Australian businesses are cutting jobs. Is AI ready to replace the work?

Almost one in three Australian employers expected to make workers redundant in the final quarter of 2025.

That figure comes from the Australian HR Institute's December 2025 Quarterly Work Outlook, which surveyed more than 600 senior HR professionals and business decision-makers.

At the same time, AI adoption is accelerating. Businesses are automating administrative work, customer enquiries, reporting, document preparation and parts of recruitment. Productivity gains are beginning to appear, and some employers are reducing headcount or choosing not to refill vacant roles.

Put the two trends together and the conclusion can seem obvious: AI has arrived, jobs are being cut, and businesses that do not automate will be left behind.

The truth is more complicated.

AI and automation are ready to take over a meaningful amount of repetitive work. They are not ready to take responsibility for an entire business function without supervision. The distinction matters — especially for small and medium businesses that cannot afford a failed implementation or the loss of experienced staff.

What the Australian employment figures actually tell us

AHRI reported that 30% of employers intended to make redundancies during the December 2025 quarter, up from 27% in the previous quarter. It was the second-highest result since the survey began in 2023.

However, this was not simply a story of businesses shrinking.

Seventy-one per cent of organisations also planned to recruit, and 91% of the organisations planning redundancies intended to hire during the same quarter.

Businesses are not necessarily replacing every employee with software. Many are changing the shape of their workforce: removing some roles, redesigning others and recruiting people with different skills.

The same AHRI survey found that 82% of business and HR decision-makers believed AI was making — or would make — their organisation more efficient and productive. Eighty-one per cent said it allowed employees to focus on more strategic work.

Yet the survey's findings on entry-level employment were surprisingly positive. Forty-one per cent of respondents reported an increase in entry-level roles due to AI, while 19% reported a decrease.

The message is not that AI automatically destroys jobs. It is that AI changes which tasks businesses need people to perform.

There is evidence of genuine displacement

A Morgan Stanley study published in February 2026 provides a clearer view of direct workforce effects.

The survey covered 935 executives from companies in Australia, the United States, Germany, Japan and the United Kingdom. All participating companies had used AI for at least a year and operated in five sectors considered particularly exposed to near-term disruption.

Globally, respondents attributed an average productivity increase of 11.5% to AI. They also reported that 11% of positions had been eliminated and another 12% had not been refilled. New hiring offset much of that reduction, leaving a net decline of approximately 4%.

Australian respondents reported some of the strongest productivity improvements. The underlying research indicated that Australian companies had eliminated or not refilled approximately 24% of positions while making new hires equivalent to 20% — also producing a net reduction of around 4%.

Those figures deserve attention, but they need context.

This was not a representative survey of every Australian business. It deliberately examined organisations in AI-exposed industries that had already been using the technology for at least 12 months. The results show what may happen among relatively advanced adopters, not what is already happening in every café, clinic, construction firm or professional practice.

They also show that workforce change is occurring in both directions. Jobs are being removed, but new positions are being created and existing employees are being retrained.

Is AI genuinely ready to replace employees?

Sometimes. But it is usually more accurate to say AI can replace tasks rather than people.

A job is a bundle of activities. An administrative employee might schedule appointments, prepare reports, answer routine questions, notice an unusual client request, calm an upset customer and remind the owner about an issue that has not yet become urgent.

Current automation may handle the scheduling, reporting and reminders extremely well. It may answer predictable questions and gather the information needed for a human to respond.

It is less reliable when a situation is ambiguous, emotionally sensitive, commercially unusual or outside the information it has been given.

This is consistent with the Jobs and Skills Australia Generative AI Capacity Study. Its analysis found that only around 4% of Australian workers were in occupations with high exposure to AI automation. By comparison, most occupations had some potential to be augmented by AI.

In plain English: AI is likely to change a large number of jobs, but completely remove far fewer of them.

Where automation is ready now

Automation is already dependable when the work is repeatable, the rules are clear and mistakes can be detected.

