AI-caused job losses have been a topic of concern for years since AI was being developed years ago. It was a concern of mine that I started thinking about all the way back in 2015 after I went to Canberra to hear what’s been discussed around AI and what was being researched at the time.
As a responsible AI advocate, the breakthroughs we have had with AI so far are an exciting step to open up humanity’s potential. However, the job displacement concerns have fallen back into my radar in the last few years as AI becomes mainstream.
Afterall, responsible AI is not only concerned with how AI is embedded into organisations or how it is used by consumers. It is also looking at AI technology from a macro lens and highlighting the potential severe risks and side-effects that may become reality in the future.
However, not everyone is convinced that AI will cause widespread job losses. Some are even convinced that more opportunities will be created than destroyed.
So with that, there are two sides to the debate when we ask “Will AI cause long-term widespread job losses?”.
Affirmative: AI will cause widespread job losses
Goldman Sachs data shows AI is reducing monthly payroll growth by 11,000 to 16,000 US jobs per month, with April 2026 marking the highest single month. The net figure is made up of about a 25,000 monthly jobs decrease from AI-driven “substitution”, and about 9,000 monthly jobs increase from data centre and infrastructure construction.
This is despite many artificial intelligence rollouts across a number of companies in 2025 experiencing a failure rate – initiatives that do not deliver on what was promised – of 80% and 42% of companies abandoning their AI initiatives completely.
It’s important to note here that although this jobs data is undeniable at this point:
- they signal lower job growth numbers, not out right job losses
- the majority of those job growth reductions are in the white-collar sector hitting entry-level and young white-collar workers hardest.
Negative: AI will not cause widespread job losses
In June, Jeff Bezos spoke at the annual VivaTech conference in Paris about his role in his new startup “Prometheus”, who are building an “artificial general engineer” that aims to significantly reduce the engineering effort of sophisticated engineering projects from 10 years to something significantly shorter.
During the talk, a point of concern was raised by Mike Massimino:
“I know there’s a lot of concern that many people have, including many smart people, that AI is going to make humans redundant and so on.”
Bezos rejected this view, and responded by saying that AI will cause labour scarcity, not mass unemployment.
Bezos explains that human imagination is limitless, we are only constrained by what we are capable of doing. When the constraints on what is possible is removed and we’re only limited by our imagination, the demand for human effort expands into new territories.
We’ve heard similar kinds of views expressed by other prominent tech experts over the last few years: enablement will unlock more abilities and lead to more jobs (than they replace).
How do we interpret both of these views?
Those are the two perspectives. Seemingly in contradiction with each other, even if both perspectives are true from the angles that they talk from. How well do these perspectives translate to long term macro-trends in the context of employment at scale?
Understanding the Affirmative view
The affirmative view directly talks to the impacts of AI in the workforce in the US’ white-collar sector. It is essentially reporting on what has happened and what onward effects are likely to happen.
We have also seen jobs lost in manufacturing and warehousing due to technology, notably in Amazon warehouses both from years ago and more recently, as well as more recently in Waymo and its self-driving cars, where drivers of Uber and other ridesharing services would have previously filled those driving roles.
As AI capabilities get more and more advanced and get implanted into more SaaS tools and eventually in robotics, we are likely going to see this trend not only continue in the sectors that are already affected, but branch out into new sectors over the next few decades.
Furthermore, even if some displaced workers end up acquiring new jobs, not all of them will. Displaced workers have to contend with:
- Upskilling or retraining, into a field that has sufficient shelf-life before this field eventually (or soon after) becomes prone to job displacement
- Supply vs demand: an increasing number of workers competing for smaller number of open job positions
Understanding the Negative view
The negative view however is much less convincing. The enablement that Prometheus is hoping to unlock may happen, but Bezos’ argument does not explain how that would translate to the workforce at large. In short, Bezos’ argument does not adequately address the concerns about widespread job displacement by AI.
It’s important to note here that Bezos’ part in Prometheus may colour his opinions in this debate. And to some extent his viewpoint is true: Prometheus will in theory drastically reduce the engineering effort and cost of future engineering based projects. We are likely going to see similar results in other AI initiatives in the near future.
However, Bezos’ argument doesn’t expand on how AI will cause labour scarcity. Let’s assume it is easier for someone with an idea to turn their idea into reality:
- It still requires people to have some level of understanding of product development. The majority of people do not have this skill
- It still requires people with material or financial resources to construct the product of the engineering solution. Notwithstanding seeking out investment or getting loans, the majority of people do not have those means
- Not all products created will end up solving a big enough problem; resource and time wastage is inevitable. A product might seem cool in your head and prospective customers might even cheer on, but you need customers to pay for it
- There will be increase in competition, and some products (or the businesses that build them) despite positive efforts will lose to more superior ones
To bring these points together, we need to ask:
“Is AI advancement, despite having obvious wins along the way and creating some employment, going to lead to continued and sustained gainful employment for the billions of people on this planet – for the adults of today and tomorrow?”
This is the heart of the concern that does not seem to be adequately addressed in many responses that reject the AI-driven “substitution” hypothesis.
How should society manage the AI transition?
It would be irresponsible for us to wait for these technological advancements to come into fruition, only to find that they have not delivered on the promises made of unlocking new employment opportunities for workers – or at least not on the scale that addresses the number of workers displaced by AI beforehand.
In the last 10 years, experts and policy makers have discussed and experimented with Universal Basic Income (UBI).
This is an obvious stop-gap for addressing the immediate needs of displaced workers, as it gives them purchasing power in the absence of gainful employment.
UBI may be able to take on increases in unemployment, but perhaps not significant job losses, and also maybe not forever. The discussion needs to go beyond plugging holes in employment with welfare programs.
Experts and policy makers need to put their heads together and figure out answers to these two key opening questions:
“What would society look like when AI and technology, on its own, is capable of building everything we need, but neglect the side-effects of AI and technology’s advancement on society?”
and:
“What would society look like when AI and technology, on its own, is capable of building everything we need, and we’ve properly managed all of the risks associated with AI and technology’s advancement along the way?”
In future posts, I plan to dig deeper into these key questions, and cover any further questions that are found as part of this discovery work.
Stay tuned.
