Why Most Employees Use AI but Few Get the Time Back

Almost everyone in a large organization now uses AI at work. Very few get much time back from it. AI training for employees is the practical gap between those two facts: the hours of learning, the change to the work itself and the support from managers that turn occasional use of a chatbot into a day or more of real capacity each week. Most organizations have paid for the first part, the licenses, and skipped the rest.

Since AI tools reached everyday work in 2023, the managers in our leadership programs have tended to describe the same frustration: the tools are in place, people say they use them, and the workload looks no lighter. That frustration usually points to the training and the work design, not to the tool.

This article explains what separates employees who gain from AI from those who only use it, why access is no longer the problem, what the gainers do differently, and how to close the gap in your own organization.

The Difference Between Using AI and Gaining From It

Surveys of AI at work usually count users. The more useful count is of people whose work has changed. EY's Work Reimagined Survey 2025, which questioned 15,000 employees and 1,500 employers across 29 countries in August 2025, separates the two:

• Basic use. 88% of employees said they use AI at work to some degree, and 37% use it daily. Most of that use is simple: searching for information (54%) and summarizing documents (38%).
• Advanced use. Only 5% qualify as advanced users, who combine several tools in their work and use AI as a thought partner rather than a faster search box. EY found they gain roughly a day and a half of extra productivity a week.

The distance between those two groups is the whole story. A team where everyone asks a chatbot to tidy an email is a team of AI users. A team where an analyst drafts the first version of a market review with AI, tests it against a second tool and spends the saved afternoon on the client's actual question is a team that gains from AI. The software is the same in both cases.

EY's own summary is blunt: employees may be saving a few hours here and there, but nothing that fundamentally changes how work gets done or how the business performs. Only 28% of the employers surveyed had built the conditions to turn AI into high-value results, and EY estimates that where those conditions are weak, productivity gains lag by more than 40%.

Why Access Is Not the Gap

When AI results disappoint, the first instinct is often to buy more: a better model, an enterprise license, a new assistant built into the email system. The evidence points the other way. In large organizations, almost everyone who wants a tool already has one, and many bring their own. EY found that between 23% and 58% of employees, depending on the sector, use personal AI tools at work, some paying for their own subscriptions.

A license is not a skill. People who have access but no training use AI the way they first met it, as a search engine that writes in full sentences. That saves minutes. It does not change what a person can take on in a week.

The pressure on employees makes the gap worse. In the same EY survey, 64% of employees said their workload had increased over the past year. People under that pressure rarely have the slack to experiment with a tool until it becomes useful. They use it for whatever is quickest and move on.

What Advanced AI Users Do Differently

The 5% who gain the most are not simply more enthusiastic. EY and BCG point to four conditions that sit behind them, and most of them are decisions an employer makes, not traits an employee brings.

1. They train for far longer

EY found that skills, meaning the training people get and how well their managers use AI, account for nearly half of what explains the hours employees save with it. Employees with 81 or more hours of AI training in the past year reported saving 14 hours a week. Those with fewer than 4 hours of training saved 3. Only 12% of employees had received the higher level. BCG's AI at Work 2025 survey of 10,635 people shows the same pattern in how often people use AI: 18% of employees with no training were regular users, against 89% of those with more than ten hours.

Eighty-one hours sounds like a lot until you divide it by the working year. It comes to about an hour and a half a week. Most organizations offer far less, often a single launch session.

2. They use AI on real work

Advanced users don't practice on exercises. They apply several tools to the documents, analyses and decisions they are paid for, and they treat the output as a draft to question. EY describes them as using AI as a colleague, coach and thought partner rather than as an automation tool. That habit is learned on the job, not in a classroom.

3. Their managers know the tools

EY found that employees who are confident their managers know how to use AI well see better results. A manager who uses AI in front of the team, and can say which tasks it suits, gives people permission to try it on work that matters. A manager who has never opened the tool makes it a private hobby.

