ABS data reveals trends on which jobs are being stolen by AI
Anyone ranging from forklift drivers, crop farmers, call centre operators as well as accounting types might like to broaden their skill base to avoid being marginalised by AI.
Was there ever a topic more anticipated than the assumed negative impact of artificial intelligence on the workforce? Perhaps the effects of Covid in early 2020, or of the Y2K bug in the late 1990s, set a precedent.
However, in many ways Covid wasn’t as devastating (in Australia) as was first speculated, and the same is true for Y2K. In both cases the predictors of Armageddon would later argue that it was their timely warnings that secured corrective action. Indeed, by 2030 we might look back and wonder what all the fuss was about regarding AI.
The threat of new technology “taking” the jobs of workers is hardly new. More than 200 years ago the Luddite movement in northern England railed against the introduction of threshing machines that separated seeds from stalks. A threshing machine could capture crop value (seed) more efficiently than a farm labourer working by hand.
Displaced, angry agricultural workers subsequently gravitated to the cities, where they worked in factories. Some made it to the colonies.
The trouble with AI as an agent of workplace change is that it’s hard to identify commonly held jobs it will replace.
The Luddites could point to a threshing machine replacing an agricultural labourer. Robots have replaced workers on assembly lines. Tram conductors have been replaced by touch-and-go ticketing. Checkout operators have been replaced by self-service checkouts. And airlines have taught passengers how to check their own baggage.
But when the tech cognoscenti talk (glowingly) of the effects of AI they invariably revert to generalities such as “any task that is repetitive and/or that can be standardised.” Or they point to industries likely to be affected such as accounting, law, advertising, logistics. This is what makes AI scarier than previous workplace revolutions: we don’t know or least we haven’t measured the jobs “lost to AI”.
Haves and have-nots
And when tech gurus talk about AI replacing jobs, does this mean that workers will be forced to change jobs (to, say, care jobs), so there is no net reduction in the workforce? Or does this mean that a subset of fewer, but far more productive, workers will command a greater share of work and therefore of wealth?
If this is the case, then surely such an outcome of AI will be a divide between those in work and those not in work? And if this is indeed then an unstoppable outcome of AI by the early 2030s, then at some point there will be tension between the haves and the have-nots.
For the moment, however, let’s focus on measuring the impact of AI on the workforce. The adoption of AI will eventually deliver greater productivity per worker (otherwise why use it?). Therefore the number of workers in some jobs must diminish over time.
The diminution of AI-impacted jobs will be first detected, I think, in a quarterly Labour Force survey published by the Australian Bureau of Statistics. In fact I think there is evidence already of where and by what margin today’s workforce is being reshaped by AI.
The ABS Labour Force survey includes estimates of the number of workers across 470 jobs comprising Australia’s workforce of 14.6 million workers. This database is based on the five-yearly census, which is updated by survey. However, a survey of the workforce across hundreds of occupations is subjected to variances due to the need for sampling.
Then again, this survey has been conducted by the ABS for 40 years so I assume it has its modelling as accurate as possible. Drawing on the Labour Force database I have tracked the rise and fall of all jobs over the past 12 months and over the last six years. These time frames should pick up short-term and medium-term trends shaping the workforce.
There is always growth and loss across the components of the workforce as the economy flexes to absorb economic opportunity. However, by considering jobs lost over the past 12 months and over the past six years, there should be some evidence of the impact of job losses triggered by AI.
Over the six years to May 2025 the Australian workforce increased by 1.8 million jobs (about 300,000 pa) and in that time most job loss, according to the census-and-survey methodology, was 94,860 workers working as checkout operators and office cashiers.
I don’t see this job loss as being entirely AI-induced. I see this job loss being caused by changes in consumer behaviour learnt during the pandemic. There may be an element of AI present in checkout facilities now being able to distinguish between, say, a pear and an avocado.
Other notable job losses over this six-year period included personal assistants down 18,046 jobs, forklift drivers down 16,614 jobs, and crop farmers down 15,460 jobs.
I suspect that working from home had a deleterious impact on demand for PAs because working from home management had no access to support and had to work it out for themselves!
The loss of forklift drivers is unusual given the surge in industrial buildings triggered by online shopping, again during the pandemic. Perhaps warehouses relying on robots might be managed by some aspect of AI.
Who’s losing out?
And crop farmers may be selling out to aggregating neighbours who see advantage in economies of scale. Indeed, there may be aspects of AI in the technology needed to manage the crop farms of Australia.
These workforce trends over the past six years reflect longer-term trends that may have been accelerated by the pandemic (eg, farms getting bigger). However, it is over the 12 months to May 2025 where there is perhaps the best evidence of how AI is shaping the job market.
Within the top 20 job losses over the past 12 months there are several jobs likely to have been marginalised by AI, including management and organisational analysts (consultants) down 21,560, accounting clerks down 14,629, bookkeepers down 10,911, bank workers down 9699, and call or contact centre and customer service managers down 9116.
To be fair, the bank worker (teller) workforce has been declining for many years due to the rise of ATMs and the popular convenience of internet banking. No need (or less need) to go into a physical bank!
Plus, it is also clear that today’s service providers would much prefer customers interact with an AI-driven chatbot than with a (more expensive) call centre operator. Video may have killed the radio star but AI seems to be strangling the call centre operator.
And as for consultants, accounting clerks and bookkeepers, it is likely that part of their remit can be, and is being, quietly siphoned off by AI. Data analysis, reconciliation of monetary flows and even a range of compliance tasks can be delegated to AI.
The Labour Force survey data isn’t perfect. The number of workers in each job can fluctuate from quarter to quarter based on survey sampling. However, it is the best source we have to capture and to measure the impact of AI.
I think the evidence shows that by mid-2025 AI isn’t devastating any specific jobs, rather it is quietly siphoning off work tasks from a range of activities from call centre operator to management consultant.
In this respect AI isn’t so much a job thief, or a job killer. Rather it seems to be quietly commandeering tasks that can be mightily helpful if you have other things of value to do.
The problem is if the task that AI does best is your core (and sole) competency, such as customer interaction, such as reconciling inflows and outflows, such as adjusting fertiliser distribution across a paddock, such as selecting a range of products from different parts of a warehouse shed.
Forklift drivers, crop farmers, call centre operators as well as consultants, accounting types and bookkeepers are those who might like to broaden their skill base to avoid being marginalised by the AI task thief.
Bernard Salt is founder and executive director of The Demographics Group; data by data scientist Hari Hara Priya Kannan
More Coverage
