Australia's workplace relations minister, Amanda Rishworth, has released a draft government guide that urges employers to tell workers plainly how AI will change their jobs and to offer them new roles. It arrives as Qantas, NAB and Canva announce cuts partly linked to AI, and it has no legal force. The more useful question it raises is what happens to the time AI frees up, because that is where the good outcome and the bad one split.
Why the same hours get heavier
AI does the easy work first. Take a clerk at a mid-sized distributor who checks 120 supplier invoices a day. About 100 match their purchase orders and take two minutes each. About 20 have a problem and take fifteen minutes each. That is a little over eight hours.
Once AI clears the matching invoices, about 200 minutes of the day disappear, and the clerk is left with five hours of problem cases and no easy ones in between to catch their breath. They also have to check the AI's work, which is the extra step the minister warned employers about when AI is added without redesigning the job.
Now the company has three options with those 200 minutes. It can cut the team by about 40%, shorten everyone's day, or keep everyone and let the invoice pile grow until the day is full again. The third option is the easiest to do and the hardest to see, and it is the one the guide's warnings about intensified or unrealistic performance standards point at. Once work flows through software, every person's output can be counted by the hour, so that standard is easy to set and easy to raise.
Employers who share the hours will want something back
In 1926 Ford, the American carmaker, moved its factory workers from a six-day week to five. Ford did not do it out of kindness alone. Its leaders wanted workers who were rested and had leisure time to spend, and who would, among other things, buy cars. The company expected something back too: more effort in the shorter week, and output did rise. The law that set the 40-hour week for everyone came twelve years later, in 1938, and was phased in over two years.
That suggests how it will go with AI. Some employers will share the hours voluntarily, and they will be the ones who need something in return: the people who handle the hard cases are the ones they cannot afford to lose, and a shorter week is a cheap way to keep them. Employers that face no such pressure will take the hours as output or as savings. A voluntary guide helps the first group explain what they are doing. It does little for the second, which is why the Greens want the same protections written into law.
The entry-level problem is a training problem
The guide singles out entry-level work as the first to be automated. The Reserve Bank has seen little evidence so far of AI disrupting Australia's labour market overall, but it sees tentative signs of weaker outcomes for younger workers and in jobs such as clerks, programmers and call centre staff.
Entry-level work is also where people learn. A junior clerk who spends two years matching invoices comes to know which suppliers are careless and which mistakes look alike. If AI takes that work, the hard cases land on experienced staff, and nobody is being trained to replace them. A firm that automates its junior tasks has to build another way to learn, such as having a new hire review every problem case the AI flags, with an experienced colleague alongside.
A manager preparing an AI rollout can do the part the guide asks for before it starts: decide what happens to the hours it frees, write it down, and tell the team. A worker can ask the same question, because the plan the minister wants employers to give staff is, in practice, an answer to it.