AI Training Beats AI Licences: Why Company Rollouts Stall When People Are Not Taught

In one year, companies grew the share of workers with approved AI tools from under 40% to around 60%, according to Deloitte, yet in BCG's global survey only a third of employees say they have been properly trained to use them. That gap between licences bought and people taught is where most AI budgets quietly disappear.

What the data shows

• Only one third of employees say they have been properly trained on AI, and regular use is sharply higher for those who get at least five hours of training with in person coaching (BCG AI at Work, June 2025).
• Only 12% of employees receive sufficient AI training; those with 81 hours or more a year report a productivity gain of 14 hours a week, against a median of eight (EY Work Reimagined Survey, November 2025).
• 95% of organisations are getting zero return from $30 to $40 billion of enterprise spending on generative AI (MIT NANDA, The GenAI Divide, July 2025).
• Only 25% of companies have moved 40% or more of their AI pilots into production (Deloitte State of AI in the Enterprise, January 2026).
• 60% of companies report hardly any material value from AI, while 5% capture outsized returns (BCG, September 2025).
• Only 39% of people who use AI at work had received AI training from their company, and 78% were bringing their own AI tools to work (Microsoft Work Trend Index, May 2024).

Buying licences is the easy part: usage stalls when nobody teaches people

The pattern is consistent across the major surveys. Access rises fast, skill does not. Deloitte's 2026 survey of 3,235 business and IT leaders in 24 countries found companies broadened workforce access to AI by 50% in a single year. Gallup's July 2026 data shows 47% of US employees now say their organisation has integrated AI tools, and 52% use AI in their role, but only 30% use it frequently and 15% daily.

Training has not kept pace. Microsoft's 2024 Work Trend Index found only 39% of AI users worldwide had received AI training from their company. Slack's Workforce Index from November 2024 found 61% of desk workers had spent less than five hours learning AI, and 30% had no AI training at all, not even self directed experimenting. The KPMG and University of Melbourne global study, published in April 2025 with more than 48,000 respondents in 47 countries, found only 47% of employees had received AI training.

Untrained use is not harmless. The same KPMG study found 66% of employees rely on AI output without evaluating its accuracy, 56% have made mistakes in their work because of AI, and 57% hide their use of AI and present AI generated work as their own. Almost half admitted to using AI in ways that go against company policy, including uploading sensitive company information into free public tools.

MIT NANDA found only 40% of companies had purchased an official AI subscription, yet workers from over 90% of the companies surveyed regularly used personal AI tools for work. People will use AI either way. Training decides whether they use the company's version, safely, on the right tasks.

MIT NANDA and BCG: AI pilots pay off only when people change how they work

MIT NANDA's July 2025 report, The GenAI Divide, drew on a review of over 300 publicly disclosed AI initiatives, interviews with representatives from 52 organisations and survey responses from 153 senior leaders. Its headline finding: 95% of organisations are getting zero return. For enterprise grade tools, 60% of organisations evaluated them, 20% reached a pilot and only 5% reached production. The authors put the failures down to brittle workflows, a lack of contextual learning and a poor fit with day to day operations, not to the quality of the technology or to regulation.

BCG's September 2025 study of more than 1,250 companies found 60% reaping hardly any material value from AI. The 5% it calls "future built" do three things differently when it comes to people. They planned to upskill more than 50% of their employees in AI in 2025, compared with 20% at laggards. They are six times as likely to carve out time for structured learning. And as a result, 50% more of their staff use AI in their daily work. BCG's rule of thumb for any technology change is that 70% of the effort should go to people and processes, 20% to the technology and only 10% to the AI itself.

Training depth changes results more than the choice of tool

Hours matter. BCG's June 2025 AI at Work survey of more than 10,600 people in 11 countries found regular use is sharply higher among employees who receive at least five hours of training and have access to in person training and coaching. EY's survey of 15,000 employees and 1,500 employers across 29 countries found people with 81 or more hours of AI training a year report a productivity gain of 14 hours a week, well above the median of eight. EY also found 88% of employees use AI at work, mostly for basic tasks such as search and summarising, while only 5% use it in advanced ways.

