Should your next hire be a person or AI? What Australian law asks

Summary

This article helps you decide whether your next hire should be a person or AI, and sets out what Australian law asks of you when AI changes a role. It is one of the Putting AI to Work series for Australian businesses and organisations, and the other articles in the series are mapped below. It is written for small, medium and large organisations alike.

Skip to the New Role Framework ↓

Background

Hire a person for the work that everyone else waits on and for the work that rests on judgement, and let AI take the rules-based admin around them. If you cannot tell which is which within a week, hire. An AI experiment is not a reason to leave a team short. When AI does change a role, the duty is the same at every size. Consult the people affected before the change takes effect. The Fair Work Act treats an employer with fewer than 15 employees as a “small business employer” and asks less of it when a job ends. There is no redundancy pay, and unfair dismissal protection starts later. So it is easy to assume the duty to consult is lighter too. It is not. Being under 15 changes what a redundancy costs, not whether you must consult first.

We read all 121 modern industry and occupation awards (the Fair Work Commission’s instruments that set minimum pay and conditions for an industry or an occupation). Every one carries the same consultation clause, word for word. No consultation clause mentions artificial intelligence, automation or software, no award uses the words “artificial intelligence” anywhere, and every clause starts only once the employer has made a “definite decision”.¹ The national safety code asks for consultation on new technology “as early as possible”, and a House committee asked in February 2025 for consultation “before, during, and after” new technology arrives. As at 18 October 2026 no award, Act or Bill has made that change.

Checked against the Fair Work Act 2009, the Fair Work Commission’s award texts, the Fair Work Ombudsman, a Full Bench of the Commission and the High Court, the four federal discrimination Acts, the Privacy Act and the OAIC, Safe Work Australia, SafeWork NSW, Jobs and Skills Australia, the ABS, the government’s response to the House committee and the Parliament’s bills lists, on 18 October 2026. This covers national-system employers of any size. It is not legal advice; it names the law and the section so you can check it.

Methodology

A long article made quickly with AI can be slop, and the risk is real. We would rather say how this one was made than have you guess.

1. The questions came first. The issues this article addresses are ones the author has worked through with clients facing them, in his consulting practice.
2. AI did the research and the first draft. Claude Code, running Claude Fable 5.1, read the primary sources and drafted the article to the house standard: every claim tied to a source, and every source opened by a person before the article leaves draft. Where an article says "we read" a document, or all 121 of something, the model read every one of them in full and did the counting; the author checks the method and a sample, not every page.
3. The author did the first review. The author read the draft end to end, cut what did not hold, and edited the body.
4. Other AI tools checked it. A second and third model read each article for errors and for one model's blind spots. This article has not had that check yet.
5. Specialists review it last. Reviewers across technology, law and human resources read the final draft. This article has not had its specialist review yet, and no law firm has reviewed it.

Why use AI at all, given the risk? Because at the rate AI is moving it would be remiss not to use the top frontier models to synthesise the current state of play, so that a business decides from the most up-to-date information there is. Feedback and corrections are welcome, and so are volunteers, law firms among them, who want to make this the best resource available to Australian businesses: write to the editor.

Sort the role’s work before you write the ad

Most hiring requests are written in the shape of the last person who did the job. When AI is in the business, that shape is the wrong starting point, because the parts of the role that are the same every week are the candidate tasks for AI to take over.

The Australian evidence is that roles change more than they vanish. Jobs and Skills Australia’s 2025 study found “Gen AI is more likely to augment jobs than replace them”, and that entry-level roles “may be more likely to transform than diminish”, with “no current evidence of widespread displacement in Australia”.² The Australian Bureau of Statistics (ABS) counted 12 per cent of Australian businesses using AI in 2024-25, and A six-to-twelve-month AI roadmap has the breakdown by size.³

Our reading of that evidence is this. Roles change rather than disappear because the bottleneck moves from processing to judgement. Today the slow step in most roles is the processing: reading, sorting, drafting, reconciling. Once AI does that, the slow step becomes the approval, the judgement and the guidance a person gives, and that is where the role’s value ends up. AI becomes a standard office technology, as the spreadsheet and email did. What rises is the expectation of the quality and volume of work a person does with it, rather than the risk of the job going. There is dislocation in the short term while an organisation works out what it has. What follows is redeployment into roles that make sense under the new structure. That is the organisation’s decision to make and to say, as How to build an AI roadmap sets out.