For a service business, that may include:

• Sending an immediate text message after a missed call.
• Capturing and qualifying a new website enquiry.
• Adding customer details to a CRM.
• Sending appointment confirmations and reminders.
• Following up an unanswered quote.
• Requesting missing onboarding documents.
• Reactivating older leads.
• Preparing a first draft of a routine email or report.
• Summarising meetings and creating follow-up tasks.
• Moving information between systems without manual copying.

None of these examples requires a business to hand control to a mysterious machine. They are defined workflows with clear triggers, actions and escalation points.

A missed call arrives. A message is sent. The caller replies. Their details are recorded. A person is alerted when human attention is required.

That is practical automation. It is measurable, testable and available now — the same five examples we've built and priced for real Australian businesses in AI Automation for Small Business in Australia: What Actually Works in 2026.

Where businesses should remain cautious

The case for full automation becomes weaker when the work involves judgement, trust, accountability or consequences that are difficult to reverse.

AI should not be left unsupervised to make decisions about employee performance, legal obligations, clinical advice, creditworthiness or unusual customer disputes. It can assist with gathering information and preparing recommendations, but a suitably qualified person should remain accountable for the result.

Businesses should also be cautious when:

• The underlying process changes from one employee to another.
• Customer or operational data is incomplete.
• The system cannot explain why it made a decision.
• Errors may create safety, privacy or compliance problems.
• The business has no reliable way to monitor performance.
• Staff are expected to trust the system without being consulted or trained.
• The proposed savings exist only in a vendor's presentation.

The technology may be impressive while the business process remains unready.

Deloitte's 2026 State of AI in the Enterprise findings illustrate that gap. Although 61% of Australian respondents reported efficiency or productivity improvements, only 28% had moved at least 40% of their AI pilots into production. Just 12% said generative AI was already transforming their business and industry.

Experimentation is widespread. Reliable implementation is much less common.

The risk of cutting people too early

Removing a role before an automation has been properly tested can create costs that do not appear in the original business case.

An employee may be performing dozens of small tasks that were never included in their job description. They know which customers need extra attention, which supplier needs to be chased and which apparent exception is actually normal.

If that knowledge leaves before it has been mapped into the new process, the business may discover that the "automated" work still requires constant human intervention.

Quality can decline. Customer frustration can increase. Remaining employees may spend their time correcting errors. In the worst cases, the business ends up rehiring while also paying for new technology.

That does not make automation a poor investment. It makes premature redundancy a poor implementation strategy.

A better approach: automate before you restructure

Before making a workforce decision based on expected AI savings, run the new process alongside the existing one.

Start with a narrow workflow that has a visible cost or lost-revenue problem. Measure how it performs for several weeks. Track the time saved, exceptions created, customer response and level of human intervention required.

Then ask:

1. Did the automation remove work or simply move it somewhere else?
2. Does it perform consistently during busy periods?
3. Can staff identify and correct errors quickly?
4. Are customers receiving a better experience?
5. Is the saving large enough to justify the cost and risk?
6. Could the recovered capacity be redirected into sales, service or billable work?

For many small businesses, the best outcome is not reducing headcount. It is avoiding the next hire, absorbing more customers with the existing team or giving the owner time back without allowing service standards to fall.

That can be more valuable — and considerably less risky — than treating automation as a redundancy exercise.

What this means for Australian SMEs

Large organisations may describe AI in terms of workforce transformation. For a trades business, allied health provider or professional services firm, the opportunity is usually much more practical.

It is the enquiry that receives a response before the prospect rings a competitor.

It is the appointment that is confirmed without a receptionist making another call.

It is the quote that gets followed up consistently.

It is the client information already prepared before the first conversation.

AI does not need to replace an employee to produce a return. It needs to remove enough repetitive work for people to concentrate on the work that requires experience, relationships and judgement.

The businesses getting this right are not asking, "How many people can we remove?"

They are asking, "What work should our people no longer have to do?"

That is the better starting point for automation in 2026. For a plain-English breakdown of where the automation ends and the AI actually begins, see What Is AI Automation? A Plain-English Guide.