Breadth matters too. In Gallup's July 2026 figures, 45% of employees who use AI for one or two purposes report a positive productivity impact, rising to 66% for three or four, 78% for five or six and 90% for seven or more. Good training widens the range of tasks people trust AI with.

And direction beats tools. BCG's June 2026 AI at Work report, based on 11,749 workers in 14 markets, found clear strategy lifts AI's impact by 25 percentage points, against only five points for better tools. 72% of respondents said the skills expected of them have shifted, but only 36% feel they have received adequate upskilling. Microsoft's 2026 Work Trend Index, a survey of 20,000 knowledge workers in 10 countries, found organisational factors such as culture, manager support and talent practices account for 67% of reported AI impact, compared with 32% for individual mindset and behaviour.

Managers and champions decide whether the training sticks

A single workshop fades if nobody keeps practising. Gallup's February 2026 survey of 23,717 US employees measured what separates frequent AI users from the rest. 78% of employees who strongly agree their manager actively supports AI use are frequent users, against 44% of those who do not. 88% of those who say AI fits well with the systems they already use are frequent users, against 55%. Clear AI policies (68% against 47%) and support for experimenting (72% against 44%) show the same gap.

BCG's 2025 survey found that when leaders show strong support for AI, the share of employees who feel positive about it rises from 15% to 55%. Only about a quarter of frontline employees say they get that support. Microsoft's 2026 research found employees were 1.4 times more likely to be high frequency users of agentic AI when their managers made it safe to experiment, yet only 26% of AI users say their leadership is clearly and consistently aligned on AI strategy.

The companies that report near complete adoption look like this in practice. Morgan Stanley said in June 2024 that 98% of its financial advisor teams had adopted its internal AI assistant, which gives advisors quick access to the firm's own research and knowledge. Moderna reported in April 2024 that within about two months of rolling out ChatGPT Enterprise, employees had built more than 750 custom GPTs, each user averaged 120 conversations a week, 40% of weekly active users had created GPTs, and its legal team reached 100% adoption. In both cases the tool was tied to the daily work of particular teams rather than handed out as a general licence.

What successful rollouts do: department training, champions and use case playbooks

For a company of 25 to 150 people, the research points to a clear pattern.

1. Train by department, on real work. A leasing team practises on its own enquiry emails and listings. Finance practises on its own reconciliations. Generic "intro to AI" sessions do not survive contact with Monday morning.
2. Give it real time. BCG's five hour threshold is a floor, not a target, and EY's biggest gains came at 81 hours a year or more. Book protected learning time; BCG's future built companies are six times as likely to do so.
3. Name a champion in every team. One person who uses AI daily, answers the small questions and shares what works. Managers must visibly use the tools themselves, since manager support is one of the largest differences Gallup measured.
4. Write a use case playbook per role. List 5 to 10 tasks each role should do with AI, the prompts that work, what information may and may not go in, and how to check the output. This is what turns a licence into a habit.
5. Measure breadth, not logins. Track how many tasks each person uses AI for. Gallup's data shows impact climbs with every extra kind of task.

For Dubai companies the urgency is higher. Stanford's 2026 AI Index puts the UAE's generative AI adoption rate at 64%, far above the 28.3% recorded for the United States. Your staff are already using AI. The question is whether they are using your approved tools, trained on your tasks, with your rules.

Questions people ask

How much AI training do employees actually need?

BCG found regular use rises sharply after at least five hours of training combined with in person coaching. EY found people with 81 or more hours of training a year report a productivity gain of 14 hours a week. Start with a focused first session by department, then keep coaching weekly through champions.

Why do most AI pilots fail?

MIT NANDA found 95% of organisations get zero return and pointed to tools that do not fit daily workflows rather than weak technology. Pilots that skip training, clear use cases and manager support rarely make it past the trial.

Is self paced online training enough?

The evidence favours hands on training. BCG ties higher regular use to in person training and coaching. Self paced courses work best as a follow up, not the starting point.

What should we measure after an AI rollout?

Measure how many different tasks each person uses AI for, how often, and whether output is checked. Gallup found 90% of employees using AI for seven or more purposes report a positive productivity impact, against 45% of those using it for one or two.

CroVIN runs in person AI training classes by department for company teams in Dubai, built around each team's own daily tasks, so the licences you buy actually get used.