So before a role is advertised or backfilled, list its tasks. The position description is a start. The person doing the job and the last three months of their sent email are better. Put each task in one of three columns.

Type of work	What it looks like	Impact
Rules-based admin	The same steps each time: entering and coding data, copying between systems, compiling a report, chasing by email, matching one list against another	Automate or assist first, with a person checking
The step others wait for	The bottleneck: the approval that holds up payment, the specialist step that queues when the work before it speeds up	Hire. A person adds the most here, and AI makes them faster
Judgement, taste or relationships	Customers, suppliers, pricing, people decisions, anything where being wrong costs trust	Hire. AI prepares; a person decides

Then ask four questions of the role.

1. If AI took the rules-based admin, how much of the role is left? Most of it: hire as planned. About half: could the role be broader, more senior or part-time? Little: move the work into an existing role with AI help, and make the hire where the bottleneck is.
2. Where is the bottleneck now, and where will it be in a year? Speeding up one step moves the queue to the next. Hire ahead of the moving gap.
3. Who will supervise the automated part? Automated work needs a named person who owns it, checks it and knows when it is wrong. Write that into a role. Who signs off when work moves into a system is in Choosing AI software and connecting it to your systems.
4. What will the person do with the time freed up? Name it. The section on sharing the benefit below says why.

Automate first only when all four of these hold, and hire when any one fails:

Automate first when	Hire when
The work is mostly the same steps, in volume	Every case is different, or the volume is small
The information is in a system the tool can reach today, or could with one small project	The work sits inside a system nobody has yet shown the tool can reach
A mistake is internal and reversible, or is caught before anything goes out	A mistake reaches a customer, a supplier or money before anyone sees it
A named person has the time to set it up and check it	Nobody has time to supervise it, so it would run unwatched

Where hires and automations sit over a year is in A six-to-twelve-month AI roadmap. Record each decision in one line: the role, hire or reshape or automate first, who supervises, and when you will look again. A decision to hire as planned is worth recording too.

If the answer is not clear in a week, hire

The sort above should take one meeting. If after a week nobody can say with confidence that the automation will carry the work, hire. This is our recommendation, not a legal rule, and it rests on three things the sources do say.

The gains are not yet measured in a business’s own back office. AI process automation: where to start read 14 government, academic and survey sources and found none that measures what AI did to an Australian business’s invoices or inbox. A team short of a person while an experiment runs carries the cost now for a gain that may come later.

A short team is a safety risk you already own. Safe Work Australia’s model code lists “Unreasonable or excessive time pressures or role overload” among job demands that are psychosocial hazards (risks to mental health from the way work is designed and managed), and lists “not allowing for drops in productivity while workers learn new processes” as poor change management.⁴ Every state and territory except Victoria has adopted the model work health and safety laws, according to the government’s own response to the House committee, and Victoria has similar duties.⁵ Leaving a vacancy open while the remaining staff absorb the work and learn a new tool is the situation the code describes.

Not replacing people can itself be the change you must consult on. The award clause counts “major changes in the composition, operation or size of the employer’s workforce” as a significant effect.⁶ A Full Bench of the Fair Work Commission (three or more members sitting together on an appeal, whose decision binds the Commission’s single members) said in July 2026 that a cut from three salespeople to two engaged the consultation clause for those who stayed as well as the one who left, because the work was “to be absorbed by those who remain”.⁷ An AI pilot that becomes a reason not to backfill is the same change by a slower route, the same work shared among fewer people.

All 121 awards carry the same consultation clause, and none mentions AI

We downloaded every award numbered MA000001 to MA000122 from the Commission’s award site on 18 October 2026, each consolidated (the text with every variation to that date written in) to a date between 29 June and 1 October 2026. One number, MA000037, returns no award text, which leaves 121.

