Is AI Really Taking Our Jobs?

In the last chapter, I was thinking about how people develop through education and experience.

But much of that development continues long after we leave school.

We learn by doing the work. By making mistakes. By working with more experienced people. And by gradually taking on responsibilities we could not have handled earlier.

That made me think differently about something I had been reading increasingly often.

AI replacing jobs.

At the same time, over the past few months, I have been spending time trying to understand smaller businesses in Singapore.

Having worked with both large and small organisations during my career, I have noticed how differently people can work in smaller companies.

Employees often wear multiple hats. Someone may be dealing with customers, coordinating with suppliers, following up on payments and helping colleagues with something entirely outside their job description.

There can be a lot to learn because the boundaries between roles are less clearly defined.

A smaller business may have technicians, salespeople and administrators, but no dedicated marketing department or data analyst.

Some responsibilities fall to the owner. Some are shared among employees. Others may simply not get done.

That made me wonder whether we are looking at AI and employment from only one direction.

We keep asking which jobs AI might replace.

But what about work that businesses never hired anyone to do in the first place?

The more I looked into it, the less satisfied I became with some of the explanations we hear.

Are we blaming everything on AI?

Every few weeks, another technology company announces layoffs.

Sometimes the company mentions AI. Sometimes the headline makes the connection. And sometimes AI seems to become the explanation simply because it is the biggest technology story of the moment.

But technology companies were already cutting jobs before generative AI became widely used.

Many had expanded aggressively during the pandemic. Then came slower growth, changing business conditions and pressure to control costs.

Now AI has entered the picture, along with enormous spending on computing infrastructure and new products.

So when a company reduces its workforce, how much of that decision is actually caused by AI?

I don't think a layoff announcement is enough to tell us.

A company may be correcting earlier over-hiring. It may be responding to lower demand. It may be redirecting investment. Or AI may genuinely have reduced the number of people needed for certain tasks.

Several things could be happening at once.

But that doesn't mean AI is having no effect.

Some of the more important effects may not appear in layoff headlines at all.

The jobs that may never be created

One piece of research caught my attention.

Stanford researchers examined payroll records covering millions of American workers. In an August 2026 update, they reported that employment among workers aged 22 to 25 in occupations highly exposed to AI was about 19% below where it would have been if it had followed the trend of comparable young workers in less-exposed occupations.

That is a substantial difference.

But the researchers were careful about what it meant.

They did not find widespread displacement across the entire economy. Nor did the findings establish that AI caused the entire employment gap.

Another observation interested me even more.

Much of the adjustment appeared to be happening through reduced hiring rather than people being dismissed.

That made me think about the jobs that may never be created.

Imagine a company that previously needed five junior analysts.

If AI allows the existing team to handle more work, perhaps it decides it now needs only three.

Nobody necessarily loses an existing job.

There is no dramatic announcement.

But two opportunities that might previously have existed no longer do.

And that brings me back to the question of development.

How does someone become an experienced professional if there are fewer opportunities to enter the profession?

We expect senior employees to exercise judgment, manage complexity and deal with customers.

But much of that ability is developed through years of doing the work.

Could we eventually need experienced professionals while providing fewer opportunities for people to become experienced?

I find that possibility more concerning than some of the layoff headlines.

And this is not simply a theoretical question.

In July 2026, Gartner reported that 22% of surveyed HR leaders said at least one business leader in their organisation had stopped hiring for entry-level roles because of AI automation.

The survey covered 110 heads of HR. It is evidence of decisions being reported by some organisations, not proof that entry-level recruitment is collapsing across the economy.

But it raises a longer-term question.

The junior work being automated is often part of how people learn a profession.

If companies remove those opportunities, they may also need to rethink how future experienced workers will be developed.

That does not mean a shortage of experienced professionals is inevitable.

But perhaps improving today's efficiency could create a problem we have not fully thought through.

At the same time, other research presents a less alarming picture.

Researchers at the New York Federal Reserve, examining job postings, found little evidence of a distinct AI-driven collapse in labour demand.

Both findings can be reasonable.

AI could be affecting particular occupations and younger workers without yet being responsible for broader weakness in the labour market.

We need to distinguish an early signal from a settled explanation.

Are we looking at the right businesses?

I then started thinking about where AI is actually being adopted.

Technology companies are among the easiest places for AI to spread.

