According to Finder’s Consumer Sentiment Tracker, 1 in 3 employed Australians, roughly 4.2 million people, worry they will lose their job to AI. Nine percent say it will “definitely” happen. The anxiety is highest among Gen Z, where 38% are concerned. That fear is not irrational. The Australian Government’s own skills body, Jobs and Skills Australia, estimates up to 1.3 million workers may need to transition roles by 2030. Displacement in routine administrative and clerical roles is real. This article is not going to tell you there is nothing to worry about. But it is going to make a specific argument: the conversation about job replacement is aimed at the wrong category of work. There is a different category worth talking about, and almost nobody is.
The work nobody trained for
Ask a financial planner what they became a financial planner to do. Not the answer for the interview. The actual answer. It is not to re-key client data before every meeting. It is not to draft compliance documentation on a Friday afternoon. Research from Docupace suggests financial advisers spend less than 20% of their time in actual client meetings. Somewhere between 35 and 45% goes to analysis, compliance documentation, and admin.
Nobody trained for that. Nobody wanted it. It accumulated.
The same pattern holds across industries. AMP Bank research from March 2025 found that 3 in 5 Australian small business owners sacrifice personal time to manage financial admin, and 42% have missed business opportunities because of time lost to it. A Sage study published in May 2025 found that small businesses collectively lose the equivalent of 13 months of work to admin in a 12-month year (international research, UK and global sample). Global research from Smartsheet across 10,000 workers found that over 40% spend at least a quarter of their work week on manual, repetitive tasks.
This is not work. It is overflow. It is the thing that arrived uninvited and never left.
What the research actually says
The loudest claims about AI, the ones generating the most fear, treat automation as inevitable and total. The peer-reviewed research is more specific.
Brynjolfsson, Li and Raymond at NBER studied AI assistance in customer service and found a 14% average productivity gain. The gains were concentrated among less-experienced workers. Senior people with established judgment saw smaller improvements. The implication is the opposite of what the headlines suggest: AI floors up novices rather than replacing experts. Skilled judgment remains the irreplaceable variable.
MIT Sloan published research in March 2025 showing the share of “human-intensive tasks” that cannot be effectively done by machines alone actually grew between 2016 and 2024, even as AI capabilities expanded. More AI in the world, more distinctly human work at the core.
Economist Daron Acemoglu’s widely cited NBER paper, “The Simple Macroeconomics of AI”, argues that even under optimistic assumptions, truly automatable tasks are a smaller share of value-added work than the headlines claim. Routine cognitive tasks are the exposed layer. Judgment-intensive and relational tasks form a much larger economic core.
Jobs and Skills Australia, the federal government’s own workforce body, concludes that augmentation outweighs automation for current generative AI. The displacement risk is concentrated in routine administrative and clerical roles. Not in roles where human judgment, relationships, or trade knowledge are the product.
Three roles and what AI actually touches
The financial planner. The work at risk is the compliance draft, the meeting prep document, the report rebuilt from three different exports. The relationship is not at risk. The planner who understands the client’s situation, who can read the room and adjust the recommendation, who has spent years building trust: that is not automatable. AI takes the four hours of admin before the meeting. The meeting stays human.
The tradie. The work at risk is the missed call while on the tools. The inquiry sitting in voicemail until 6pm. The job that went to someone else because the phone was not answered. The trade skill, the knowledge of what a job actually involves, the judgment about whether a quote is right: that stays. AI answers the phone. The tradie does the job.
The brand founder. The work at risk is the content backlog. The post that never went up because the week ran out. The document that needed to exist before the next sale but kept getting pushed. The voice, the vision, the understanding of the audience: that stays. AI writes the first draft. The founder shapes what it becomes.
In each case, AI absorbs the overflow. The work the person trained for, the reason they entered the field, remains theirs.
The design question
This is not automatic. AI built to displace is designed differently to AI built to augment.
Acemoglu and Johnson make this point directly in their 2026 Brookings piece, “Building Pro-Worker AI”: the outcomes we see are not technologically determined. They are design choices. AI that automates pushes humans out. AI that absorbs admin lifts humans up, particularly those earlier in their careers, which matches what Brynjolfsson’s research found.
The distinction matters when you are deciding what to build. Every product Upgraded makes is designed around one question: what is the work this person never wanted to do, and how do we absorb it? Not: what role can we replace?
The answer to that question determines whether AI takes from people or gives back to them.
The work nobody trained for is the right target. It is the category that has been silently consuming the hours that should have gone elsewhere. That is where AI belongs.
If you are trying to protect time for the work that actually matters, that is what this is built for. Start with a conversation.