In all 121 the consultation clause opens with the same sentence, and once clause numbers are set aside, 108 are identical to the letter, and the other 13 differ by a comma, a colon, the word “clause”, “at” for “in”, or a typing slip.¹ The Fair Work Ombudsman says the same thing in one line: “All awards and agreements have a consultation process for when there are major changes to the workplace that are likely to have significant effects on employees.”⁸

We then searched every award text on the site, 155 including the enterprise and state public sector awards, for “artificial intelligence”, “AI”, “algorithm”, “generative”, “machine learning” and “chatbot”. None appears in any award. In the 121, “automated” appears in two, both in classification descriptions of what a worker operates, and no consultation clause uses “automat”, “computer” or “software”. The one award that names automation as a cause of job loss is an enterprise award for a single grain handler, which requires “not less than three months’ notice of termination” for redundancy “on account of the introduction or proposed introduction by the employer of automation or other like technological changes”.¹

That does not leave AI outside the clause. The trigger is technology in general. In the Clerks award, which covers clerical work defined to include “invoicing, billing, charging, checking”, clause 38 reads:⁶

If an employer makes a definite decision to make major changes in production, program, organisation, structure or technology that are likely to have significant effects on employees, the employer must: (a) give notice of the changes to all employees who may be affected by them and their representatives (if any); and (b) discuss with affected employees and their representatives (if any): (i) the introduction of the changes; and (ii) their likely effect on employees; and (iii) measures to avoid or reduce the adverse effects of the changes on employees; and (c) commence discussions as soon as practicable after a definite decision has been made.

The employer must also give “in writing … all relevant information about the changes”, need not disclose confidential information “contrary to the employer’s interests”, and “must promptly consider any matters raised”. “Significant effects” include termination, “major changes in the composition, operation or size of the employer’s workforce or in the skills required”, loss of job or promotion opportunities, loss of job tenure, altered hours, “the need for employees to be retrained or transferred to other work or locations”, and “job restructuring”.⁶

A tool that takes a day a week of data entry out of a role, and moves that person to checking and exceptions, is a change in technology with a need to retrain. The government’s March 2026 response to the House committee says significant workplace changes “could include AI and ADM technologies introduced in the workplace” (ADM is automated decision-making).⁵

What the award clause does and what it does not.

The award clause	What it means for a business adding AI
Starts after a “definite decision”	You may decide first and consult second. The Fair Work Ombudsman notes that employers who consult “still have the right to make the final decision”9
Notice, written information, discussion of effects and measures to reduce them	A conversation is not enough; the information goes in writing
“Significant effects” include retraining, restructuring and workforce size	Moving work from a person to a tool usually touches the skills required, job restructuring or retraining, three of the seven
Applies whatever the employer’s size	A Full Bench confirmed this in July 2026 (next section)7
Mentions no technology by name	Nothing in it is specific to AI; it covers technology in general

The enterprise agreement model term asks a little more. Section 205 requires every enterprise agreement (a pay and conditions agreement between one employer and its employees, approved by the Commission) to carry a consultation term about “a major workplace change that is likely to have a significant effect on the employees”. Where an agreement has none, the Commission’s model term (a standard consultation clause that is read into the agreement) applies.¹⁰

The model term, made in February 2025, adds three things the award clause lacks: the employer must give “the reasons or justification for the change”, must give “prompt and genuine consideration” to what staff raise, and must “take reasonable steps to communicate the outcome”, and its list of effects adds “job security”.¹⁰

The two instruments do not agree on which model term applies. Regulation 2.09 of the Fair Work Regulations, in its compilation of 1 October 2026 (a compilation is the official text with every amendment to that date written in), still says “For subsection 205(3) of the Act, the model consultation term is set out in Schedule 2.3”, which is the older text.¹¹ The Act keeps the old provisions for agreements voted on before the amendments to s 205 known as Closing Loopholes No. 2 commenced. If your agreement has no consultation term, ask the Commission which model term it carries.

The safety code asks you to start earlier. Separately from the award, work health and safety law requires consultation with workers on health and safety matters. Safe Work Australia’s code lists “changes to tasks, workload, duties and working arrangements” and “new technology” among the changes to consult on, and says “Consultation on changes that may affect work health and safety should occur as early as possible.”⁴ The award clause runs from the decision. The code speaks of proposed changes and of consulting as early as possible. A business that waits for its decision has met the award and may not have met the code.

