AI Jobs Australia Has Two Official Counts. One Counts What Could Change, the Other What Has.
The exposure map says 45% of North Sydney's office tasks could change. The ABS survey says 82% of businesses felt nothing. The gap between the two is the story.
North Sydney holds the most AI-exposed jobs in Australia, and the first official count of what AI has actually done to jobs says almost nothing has happened yet.
SGS Economics and Planning put the number at 45.3 per cent: the share of tasks done by people working in North Sydney that its researchers scored as exposed to AI automation, covering 80,632 workers. The City of Sydney comes in above 45 per cent, Melbourne at 44.5, Ryde at 43, and Brisbane’s CBD and inner suburbs at about 45. Alison Holloway, the consultancy’s principal, and senior consultant Jayden Perry built the map from 2021 Census employment data and Jobs and Skills Australia’s exposure scores. They wrote that the disruption “looks set to concentrate instead in the professional heart of our cities”. Exposure, they cautioned, measures tasks that could change, not jobs that will disappear.
Three days later the Australian Bureau of Statistics published its first count of what has happened. “About 82 per cent of businesses reported that AI did not impact the level of job vacancies in August 2026,” said Sean Crick, the bureau’s head of labour statistics. Another 3.8 per cent said AI had reduced their vacancies, and less than one per cent said it had increased them. The effect concentrates where the map would expect it. Sixteen per cent of information, media and telecommunications businesses said AI had cut vacancies, against 8 per cent in wholesale trade and 6 per cent in manufacturing. Electricity, gas, water and waste services went the other way, with 10 per cent saying AI had added vacancies.
The search phrase is AI jobs Australia, and the two counts answer it in different tenses: one says what could change, the other says what has. The map counts tasks, not jobs, and applies one national exposure score to each occupation wherever the work is done. The survey counts vacancies, not jobs, and asks employers. Thirteen per cent said they did not know what effect AI was having, and businesses with four or fewer staff are left out entirely. Total vacancies stood at 325,000 in August, 31 per cent below the May 2022 peak. The hiring slowdown is real. AI’s share of it is what neither count can yet show.
Employment Minister Amanda Rishworth told the Sydney Morning Herald there was “no evidence so far of widespread AI-driven disruption to the labour market”. Her own department’s February note said unemployment in the occupations most exposed to AI had risen sharply against the least exposed, a gap at a level not seen in 20 years. Its July monitoring report found the same faint signal and warned against reading it as proof. Occupations in the most-exposed fifth grew 5.6 per cent since ChatGPT arrived, against 9.5 per cent for the least exposed. The department’s own instruction is that the result is “justification for ongoing monitoring rather than clear evidence that AI has reduced employment”. Richard Denniss of The Australia Institute, testifying to parliament’s AI committee in October, put it the other way: “the business model of AI is to generate mass unemployment”.
Jobs and Skills Australia spent 27 October correcting a number the debate had already adopted. Its statement said media reporting had turned a modelled scenario into 139,000 jobs lost to AI. The scenario showed 106,000 more workers in trades and technical occupations and 139,000 fewer in professional occupations than baseline growth over the next decade. The agency added that the scenario did not consider AI at all. It never counted jobs lost.
The department’s report found the most-exposed fifth, general clerks, programmers, accountants and receptionists, is majority female and 43.7 per cent degree-qualified. The least-exposed fifth is 69.5 per cent male and 14.9 per cent degree-qualified. Software developers, the occupation most often named in “AI took my job” stories, employed 199,000 people in February, up 25 per cent since November 2022. The squeeze is at the junior end. Barracuda found 45 per cent of Australian firms had replaced or removed entry-level cybersecurity roles with AI, against 33 per cent for senior roles, and the advertising industry’s own talent review put entry-level roles at 1 per cent of vacancies. A study of 57,134 job ads found only 4.9 per cent of Australian entry-level roles involve AI at all.
Amanda Sumner, a business relationship manager at Go Computers in Australia, wrote in October that her workload had not fallen since she started using AI. “My workload hasn’t decreased. If anything, it’s increased,” she wrote. “AI isn’t doing my job. It’s helping me do more with it.” Inka Magnaye, a voice-over artist, shared a message from her mother in October: “I just lost a voice over job to AI.” Voice-over work is one of the few occupations Jobs and Skills Australia found had already seen less demand. The Finance Sector Union, speaking for the 45 Commonwealth Bank workers whose roles were declared redundant for an AI voice bot, said members had told it the bank’s claim about falling call volumes “was an outright lie”. Rishworth’s “no evidence” and the workers’ accounts are both true, because they count different things. The survey counts vacancies. The workers are talking about jobs that already exist.
Commonwealth Bank told 45 customer-service workers in July 2025 that their roles were redundant because a voice bot had cut calls by 2,000 a week. The union disputed the claim, saying volumes were rising, and took the case to the Fair Work Commission. Twenty-three days later the bank reversed the decision and apologised, saying its assessment “did not adequately consider all relevant business considerations”. The roles were not redundant. A survey of 2,036 workers in the US and Britain found about one in six of those replaced by AI had later been asked back. From 20 October, anyone lodging documents with the Fair Work Commission has had to disclose whether generative AI was used to prepare them.
CSIRO found the opposite pattern among the firms that adopted AI early. Studying job advertisements from more than 4,000 Australian firms, it found AI adopters posted 36 per cent more non-AI job advertisements over time than comparable firms that had not, after accounting for firm size, industry and location.
