employees don’t want to use AI. It’s that most people need to see it applied to their specific job before they believe it will actually help them. Abstract demonstrations don’t move the needle. Showing an HR manager how to use AI to shortlist CVs in half the time does.

This isn’t a criticism; it’s a gap the market is still figuring out. But the organisations that close it fastest will compound the advantage significantly.

What leaders should do differently

The good news is that this gap is fixable. It starts with leaders getting honest about their own AI literacy before trying to design AI strategy for others.

Start by using AI yourself, consistently. Not in a demo. In real work. If you’re a manager and you haven’t used AI to draft a document, prep for a meeting, or analyse a dataset in the past two weeks, you’re not leading from experience. That matters.

Ask the right questions in your next 1:1s. Not “are you using our AI tools?” but “show me something you’ve done recently with AI. Walk me through it.” You’ll learn more about your team’s actual usage and needs in that conversation than any survey will tell you.

Create psychological safety around AI use. If your team is hiding their AI use because they’re not sure it’s allowed, that’s a policy clarity problem you need to solve. Publish a simple, pragmatic AI use policy. The goal isn’t control; it’s clarity about what’s fine, what requires caution, and what’s off limits.

Invest in structured training, not just tools. The 92% vs 1% paradox (high investment, near-zero maturity) is a capability problem. Fix it with real training programmes, not one-off sessions. The research on enterprise AI adoption consistently points to training as the single biggest lever.

What employees should do in the meantime

If you’re an employee who’s already using AI quietly, or who wants to but doesn’t know where to start, here’s the practical path.

First, don’t wait for permission if there’s no policy. If your company hasn’t published an AI use policy, that’s not the same as a prohibition. Use common sense: don’t input confidential client data into a free AI tool, don’t paste sensitive financials into a public interface. Within those guardrails, explore.

Second, keep a log of what works. When you find a prompt or a technique that saves you meaningful time, write it down. This serves two purposes: it helps you reuse it, and it gives you concrete examples when the conversation about AI at your company eventually happens (and it will).

Third, share what you learn. The gap between the 13% and the majority is often a knowledge gap, not a motivation gap. If you’ve found something that works, show a colleague. The informal sharing of effective AI techniques is how ground-level adoption actually spreads.