The committee asked for “before”. Nothing has changed. The House Standing Committee on Employment, Education and Training tabled The Future of Work, dated January 2025, on 11 February 2025.¹² Its recommendation 5.102 asks the government to “strengthen obligations on employers to consult workers on major workplace changes before, during, and after the introduction of new technology”, and recommendation 2.92 asks it to classify AI used for “recruitment, referral, hiring, remuneration, promotion, training, apprenticeship, transfer or termination” as high-risk.¹²

The government’s response of March 2026 says the Fair Work Act and safety laws “are principle-based and designed to be technologically neutral”, that “Additional consultation requirements may be agreed through enterprise bargaining or Fair Work Commission (FWC) processes”, and that the Department of Employment and Workplace Relations “is progressing a regulatory analysis of workplace laws”.⁵

On the Senate bills list of 16 October 2026 and the House list of 15 October 2026, no Bill would require an employer to consult about AI or regulate AI in employment decisions, and the one automated decision-making Bill covers Commonwealth administrative action only.¹³

NSW has passed a work health and safety duty over “digital work systems”, defined to include artificial intelligence, covering the allocation of work by such a system, and it awaits proclamation (the government’s notice fixing the day it starts).¹⁴

Under 15 employees or over: the money changes, the duty does not

The Fair Work Act’s “small business employer” is one that “employs fewer than 15 employees”, counting every employee, regular casuals but not irregular ones, the employees of associated entities (related companies, such as a parent or a sister company), and the people being dismissed.¹⁵ It is a different line from the ABS bands this site uses (small is under 20) and from the Privacy Act’s $3 million turnover test.¹⁶ Here is what it changes when AI reshapes a role.

When AI changes or removes a role	Fewer than 15 employees	15 or more employees
Award consultation clause	Applies in full7	Applies in full6
Consultation in an enterprise agreement	The agreement’s term, or the model term10	Same
Safety consultation on new technology and workload	Applies4	Applies
National Employment Standards (NES) redundancy pay	None, unless the award has an industry scheme that covers small business or the employer is insolvent after downsizing158	4 to 16 weeks’ pay by length of service, for employees with at least a year’s service15
Unfair dismissal protection starts	After 1 year17	After 6 months17
Genuine redundancy defence	Same test: job not needed by anyone, consultation done, redeployment not reasonable17	Same
Small Business Fair Dismissal Code	Covers conduct and capacity dismissals, not redundancy718	Does not apply
Notice to Centrelink and unions	Cannot arise	When 15 or more are to be dismissed for “economic, technological, structural or similar” reasons19

In Orientile Pty Ltd v Zhang, the Full Bench decided the small business question. Decided on 7 July 2026, the case concerned a small business covered by the Storage Services and Wholesale Award that cut its sales team from three to two, and told the employee chosen, at a single meeting, that the position was redundant.⁷ The Commission found the dismissal unfair and ordered compensation, and the Full Bench dismissed the appeal.

Its catchwords (the one-line summary of what the decision holds, printed at its head) state the rule: “Modern award consultation obligations apply in the same way to all employers including small business employers”, and “Partial or substantive compliance is not sufficient”.⁷

The Bench said “The fact that the Appellant is a small business, or that one employee was to be dismissed, does not diminish the significance of the change”, that it is “no answer” to say the same person would have been dismissed anyway, and that “consultation is of particular importance to employees of small businesses who are not entitled to the redundancy payments set out in the scale in s. 119”.⁷ The first-instance decision (the single member’s decision that was appealed) also found the Small Business Fair Dismissal Code had no application because the dismissal was for redundancy, not conduct.⁷

The case was not about AI. Its facts are the facts of an AI change in a small team. Fewer people, the same work, redistributed among those who stay.

AI can be the reason for a redundancy, and the role you are about to hire can undo it

The government’s response says it plainly: “Redundancy provisions under the Fair Work Act apply to AI-related redundancies.”⁵ The Fair Work Ombudsman lists among the causes of redundancy a business that “introduces new technology, for example, the job can be done by a machine”.⁸ Section 389 sets the test.

A dismissal is a genuine redundancy if the employer “no longer required the person’s job to be performed by anyone because of changes in the operational requirements of the employer’s enterprise” and “has complied with any obligation in a modern award or enterprise agreement … to consult about the redundancy”. It is not genuine “if it would have been reasonable in all the circumstances for the person to be redeployed within” the business or an associated entity.¹⁷ If it is genuine, there is no unfair dismissal remedy. If it is not, there may be.⁸

Three parts of that test meet the hiring decision.

“The job” means the work, not the title. The High Court said in August 2025 that s 389(1) “does not look to whether the employee’s position, in terms of job title, was no longer required, but whether their ‘job’, in the sense of the nature of the work they performed, was no longer required”.²⁰ If AI takes part of the work and a person is still needed for the rest, the job has not gone.