Much of their work already happens through software. Writing code, preparing documents, analysing information and creating digital products are activities AI can increasingly assist with.

But the economy is much larger than technology.

A technician still has to repair an air-conditioning system.

A nurse still has to care for a patient.

A construction worker still has to build something.

AI may change how these people plan, diagnose, coordinate or document their work. And over time, other technologies may change the physical tasks too.

But there is a practical difference between work that already happens digitally and work that still requires someone to interact with the physical world.

I find that distinction more useful, at least for now, than simply separating white-collar and blue-collar jobs.

It also made me question whether we give too much attention to what happens inside large technology companies.

OECD figures from 2025 showed AI adoption among roughly 57% of information and communications firms, compared with about 20% of businesses overall.

Technology companies may be showing us the leading edge of adoption, not the experience of the entire economy.

And smaller businesses account for a substantial share of employment across the world, even though their individual stories rarely make international headlines.

That brought me back to the companies I had been examining in Singapore.

Could the employment impact of AI look quite different in a smaller business?

I started looking for evidence from other markets.

What if there was nobody to replace?

An OECD study published in November 2025 surveyed more than 5,000 small and medium-sized enterprises across seven countries, including Canada, Germany, Japan, South Korea and the United Kingdom.

Among surveyed SMEs using generative AI, about 65% reported improvements in employee performance.

But another finding was particularly interesting.

35% reported using generative AI to perform new tasks they could not perform before.

And 83% reported no change in their overall staffing needs.

These are self-reported findings, not proof that AI has transformed small-business productivity.

But they are close to the possibility I had been thinking about.

Imagine a small service business.

It has customers, employees, suppliers and day-to-day operational problems.

The owner may want better marketing, more systematic customer follow-up or a clearer picture of how the business is performing.

But the company may never have been large enough to justify hiring specialists for every one of those activities.

Now some of those capabilities may become more accessible.

AI could help existing employees prepare marketing material, analyse business information, organise customer communication or carry out administrative work that previously received little attention.

The business is not necessarily replacing somebody.

There may never have been somebody to replace.

That is what I find interesting.

For a large corporation, AI may make it possible to do the same work with fewer people.

For a smaller business, it may make it possible for the same people to do work the company could never previously afford.

Of course, neither outcome is automatic.

A small company still needs someone to understand the problem, organise the information, introduce a useful workflow and check whether it works.

And even if the company becomes more capable, that does not mean it will hire more employees.

It may simply operate better with the people it already has.

But that is still a meaningful change.

It suggests that the employment story may not be only about replacement.

It may also be about capabilities becoming accessible.

Faster is not necessarily more valuable

AI can already make certain tasks faster.

But what happens to the time saved?

If I can write an email in five minutes instead of fifteen, I have saved ten minutes.

Do I serve another customer?

Do I make another sale?

Or do I simply write three more emails?

A company doesn't automatically become more successful because employees produce more reports, presentations or messages.

The additional productivity has to become something useful.

More revenue. Better service. Lower costs. Fewer mistakes. Or work that the business could not previously undertake.

A task becoming cheaper does not tell us where the economic value will ultimately end up.

This matters for employment too.

If AI simply reduces the cost of existing activities, companies may need fewer people.

But if it enables new activities, reaches new customers or allows businesses to expand, the consequences could be different.

And those outcomes depend as much on business decisions as on the technology itself.

Buying a software subscription is relatively easy.

Changing how a business operates is much harder.

Where I have landed

When I started looking at AI and employment, most of what I was reading was about jobs disappearing.

I still think that deserves serious attention, particularly if younger people find it increasingly difficult to enter certain professions.

But my experience with smaller businesses has made me consider another side of the story.

Some businesses may not need AI to replace anybody.

They may use it to do work they could never previously afford.

That doesn't mean the opportunities created will compensate for the jobs lost.

They may involve different people, skills and industries.

And a business becoming more capable doesn't necessarily mean it will employ more people.

But it changes how I think about the employment question.

Perhaps we should look beyond headcount and pay more attention to what people and businesses are becoming capable of doing.

The junior position that is never advertised.

The employee whose responsibilities gradually change.

The small business that can suddenly undertake work it previously could not afford.

These changes may tell us more about how AI is reshaping work than another headline about layoffs.

AI may change the work long before it changes the job title.