Ikhaya Automations

Ikhaya Automations

Women Are Using AI Less Than Men. The Reason Matters.

Women are using generative AI less than men. But telling women to be more confident with technology may completely miss the point.

There is a gender gap in AI adoption.

Across more than 300,000 people and over 100 countries, research reviewed by the Harvard Business School AI Institute found that 47.8% of men had adopted generative AI compared with 39.3% of women.

The gap has narrowed since generative AI first entered the mainstream. But it hasn’t disappeared.

In the UK, Deloitte found a particularly noticeable difference: in its 2024 research, 43% of men surveyed were using generative AI compared with 28% of women.

It would be easy to look at numbers like these and conclude that women need more AI training, more confidence or more encouragement to embrace technology.

But that explanation may be far too simple.

Because emerging research raises a much more uncomfortable possibility.

What if some women are not simply less confident about using AI at work? What if they have more reason to be cautious about being seen using it?

Using AI can come with a competence penalty

One of the most interesting pieces of research into workplace AI use did something deceptively simple.

Researchers asked engineers to assess a piece of computer code.

The code didn’t change.

What changed was what the reviewers were told about how it had been produced.

When reviewers believed an engineer had used AI assistance, they rated that engineer as less competent than when they believed the same work had been completed without AI.

Same output. Different perception.

And the penalty wasn’t equal.

The research, involving a wider company dataset of almost 29,000 software engineers and a controlled experiment with 1,026 engineers, found evidence that women could face a greater competence penalty for using AI than men.

That changes the conversation.

If someone believes using AI might make colleagues question whether the quality of their work comes from their own ability or from the machine, hesitation isn’t necessarily evidence that they don’t understand AI.

It could be a rational response to the environment around them.

And that matters because AI adoption isn’t happening in a vacuum.

The problem may be permission, not capability

There is another clue in the research.

McKinsey’s Women in the Workplace 2025 found that only 21% of entry-level women said their manager encouraged them to use AI, compared with 33% of entry-level men.

That might sound like a relatively small management behaviour.

It isn’t.

McKinsey also found that employees whose managers encouraged them to use AI were more than 50% more likely to use it.

Think about what that means in practice.

Two equally capable employees could sit in the same organisation with access to exactly the same technology.

One hears: “Try it. Experiment. See what you can do with it.”

The other hears nothing.

One sees colleagues using AI openly and being recognised for finding faster ways to work.

The other isn’t quite sure whether using it will be considered innovative, lazy or even cheating.

Technically, both employees have access to AI.

In reality, they do not have the same permission to experiment.

And buying another AI licence won’t fix that.

This is where businesses need to be careful

There is a tendency to treat AI adoption as a technology project.

Choose the software. Buy the licences. Arrange some training. Announce that the business is embracing AI. Done.

But increasingly, the interesting questions aren’t about the technology at all. They’re about people.

  • Who actually uses it?
  • Who experiments with it?
  • Who feels comfortable admitting that they used it?
  • Who gets praised for finding a better way to complete a task?
  • Who worries that using the exact same tool will make people question their competence?
  • Who has enough time to learn through experimentation in the first place?

These questions matter because access to AI and adoption of AI are not the same thing.

The UK is already giving us an interesting illustration of that.

Office for National Statistics research published in 2026 found that around 35% of UK businesses with 10 or more employees reported using at least one AI technology.

Yet around 55% of surveyed workers reported using AI for work or education.

Employees are not necessarily waiting for their employers to develop formal AI programmes. They are experimenting already.

The question for businesses is whether everyone has an equal opportunity to do so.

Because the advantage from AI can be real

This wouldn’t matter nearly as much if generative AI were simply another overhyped technology producing marginal improvements.

But some of the strongest workplace research suggests it can materially improve performance.

In a major field study of customer-support workers, researchers found that access to a generative AI assistant increased productivity by around 15%.

But the most interesting finding wasn’t the average improvement. It was who benefited most.

The largest gains went to less experienced and lower-skilled workers.

People with only a couple of months’ experience could perform more like colleagues who had been in the role considerably longer.

