An exterior view of Australian data centre operator NextDC's Artarmon facility, Sydney. Photo by Saeed KHAN | AFP via Getty Images

CANBERRA — Artificial Intelligence could make legal advice cheaper and courts more expensive. It could make retailers more efficient in day-to-day operations while increasing workers’ compensation claims. Those who build the infrastructure to power these technologies will likely compete with the struggling residential housing market for electricians, electricity and water.

These are three of the more surprising economic impacts of AI, outlined in submissions to a Joint Select Committee on Artificial Intelligence, due to report back in November.

They show AI does more than prompt productivity gains and job losses: It moves costs around the economy, so that as some organizations save money through AI adoption, others can be left to pick up the tab.

Here are nine instances where AI is having, or could have, surprising economic impacts:

The first AI job losses may not look like job losses

What if people lose their careers and it doesn’t show up in unemployment figures?

Ramp’s chief economist, Ara Kharazian, see freelancers as the front line of AI-induced job losses. The U.S. firms which spent the most on freelancers pre-generative AI have cut their costs by 97 percent when they replaced that freelance support with AI support.

It’s easy to replace freelancers: no need to announce layoffs, negotiate severance or restructure a department. The organization simply stops paying people for tasks such as logo creation, translation or coding.

Cheaper lawyers, busier courts

As AI reduces the price of accessing legal services, more people can uphold their rights by pursuing claims. When those claims are spurious or poor quality, the private productivity gain becomes a public-sector capacity problem.

Maurice Blackburn, Australia’s largest plaintiff law firm, argued to a previous New South Wales government inquiry that AI can shut out of justice today for price reasons. “AI will let law firms deliver high-quality services at scale,” it said.

That doesn’t mean law firms aren’t themselves at risk: Legal AI tools can cut them out of some markets, by offering free tools, for services such as drafting and reviewing contracts.

For courts and tribunals this may mean a surge of applications including from self-represented litigants. The Fair Work Commission has seen this in the burgeoning number of unfair dismissal cases, and courts are concerned that AI is inserting low-quality or outright false material into submissions, wasting court time.

AI could kill the lazy savings account

By removing friction from an action, AI can transfer value from incumbent businesses to consumers.

Meta’s launch of Muse, its personal AI assistant, is bringing attention to the ways people can use AI agents to easily shift their money into competing lenders. Banks usually offer depositors less than 1% interest on savings, while some non-bank firms provide rates upwards of 4%.

One person’s market competition is another bank’s nightmare: If AI use causes their cheap source of funding to dry up.

Ethan Bloch, who handles personal finance for OpenAI, pushes back on the idea that the rise of agents will radically shift the calculus for banks. “Consumers want a name brand they can trust with their life savings,” he said in a post on X.

Will AI efficiency generate workers compensation and sick leave costs?

A 2025 survey from the SDA retail workers union, of more than 11,000 retail workers, found 72% reported burnout and 77% were assessed at high risk of psychosocial harm.

Their fear is that AI-powered rostering and task allocation, and employee monitoring, could deliver short-term efficiency offset by workers compensation, sick leave and other mental health costs.

The Electrical Trades Union warned that AI-enabled workplace surveillance could lead to “unsustainable work intensification,” in a submission to a separate Senate inquiry on data centers and AI.

NSW passed digital work system reforms to work health and safety laws in February 2026, making employers legally responsible for health and safety risks caused by AI and automated rostering tools. Victoria is considering similar laws, but business lobby groups and tech companies are appalled at measures they say will blunt the productivity improvements of AI.

AI is physical and it’s competing for your electrician

Data centers compete with housing, defense and energy transition companies for electricians and engineers — demonstrating that AI is a set of tools with a physical footprint, and human impacts.

To cope with that the Australian Academy of Technology and Engineering explicitly wants data center workforce requirements incorporated into national skills and infrastructure planning. The academy calls for public reporting of energy use and both direct and indirect water use.

The Business Council of Australia supports expanded AI infrastructure while recommending that data centers pay the costs of grid connections and network upgrades they trigger.

A September consultation paper from Department of Prime Minister & Cabinet cites energy regulator calculations that data-center electricity use could grow seven-fold over the next decade. Sydney Water estimates that data centers could consume up to 20% of Sydney available drinking water by 2035.

The Albanese government proposed a national framework with minimum requirements for large data centers including rules on energy usage, adopted by the national cabinet in August. NSW has a separate policy offering fast-track approval for data centers that pay for their connections and invest in renewable power.

Data Centres Australia has warned if the bar is set too high, investment may shift to other countries.

The great AI train carriage pile-up

AI has upstream effects in the economy, well before it touches consumers. One example is a shortage of railcars is set to slow America’s artificial intelligence boom — and make electricity more expensive.

The need for specialized train cars capable of moving heavy-duty power equipment has soared with the data center boom, and there aren’t enough of them. Siemens Energy, GE Vernova and Mitsubishi Heavy Industries are all reporting record-high orders, with years-long waiting lists for components such as gas turbines and transformers.

In the U.S. securing a rail route that is wide enough and lined with bridges strong enough to bear the heavy loads can take up to a year. Then there are the train cars themselves. The average train car has four axles, but specialized cars can have 12 to 36 axles.

Antoine Vagneur-Jones, an analyst who tracks supply changes at BloombergNEF, sees pinch-points all along the AI supply chain. “You’ve suddenly got this bottleneck that’s migrated from chips to now power equipment to the grid,” he said.

Fraud controls impose costs on legitimate businesses

Australian Chamber of Commerce and Industry warns that increasingly sophisticated scams and scam-detection systems create the risk that a business may lose access to customers or platforms because a detection system makes the wrong call, a decision small firms may lack the resources to challenge.

A recent Auditor-General report described the cost of false positives in government screening. One government AI fraud detector was wrong nearly four times out of five.

Of 82 potential Medicare non-compliance matters reviewed from an AI-enabled model, 18 were true positives and 64 were false positives, a 22% true-positive rate.

The department stopped using the model after describing it as a drain on manual-review resources.

AI can create customers while blowing up creative markets

The Communications Department told the inquiry that AI-generated summaries can intercept readers, reduce clicks to news websites and divert financial value from publishers to AI companies.

The Local & Independent News Association calculates that roughly 300 Australian newsrooms have closed or downsized since 2019. With smaller publishers more reliant on referral traffic and having less capacity to negotiate or enforce licensing deals, LINA says AI systems that retrieve and summarize news could reduce that traffic further.

The Association of Artist Managers warns that a flood of AI-generated music could dilute the revenue pool for human artists, including income from music licensing. These submissions make copyright an issue of market access and revenue distribution, as well as rights over AI training material.

AI could also reduce costs by shifting spending toward prevention

The Environment Institute of Australia and New Zealand, an environmental professionals body, says AI can speed up climate-risk assessment and the analysis of large environmental data sets.

Its submission to the AI inquiry argues this could help governments, businesses and investors fund mitigation before damage occurs, shifting some spending from emergency response and recovery to prevention. EIANZ says expert verification remains necessary.

Home

AI promises to save Australia money. Here are the costs nobody talks about.