Redeployment does not need a vacancy, and can include training. The same judgment says “‘Redeployed’ does not require there to be a vacant position”, and that “all the circumstances” can include “whether redeployment would require the employee to undergo further training” and “a position being performed by a contractor while waiting for an employee to be hired”.²⁰ Suppose the sort above points to a new role at the bottleneck, and the person whose admin work is moving could do that role with training. Making that person redundant while you advertise the new role is then open to challenge under s 389(2).

Re-filling the role is not a redundancy. The Small Business Fair Dismissal Code, as the Fair Work Ombudsman publishes it, says: “Re-filling the position with a new employee is not a genuine redundancy.”¹⁸

The consultation step is part of the test. Section 389(1)(b) makes award consultation a condition of a genuine redundancy. That is why the failure to consult in Orientile cost the employer its defence even though the job had gone.⁷¹⁷ Read the consultation clause before the meeting, not after it.

We found no Commission or court decision in which AI or automation was the operational reason for a redundancy. See “What we could not confirm” below.

If AI screens applicants, discrimination and privacy law follow the tool

Discrimination law reaches the screening step. The Sex Discrimination Act makes it unlawful for an employer to discriminate on a protected ground “in the arrangements made for the purpose of determining who should be offered employment”, and the Age and Disability Discrimination Acts use the same words, and the Racial Discrimination Act makes it unlawful “to refuse or fail to employ” a person by reason of race.²¹ A screening tool is an arrangement for deciding who is offered a job.

Under the Disability Discrimination Act, it is also discrimination to impose a requirement a person with a disability could meet only with “reasonable adjustments” that are not made, which can bear on timed online tests and recorded video interviews.²¹ The Fair Work Act separately prohibits adverse action (refusing to employ someone, for example) against a “prospective employee” because of a protected attribute.²² The government’s response adds that this applies “irrespective of the use of AI”, and that in a claim “the employer will bear the onus of proving that adverse action was not taken for the alleged reason”.⁵

The Human Rights Commission has published a checklist. In 2024 the Australian Human Rights Commission published an AI and recruitment compliance checklist, developed by LexisNexis “in collaboration with the Commission”, covering privacy and data protection, the safety and security of the system, “giving candidates the ability to challenge unfair outcomes” and “Keeping humans in control of decisions”.²³

The employee records exemption does not cover applicants you do not hire. If the Privacy Act covers you (the turnover test is in AI regulation in Australia in 2026), its exemption for employee records applies only to acts “directly related to … a current or former employment relationship”.²⁴ The Office of the Australian Information Commissioner (OAIC) says “the exemption will not apply to the collection of personal information about prospective employees who are subsequently not employed by an organisation, such as unsuccessful job applicants”.²⁴

From 10 December 2026, AI screening goes in the privacy policy. From that date a covered business’s privacy policy must describe decisions a computer program makes or substantially shapes that significantly affect a person, and the rule itself is in AI regulation in Australia in 2026.²⁵

The OAIC’s guidelines, updated on 30 September 2026, give as their first example of such an arrangement “an entity procures another entity’s AI system to screen and rank job applications leading to a decision being made on who to employ”, and list “access to employment opportunities” among the interests that can be significantly affected.²⁶ A person in the loop does not take you outside the rule, a point Who is accountable when AI makes the decision takes further.

The practical rule for a business is that a person reads every application the tool rejects before it is rejected, or that the tool only sorts and a person decides. That is not a legal requirement in those words, and it does not remove the privacy-policy duty. It is the cheapest way to show that the hiring decision, and any discrimination claim about it, rests on reasons a person can give. How the staff AI policy and any monitoring of staff are rolled out is in How to write an AI policy.

When roles get lighter, decide who gets the time before you say anything

No Australian law tells an employer how to share the time AI saves. The government’s response puts it as aspiration: “With the right regulatory settings, AI adoption has the potential to improve business productivity and deliver better wages, job satisfaction and stability for workers.”⁵ What the law does set is the edges of the choice.