That’s important.

It suggests AI doesn’t necessarily make the strongest employee even stronger while everyone else falls further behind.

Used well, it can do the opposite.

It can help people access knowledge, learn faster and close experience gaps.

That makes generative AI potentially one of the most powerful workplace equalisers we’ve had in years.

But only if people actually get to use it.

AI could narrow the gender gap — or quietly widen it

This is the tension businesses need to understand.

There is nothing inherently inevitable about AI increasing workplace inequality.

In fact, there are good reasons to think it could reduce some of it.

Imagine someone early in their career who can ask AI to explain an unfamiliar financial concept before walking into a meeting.

Someone moving into management who can use it to challenge their thinking before making a difficult decision.

An administrator who can automate repetitive work and spend more time solving problems.

A non-technical employee who can analyse information they previously would have needed specialist help to understand.

Someone who doesn’t want to ask a colleague what feels like a “stupid question” but is perfectly comfortable asking an AI assistant to explain it five different ways.

Used like this, AI doesn’t replace human capability. It expands it.

And if the biggest productivity benefits really do accrue to people with less experience, the opportunity is enormous.

But there is another possible future.

Some employees experiment freely while others hesitate. Some are encouraged by their managers while others aren’t. Some gain hours back every week and use that capacity for higher-value work. Some become increasingly capable at working alongside AI while others never get the opportunity to develop that skill.

And some employees are praised for using AI efficiently while others risk having their competence questioned for doing exactly the same thing.

Over months and years, those differences could matter.

Not because AI automatically creates inequality.

Because small differences in access, experimentation and recognition can compound.

Don’t solve this by telling women to use AI more

That would be the wrong lesson to take from the research.

Businesses don’t need another initiative telling women that they need to become more confident with technology.

They need to look at the environment in which people are being asked to adopt it.

Are managers explicitly giving everyone permission to experiment?

Don’t assume employees know what is acceptable. Explain which tools can be used, what information can be put into them and what responsible experimentation looks like.

Is AI training happening inside people’s working day?

If learning is optional, self-directed and expected to happen after work, access isn’t as equal as it first appears. Give people protected time to learn.

Are you teaching AI or teaching useful work?

“AI training” can sound intimidating and abstract. “Here’s how to use the tools you already have to turn these meeting notes into actions” is concrete. Start with actual work.

Are people being judged on the quality of their thinking and output, or on whether AI helped them produce it?

AI assistance shouldn’t remove accountability. Humans still need to check, challenge and take responsibility for what they produce.

But using a calculator doesn’t make an accountant less capable. Using Excel doesn’t mean the spreadsheet deserves the credit.

AI should increasingly be viewed in the same way: a tool whose value depends on the judgement of the person using it.

And do you actually know who is using AI?

A company can buy licences for 100 employees and still have an adoption problem.

Don’t measure AI transformation by how much software you’ve purchased.

Look at who is using it, what they are using it for, what is stopping everyone else and whether it is genuinely making work better.

The AI gender gap is a management problem too

The gender gap in AI adoption is real, although its size varies considerably between countries, occupations, age groups and measures of use.

There are also encouraging signs that it can close.

Women’s adoption of generative AI in the US has grown rapidly, and the gender gap in recorded AI skills has narrowed across much of the world.

So this isn’t a story about women inevitably being left behind by technology.

Nor is it evidence that women simply need to become more confident.

It’s a warning about something much more practical.

Giving people access to AI is not the same as giving them an equal opportunity to benefit from it.

Businesses have more control over that than they might think.

They can create environments where experimentation is normal rather than risky. They can give employees time to learn. They can make expectations around responsible AI use explicit. They can teach practical applications rather than technical theory. And they can make sure employees are judged on the quality of their judgement and work, rather than whether a machine helped them get there.

AI has the potential to make people more capable.

It can help inexperienced employees learn faster, remove repetitive work and give ordinary people access to capabilities that once required specialist knowledge.

That’s an extraordinary opportunity.

But technology doesn’t decide who gets that opportunity.

Workplaces do.

Sources and further reading