# The AI Training Gap: 70% of Managers Want It, 14% of Workers Get It

> 70% of managers say an AI-trained workforce is critical, but only 14% of workers have been offered training. Here's what's going wrong.

- URL: https://hyperfocusam.com/blog/ai-training-gap-managers-vs-workers/
- Language: en
- Author: Sam Wong (黃力桐), https://hyperfocusam.com/

---

[Sam Wong](https://hyperfocusam.com/en/)

[Sam Wong](https://hyperfocusam.com/about/) · 24 February 2026

Google [published a stat](https://www.ipsos.com/en-us/googleipsos-ai-works-america-poll) that stopped me mid-scroll: 70% of managers believe an AI-trained workforce is critical for success, but only 14% of workers have actually been offered AI training.

I've spent two years inside that gap. Not studying it. Living in it. I walk into rooms where the CEO has read three McKinsey reports and the team has never opened an AI tool. The disconnect isn't subtle.

## Who Actually Owns This?

The structural problem I see in almost every engagement: nobody owns AI training. HR thinks it's IT's job. IT calls it a "user behavior issue." L&D doesn't have budget. The workforce is left figuring things out from YouTube.

At one company, the IT director told me his team flags any AI tool as a security risk by default. I've seen this pattern at least a dozen times. The blocking doesn't stop AI usage. It just pushes it underground where nobody can govern it.

## The Difference Between Tool Training and Behavior Change

This is the part companies keep getting wrong. "AI training" in most organizations means someone shows Copilot features for 90 minutes, gets a 9/10 satisfaction score, and leaves. Three weeks later, nothing has changed.

When I [ran a 6-session program for the food manufacturer's HR team](https://hyperfocusam.com/en/cases/garden-ai-pioneer-programme/), the breakthrough wasn't in session two when they learned Copilot. It was in session five when a participant said "I can figure out what to ask" instead of asking me for a prompt list. That shift — from "give me the 100 prompts" to "I can think through this myself" — is what actual training produces. A webinar can't do that.

When I [trained 1,500+ participants at a major Hong Kong bank](https://hyperfocusam.com/en/cases/hong-kong-bank-ai-training/), most of them had never touched an AI tool before walking in. By the end, they were applying the 70/30 split to their daily work. The gap between "never used" and "productive" is smaller than executives think. But someone has to actually bridge it. A memo from HR doesn't count.

## What I Keep Coming Back To

Start with 15 willing people, not 200 reluctant ones. Solve the security question before the first session — tier 1, tier 2, tier 3, what's safe and what's off-limits. And stop measuring sessions attended. Measure whether the weekly report takes 15 minutes now instead of 90.

The uncomfortable truth about the 70/14 gap is that it's not a resource problem. Most organizations have the budget. They have the tools: Copilot Chat is included at no extra cost with an eligible Microsoft 365 subscription. What they don't have is someone who wakes up in the morning thinking "AI adoption is my job." Until that person exists, the gap holds.