Time. For part-time employees under the Clerks award, changes to agreed hours “must be agreed in writing between the employer and employee”.⁶ A business cannot turn freed time into fewer paid hours by announcement. Altered hours are also one of the award’s “significant effects”.⁶

Pay. A role that now supervises an automation, handles the exceptions and signs off output carries more judgement than the data entry it replaced. Read the award’s classification definitions against the new role. If it now fits a higher level, the award sets the minimum. Moving someone to lower-paid duties because of redundancy needs notice or pay for the difference, under clause 42.1 of the Clerks award.⁶

Training. Safe Work Australia’s code lists “not allowing time for workers to learn new tasks” and “providing insufficient training to support changes (e.g. how to perform a new role or use a new process)” as poor change management.⁴ The High Court treats retraining as one of the circumstances in which redeployment may be reasonable.²⁰ Training is both the fair answer and the defensible one.

The work that comes next. Safe Work Australia lists “work that is too easy (e.g. significantly below a worker’s skills or abilities)” and “having too little to do” as job demands that can be hazards.⁴ A role stripped of its routine and given nothing in its place is not neutral. Name the next piece of work in the same conversation: the customer follow-up that never got done, the supplier review, a first piece of the judgement work from the sort above.

What not to do. Do not let the freed time become a headcount cut that nobody announced, where the decision not to backfill is never said and the work is spread across those who remain. That is a change in “the composition, operation or size of the employer’s workforce”, and the Full Bench applied the clause to the people who stayed.⁶⁷

The benefit most people feel first is a role that fits in its hours, rather than fewer hours. The core work gets done in less time, the week finishes on time more often, and more of it goes on the work that needs judgement. Picture the person it reaches first. They have routinely done five hours of unpaid overtime a week, and now they close their computer at 5pm. Section 16 of A free AI policy template is where a business writes that aim down for staff, and How to write an AI policy says why the promise about jobs stays spoken.

For a business of a few dozen staff, a workable default is to decide the split before anyone’s role changes, write it into the role’s new description, and review it with the person at three months. One share to the person (shorter weeks for those who want them, by written agreement), one to training, one to named next work, and a pay review where the role has moved up. The proportions are yours to decide. Deciding them first is the point.

When you do hire, interview for how people use AI

Hire people who are curious, who check, and who know where the lines are. They do not need to be AI specialists.

Ask	A good answer	A concern
“Tell me about a task you handed to an AI tool. How did you check the result?”	A real task, a real check, and something it got wrong	“It’s always right”, or no curiosity about it
“What wouldn’t you put into an AI tool at work?”	Customer details, pay, passwords, anything given in confidence; the difference between a company and a personal account	Pastes anything anywhere
“How would you get an AI tool to do [a task from this role] every week?”	Steps: where the information is, what to tell it, what to check before it goes out	Expects one perfect prompt
“What did you last learn to do that you couldn’t six months ago?”	Anything	Nothing comes to mind

For managers, add: “How would you tell someone their role is changing because part of it is now automated?” It is the conversation the consultation clause requires. Say in the ad that the role uses AI tools daily and that your AI policy applies. The template is in A free AI policy template.

The New Role Framework

Use this before a new role is advertised. Copy it into a document and fill it in as the decision is made. It holds the role’s work sorted into the three types, the four questions, the conditions for automating first, the decision and who made it, who supervises any automated part, what the person will do with the time freed, and the questions to ask at interview.

The Role Backfill Framework

Use this when someone leaves and the role is about to be re-filled. Copy it into a document and fill it in as the decision is made. It starts from the leaver, asks how much of the role is left if AI takes the rules-based admin, and records the option chosen, who supervises any automated part, who gets the time freed, and whether any other role changes as a result.

The Redundancy Framework

Use this when a role will be removed or substantially changed because of AI. Copy it into a document before the definite decision is made, and fill it in as each step is done. It holds the genuine redundancy test, the consultation steps the award clause requires, what changes with the size of the business, notice, what is said to the person and by whom, who gets the time freed, and the decision made at the top.

What changed

This is version 1.0, first published 18 October 2026 and last updated 19 October 2026. The version moves at a major change, such as a law or a piece of guidance the guide rests on changing; a smaller fix moves the updated date only. Earlier versions are kept from the first major change on.

No changes since first publication.

