Too Old to Start Again, Too Young to Call It a Day
AI is actually doing to Australian jobs - and what history says about surviving the middle of a revolution
In this article you will learn:
• What the current Australian data says about AI and employment, as opposed to the headlines
• The difference between a job being exposed and a job being automated
• What previous technological upheavals did to workers, here and overseas
• Why the aggregate story is reassuring and the individual story is not
• What is genuinely worth doing about it in midlife
Introduction: The Question Underneath the Question
Every conversation I have about AI at the moment arrives at the same place.
Not ‘will AI take my job’. The real question is quieter and much more frightening.
What do I do if I am sitting in the middle of a revolution? Too old to start again from scratch. Too young to call it a day. Twelve or fifteen working years left, a mortgage, and no appetite for burning it all down.
It is a fair question, and it deserves better than either the doom or the LinkedIn optimism. So here is what the evidence actually says.
What Is Happening in Australia Right Now
The Department of Employment and Workplace Relations published its report on AI and employment in Australia in July 2026. Its central finding is that there is no evidence of broad AI-driven upheaval in the Australian labour market. Unemployment sits around 4.4% and workforce participation is near record highs.
But it also found something more interesting. Since late 2022, employment in the most AI-exposed group of occupations grew by 5.6%, while the least exposed grew by 9.5%. Occupations well above average exposure are sitting around 2% below where the previous trend would have put them.
Nothing has collapsed. Growth has quietly slowed in the places where the tasks are exposed. That is a very different thing, and it is much harder to notice.
Exposed Is Not the Same as Automated
These two words get used as if they mean the same thing, and they do not. Conflating them is what produces most of the panic and most of the bad advice.
Jobs and Skills Australia modelled almost a thousand occupations and found that around 4% of the workforce sits in roles with high automation exposure, while about 79% have low automation exposure but medium to high potential for augmentation. Their conclusion was that augmentation generally outweighs automation, with the automation potential concentrated in routine work - clerical and administrative roles most of all.
Meanwhile the DEWR data shows that in the most exposed occupations, 43.7% of workers hold a bachelor’s degree or higher, against 14.9% in the least exposed.
Read together, those two findings say something that surprises almost everyone:
• The most educated, cognitive work has the most tasks exposed - drafting, analysing, summarising, researching.
• The most routine clerical work has the most tasks genuinely automatable.
• And the work that is hardest to touch at all is work that needs a body in a room and a relationship.
A teacher’s aide managing a dysregulated child is doing something no model can do. So is a nurse, a counsellor, a physiotherapist, a support worker. Those roles score badly in exposure models because the models look at the paperwork - but the paperwork is not the job.
There is a term gaining ground for where this is heading: hybrid intelligence. The idea is that the future of most work is not humans versus machines, but humans working alongside them - the machine handling the volume and the pattern-matching, the human bringing judgement, context and the ability to read a room. It is augmentation given a name, and if the futurists are right it is the direction most of us are travelling in.
The safest work is not the most skilled work. It is the work that requires presence.
We Have Done This Before
Here is where it gets genuinely reassuring, and it is worth sitting with.
Research published in the Quarterly Journal of Economics in 2024 built a database of job titles from 1940 to 2018 and found that roughly 60% of employment in 2018 sat in job titles that did not exist in 1940. Sixty per cent. Nobody in 1940 could have named the jobs their grandchildren would do.
Australia has run the same story. Reserve Bank data shows manufacturing employed around 26% of Australian workers in the 1960s, falling to 17% by the 1980s and 11% by the 2000s, while services rose from 63% to 84%. And here is the detail that tells you what really happened: between 1978 and 2000, manufacturing’s share of employment fell from 21.2% to 12.6%, but the actual number of people employed in it only went from about 1.28 million to 1.14 million.
Manufacturing did not empty out. The rest of the economy grew around it. That is what an evolution looks like from the inside.
And Now the Part Nobody Tells You
The aggregate always recovers. The people do not always recover with it.
During the British Industrial Revolution, from roughly 1790 to 1840, output per worker climbed steadily while real wages for working people stayed flat and the rate of profit doubled. Economists call it Engels’ Pause. Fifty years of progress in which the gains went almost entirely to whoever owned the machines. Two generations of people lived their entire working lives inside the gap.
Australia has its own quieter version. Treasury’s work on structural change points to an estimated shortfall of around 247,600 jobs in coal mining regions that never showed up in the unemployment figures at all.
That sounds impossible until you understand how the counting works. To be recorded as unemployed you have to be actively looking for work and available to start. Stop looking and you are not unemployed - you have left the labour force and disappeared from the number. So when those jobs went, people retired earlier than they meant to, moved onto a pension, went into the family business, relocated, or simply gave up after enough knock-backs. Treasury calls them hidden.
The region’s unemployment rate held up. The people did not. The Productivity Commission put it plainly enough: the effects were moderate in aggregate, but considerably bigger for particular industries and particular regions.
The same research that gives us the hopeful 60% figure also found that automation eroded twice as many jobs between 1980 and 2018 as it did between 1940 and 1980, and that new work for people without a degree shifted out of middle-paid roles and into lower-paid personal services.
The middle has been hollowing out for forty years. AI did not start that. It is simply the newest thing accelerating it.
So What Does This Mean Mid-to-Late Career?
It means the honest answer to ‘will it be fine’ is: the economy will be fine, and that is close to useless information for you personally.
What matters is not the aggregate. It is whether your particular work sits in the group the transition passes through or the group it lands on - and unlike the handloom weavers, you can actually find that out. This is the same argument I have made about traditional career models - the system-level story was never designed to answer an individual woman’s question.
Three things follow from the evidence:
• Know your own exposure honestly. Not your industry - your actual tasks. Which parts of your week could be drafted, summarised or processed by a machine, and which parts need you in the room?
• Move toward presence, judgement and relationship. Not because those things are noble, but because they are the hardest to replicate and the evidence is consistent on it.
Build in a way that respects your capacity and your season of life. Futureproofing does not mean a degree at 53. It usually means one deliberate skill layered onto twenty years of experience you already have.
How Career Counselling Supports This Shift
Working out where you actually sit is not something most people can do alone, because it requires seeing your own work as a set of tasks rather than a job title. That is exactly what career counselling is built for.
• Breaking your role into tasks and looking honestly at which are exposed
• Identifying the parts of your experience that are hardest to automate and easiest to sell
• Choosing upskilling that builds on what you have, rather than starting over
• Making the plan while you still have time and reserves, rather than after a redundancy
The Invitation
You are not too old, and you are certainly not finished. You are living through a transition, and transitions have always been survivable for the people who could see them clearly.
This week, do one thing. Write down everything you did at work last week - properly, task by task. Then mark each one: could a machine do this, could a machine help me do this faster, or does this need me?
That list is the most useful career document you will make this year, and it costs you twenty minutes.
And if you want help reading it, a Soul Strategy Call is a good place to start.
Sources
Department of Employment and Workplace Relations, AI and Employment in Australia, July 2026.
Jobs and Skills Australia, Our Gen AI Transition (Paper A: Exposure).
Autor, Chin, Salomons & Seegmiller, New Frontiers: The Origins and Content of New Work 1940-2018, Quarterly Journal of Economics, August 2024.
Allen, R.C., Engels’ pause: technical change, capital accumulation and inequality in the British industrial revolution, Explorations in Economic History, 2009.
Reserve Bank of Australia, Structural Change in the Australian Economy, Bulletin, September 2010, and The Manufacturing Sector: Adapting to Structural Change, Bulletin, March 2001.
Australian Treasury, Experiences of Structural Change; Productivity Commission, Trends in Australian Manufacturing.