4. Their work has been redesigned

BCG compared AI users at companies that only rolled out AI tools with AI users at companies redesigning whole workflows around AI. In the redesigning companies, 55% of employees saved more than an hour a day, against 29% where tools were simply handed out. Those companies also trained more: 67% of their employees had more than five hours of upskilling, against 49%.

If the work stays the same, the saved hour gets spent on the same work. Redesign is the step that decides where it goes.

How to Close the Gap With AI Training for Employees

None of these steps needs a large budget. They need a decision about where AI should change the work, and the patience to train people inside that work rather than around it.

1. Start with two or three workflows

Pick a small number of tasks that take real time and recur every week: a monthly report, proposal drafting, customer replies, a reconciliation. Map how they run today and decide what AI should take over. Company-wide rollouts spread training thinly across everything and change nothing in particular.

2. Train inside the work

Build the training around those workflows, using the team's own documents. Short, regular sessions with someone who already uses the tools well work better than a single workshop. Aim for hours across the year, not a launch day. EY reports that it built its own AI training into daily tasks rather than separate learning events, and that more than 83% of its employees completed foundational AI learning.

3. Train managers first

Managers decide whether experimenting with AI counts as work or as a distraction from it. Train them before their teams, and expect them to use the tools where their people can see them.

4. Decide what stops

For each redesigned workflow, write down what the team will stop doing and what the saved time is for: more client work, the backlog, learning, or a lighter week. EY's advice is to determine what employees should stop doing as AI comes in, and to make role redesign a standard part of every AI rollout rather than an afterthought.

5. Measure hours and quality, not logins

Usage dashboards show who opened the tool. They don't show whether the monthly report now takes two days instead of four, or whether its errors went down. Measure the workflow before and after. As EY's Global Chief Innovation Officer Joe Depa puts it, "Organizations fixated solely on hours saved miss critical dimensions like improved accuracy, reduced error rates, and enhanced decision-making quality."

What This Means for Kenyan Employers

EY's 88% is a large-company figure. Its survey included only organizations with 1,000 or more employees, and about half of its respondents worked in the Americas. Most Kenyan organizations are far smaller, and AI use across the country is much lower.

Microsoft's AI Economy Institute estimates that 8.1% of Kenya's working-age population used generative AI tools in the second half of 2025, up from 7.8% in the first half. Across the Global South the figure was 14.1%, and across the Global North 24.7%.

That makes the training question more pressing for Kenyan employers, not less. With fewer people arriving already fluent, organizations can't rely on staff teaching themselves. And where staff do teach themselves, they often do it with personal accounts, on their own phones and data bundles, with client information that Kenya's Data Protection Act 2019 expects the employer to protect. A training plan that names the approved tools and shows people how to use them on real work closes both gaps at once.

There is a retention side too. Kenyan professionals with strong AI skills can now be recruited by regional and international employers, often for remote work. An employer that trains people and leaves their jobs unchanged gives them a better CV and no reason to stay.

Three Common AI Training Mistakes

Counting licenses as adoption

A dashboard showing that 90% of staff have logged in says almost nothing about whether anyone's work has changed. Track the workflows you redesigned and the time and quality they gained.

Treating training as an event

A launch workshop teaches people where the tool is. The hours that make a difference come afterward, spread across the year and tied to real tasks. Budget for those hours the way you budget for the licenses.

Training people without changing their jobs

Training people and leaving their job unchanged builds a CV, not a capability. The skills go unused at work, the saved time leaks into more of the same tasks, and the best-trained people take their skills somewhere that will use them.

Frequently Asked Questions

Conclusion

Almost everyone uses AI now, so use is no longer an advantage. The return comes from hours of training on real work and jobs redesigned to use what that training builds. Both are decisions an employer makes.

Pick one person on your team who uses AI every day. How many hours of AI training have they had this year, and has anything about their job changed since they started using it?

If you want your managers to lead that change, our Leadership and Organizational Change Management program can help. Request a proposal and we will start from the workflows your teams run every week.