What we could not confirm

• The Commission’s own count of current modern awards. Our 121 rests on the award site’s numbering (MA000001 to MA000122, less MA000037, which returns no text). We did not open a Commission page that states the number.
• Whether the Commission has any award review on foot that touches AI or consultation. We did not find one. Our search for it was cut short.
• A decision in which AI or automation was the reason for a redundancy. One web search found none, and our search allowance ran out before we could search the Commission’s decisions database itself. Orientile and Helensburgh Coal are about a sales team and a coal mine.
• Whether “anyone” in s 119 and s 389 reaches work moved to software. The sections speak of the job no longer being done “by anyone”. We read no decision on whether software counts as anyone, and do not infer one.
• That the Small Business Fair Dismissal Code we quote is current. We read the copy the Fair Work Ombudsman hosts, marked “Updated 1 January 2011”.
• Which model consultation term applies to an agreement without one. The Act and the Commission’s 2025 Determination point to the new term. The Regulations’ compilation of 1 October 2026 still points to Schedule 2.3. We report both and do not pick.
• The Human Rights Commission’s checklist itself. We read its landing page, not the checklist document.
• Copies read through a text proxy. The Commission refused direct requests for Orientile, so the decision was read through a text proxy from its official PDF. The Fair Work Ombudsman, OAIC, Human Rights Commission, SafeWork NSW, ABS and Parliament pages were read the same way. The Fair Work Act, the discrimination Acts, the Privacy Act, the Model Terms Determination, the High Court judgment, the government response, the Safe Work Australia code, the Jobs and Skills Australia report and the award texts were read directly.

We will re-check this piece every quarter, and on 10 December 2026, when the automated-decision duty in the Privacy Act starts. If the Commission varies the consultation clause or a Bill on AI and workplace consultation is introduced, the count and the table change, and so does the date at the top of the page.

Notes

1. Our count, 18 October 2026: the consolidated texts of modern awards MA000001 to MA000122 published on the Fair Work Commission’s award site, beginning at MA000001, downloaded directly, each consolidated between 29 June and 1 October 2026. MA000037 returns no award text. MA000123 to MA000157 are enterprise and state reference public sector awards; MA000154 and numbers above MA000157 return no award. Clause text compared after removing clause numbers and variation notes; word searches run over the full text of all 155 awards with content. “Automation” appears only in MA000136 (clause 12.3, redundancy due to automation) and MA000156 (a classification description), both enterprise awards. ↩ ↩ ↩

2. Jobs and Skills Australia, Our Gen AI Transition: Implications for Work and Skills, 2025, key findings and p 8. Read 18 October 2026. ↩

3. Australian Bureau of Statistics, Business adoption of artificial intelligence accelerates in 2024-25, media release, 25 June 2026. Read 18 October 2026 through a text proxy. ↩

4. Safe Work Australia, Model Code of Practice: Managing psychosocial hazards at work, July 2022, p 14 (consulting on changes), p 16 (job demands), p 34 (low job demands) and p 40 (poor organisational change management). Read 18 October 2026. ↩ ↩ ↩ ↩ ↩

5. Department of Employment and Workplace Relations, Australian Government response to the House of Representatives Standing Committee on Employment, Education and Training report: The Future of Work, March 2026 (page last modified 1 April 2026), pp 3, 12 and 14 to 17. Read 18 October 2026. ↩ ↩ ↩ ↩ ↩ ↩

6. Fair Work Commission, Clerks—Private Sector Award 2020, consolidated to 1 July 2026 (PR799280): clause 2 (definition of clerical work), clause 10.3 (part-time hours), clause 38 (consultation about major workplace change, 38.1 to 38.6) and clause 42.1 (transfer to lower paid duties). Read 18 October 2026. ↩ ↩ ↩ ↩ ↩ ↩ ↩ ↩

7. Fair Work Commission, Orientile Pty Ltd v Zhang [2026] FWCFB 162, Full Bench, 7 July 2026, catchwords and paragraphs [2], [45] to [47]. Read 18 October 2026 through a text proxy; the Commission refused direct requests. ↩ ↩ ↩ ↩ ↩ ↩ ↩ ↩ ↩ ↩

8. Fair Work Ombudsman, Redundancy (“Not a genuine redundancy”; “Consulting with employees about major workplace changes”), Redundancy pay (published 16 October 2026; the six awards with industry-specific redundancy entitlements) and Who doesn’t get redundancy pay (“An award or agreement can contain industry specific terms that cover small businesses”). Read 18 October 2026 through a text proxy. ↩ ↩ ↩ ↩

9. Fair Work Ombudsman, Consultation and cooperation in the workplace best practice guide, “Consultation about major workplace change”. The copy read carried no date. Read 18 October 2026 through a text proxy. ↩

10. Fair Work Act 2009 (Cth), s 205, compilation 73; and Fair Work (Model Terms) Determination 2025, registered 25 February 2025, Schedule 2, subclauses (6), (8), (9) and (11). Read 18 October 2026. ↩ ↩ ↩

11. Fair Work Regulations 2009 (Cth), reg 2.09 and Schedule 2.3, compilation 57, compilation date 1 October 2026; the transitional rule is Fair Work Act 2009, Schedule 1, Part 16, clause 107. Read 18 October 2026 in the Register’s PDF. ↩

12. Parliament of Australia, The Future of Work: list of recommendations, recommendations 2.92 and 5.102; tabling date from the House of Representatives media release ‘The Future of Work’ report released, 11 February 2025. Read 18 October 2026 through a text proxy. ↩ ↩

13. Parliament of Australia, Senate Bills List, 16 October 2026, entry for the Automated Decision-Making (Safeguards and Transparency) Bill 2026; and House of Representatives Daily Bills List, as at 15 October 2026. The absence of a Bill is our reading of every Fair Work, workplace, discrimination and AI entry on both lists. Read 18 October 2026 through a text proxy. ↩

14. SafeWork NSW, Development of the Digital Work Systems Guidelines, published 13 August 2026, “What are digital work systems”, “Key changes introduced by the DWS Act” and “When the changes start”. Read 18 October 2026 through a text proxy. The Act itself, Work Health and Safety Amendment (Digital Work Systems) Act 2026 No 5 (NSW), assented to 18 February 2026, read in full on 19 October 2026; s 2 commences the s 21A duty “on a day or days to be appointed by proclamation”. ↩

15. Fair Work Act 2009 (Cth), ss 23, 119 and 121, compilation 73, compilation date 7 July 2026, read 18 October 2026 in the Register’s PDF. ↩ ↩ ↩

16. OAIC, Small business: “A small business is one with an annual turnover of $3 million or less.” Read 18 October 2026 through a text proxy. ↩

17. Fair Work Act 2009 (Cth), ss 383, 385 and 389, compilation 73, read 18 October 2026. ↩ ↩ ↩ ↩ ↩

18. Fair Work Ombudsman, Small Business Fair Dismissal Code, copy marked “Updated 1 January 2011”, “Application”. Read 18 October 2026. ↩ ↩

19. Fair Work Act 2009 (Cth), ss 530 and 531, compilation 73. Read 18 October 2026. ↩

20. High Court of Australia, Helensburgh Coal Pty Ltd v Bartley [2025] HCA 29, 6 August 2025, Gageler CJ, Gordon and Beech-Jones JJ at [30], [36] and [40]. Read 18 October 2026. ↩ ↩ ↩

21. Sex Discrimination Act 1984 (Cth), s 14(1)(a), compilation 47; Age Discrimination Act 2004 (Cth), s 18(1)(a), compilation 57; Disability Discrimination Act 1992 (Cth), ss 6(2) and 15(1)(a), compilation 41 (8 October 2026); Racial Discrimination Act 1975 (Cth), s 15(1)(a), compilation 19. Read 18 October 2026 in the Register’s PDFs. ↩ ↩

22. Fair Work Act 2009 (Cth), s 351(1), compilation 73. Read 18 October 2026. ↩

23. Australian Human Rights Commission, AI and Recruitment: Compliance checklist for employers (2024). Read 18 October 2026 through a text proxy; the checklist document itself was not read. ↩

24. Privacy Act 1988 (Cth), s 7B(3), compilation 104 (4 June 2026); and OAIC, Employee records exemption. Read 18 October 2026; the OAIC page through a text proxy. ↩ ↩

25. Privacy and Other Legislation Amendment Act 2024 (Cth), as made, Schedule 1, Part 15, item 88 (APP 1.7 to 1.9), commencing 10 December 2026. Read 18 October 2026. ↩

26. OAIC, APP Guidelines, Chapter 1: APP 1, updated 30 September 2026, the list of examples of “arranged for” following paragraph 1.45, and the list of significant services and supports. Read 18 October 2026 through a text proxy. ↩