AI Wrote the Email. The Fair Work Commission Read It. Here’s Why That Should Worry Your Business

Your manager is fed up with a difficult employee. It’s 9pm. They open ChatGPT, type “write a firm email about their ongoing behaviour issues”, and hit send on whatever comes out.

Three months later, that email is sitting in front of a Fair Work Commissioner.

This isn’t a hypothetical any more. In the last few months, the Commission has banned two parties from using AI to write to each other, a court has ordered a former HR professional to pay $45,000 in costs after filings full of AI errors, and the Commission has brought in formal rules for anyone using AI in its cases.

AI isn’t the problem. Lazy AI is. Here’s what’s happened and what to do about it.

The short version What’s changed: The Fair Work Commission and the courts are now calling out AI-written workplace communication and AI errors in legal documents, and they’re attaching consequences. The risk: Long, accusatory, inaccurate AI emails make disputes worse and can become evidence. AI errors in documents can cost you credibility and money. The fix: A simple AI rule for your managers: AI can tidy, humans decide, and nothing goes out unchecked.

Has the Fair Work Commission Really Banned AI Emails?

Between two specific parties, yes.

In late July 2026, Commissioner Sarah McKinnon dealt with an anti-bullying application from a caretaker working for a Queensland body corporate, as reported by the ABC. The caretaker said the body corporate repeatedly refused to clarify their duties and gave confusing, shifting directions.

The Commissioner found the caretaker had been bullied.

But she also took aim at the emails the caretaker had been sending, which appeared to have been prepared with an AI tool. She described them as long, sweeping, full of generalisations, repetitive and often accusatory. She said it wasn’t surprising the body corporate eventually stopped engaging.

Her orders required all future communication between the parties to be accurate and respectful, kept brief and to the point, with “no use of artificial intelligence tools in the preparation of correspondence between the parties”. The order took effect immediately.

Why Does This Matter If the Worker Still Won?

Because this is where it gets interesting for employers.

The AI emails didn’t wipe out the bullying complaint. The substance of the caretaker’s concerns still stood. But the Commissioner also found that the body corporate’s response, essentially disengaging and digging in, meant it never properly dealt with many of the issues being raised. That hurt everyone involved.

So there are two lessons in one case:

• If your staff are writing with AI, they can bury a real issue under so much noise that nobody wants to read it, and the dispute drags on.
• If you’re receiving AI-written complaints, you still have to deal with what’s underneath them. Ignoring a complaint because it’s badly written won’t protect you.

Every business owner has had one of those emails. Three pages, twelve bullet points, a lot of “furthermore”. The answer isn’t to stop replying. It’s to pick up the phone, work out what the person actually wants, and respond to that.

Can AI Mistakes Really Cost You Money in a Workplace Dispute?

Yes, and a former HR professional just found out how much.

In [2026] FCA 1415, the Federal Court ordered a former Stantec Australia employee who had worked in a professional HR role to pay the company $45,000 in costs. He had asked the court to review two Fair Work Commission decisions. The court dismissed his case, describing it as always doomed to fail.

Along the way, the court noted “apparent hallucinations” in his documents, including references to written submissions that didn’t exist. He said he had only used AI to check grammar and spelling. The court thought it likely that some of the problems came from indiscriminate use of AI.

Costs orders like this are rare in Fair Work matters. Each side usually pays its own way. But a court can order costs where a case is brought without reasonable cause, or where unreasonable conduct runs up the other side’s bill. The court found both applied here.

Two details are worth noting for employers:

• Stantec’s early offer helped. In June 2025, its lawyers invited him to withdraw his Commission claim with no costs order, because it had no merit. He refused. That kind of early, written offer can matter a lot later.
• Stantec’s own numbers had errors. The court found mistakes in the company’s costs evidence and cut its claim from about $56,000 to $45,000. Accuracy cuts both ways.

It’s not the first time

Earlier this year, in Hoverd v M & J D Pty Ltd [2026] FWC 1013, an applicant told the Commission he had used AI to organise and draft his submissions. According to the FWC bulletin, he relied on contract terms and award provisions that didn’t exist, and the Commission found he had deliberately misrepresented facts.

Made-up award clauses. Made-up contract terms. Submitted to the tribunal that writes the awards. You can see the problem.

What Are the Fair Work Commission’s New AI Rules?

From 20 October 2026, the Commission’s GenAI requirements apply to its matters. If you use generative AI to help prepare an application, submission, witness statement or other document, you must:

• Disclose that you used it.
• Explain how you used it.
• Check the content is accurate and relevant.
• State that those checks have been done.

The Commission also warns against putting personal or confidential case information into public AI tools.

These rules apply to employers as much as employees. If your HR person, manager or adviser drafts your response to an unfair dismissal claim using AI, the disclosure obligation is yours.

Should Your Business Ban AI for Workplace Emails?

No. A blanket ban is unrealistic, and AI is genuinely useful for tidying up writing, checking tone and saving time.

But you do need rules, especially for anything sensitive. Here’s where we’d draw the line:

Fine to use AI for	Don’t use AI for
Checking spelling, grammar and clarity	Writing warnings, show cause letters or dismissal letters from scratch
Making a long email shorter	Responding to complaints, grievances or bullying allegations
Drafting routine updates and roster notices	Anything involving personal, health or confidential information in a public tool
Summarising your own notes	Stating facts, dates, award terms or legal positions you haven’t checked
Brainstorming how to approach a conversation	Documents for the Fair Work Commission without disclosure and checking

 

Fun fact from the ABC’s coverage: one of the giveaways experts mention for AI-written text is the humble long dash. We’ve banned it from our writing for years. Turns out we were ahead of our time.

How Do You Write an AI Policy That Actually Works?

Keep it short enough that people will read it. One page is plenty for most SMEs. Cover:

1. What AI tools are approved. Name them. If you have a business version with better privacy settings, say staff must use that one.
2. What can never go into AI. Personal information, health details, performance and disciplinary matters, client data and anything confidential.
3. Who is responsible for the output. The person who sends it. Every time. AI is a tool, not an excuse.
4. What needs human writing. Disciplinary, grievance, complaint and termination communication should be written or substantially rewritten by a person, then checked.
5. The check before sending. Is every fact correct? Is it shorter than it needs to be? Would you be comfortable reading it out to a Commissioner?
6. Disclosure in legal matters. Anything going to the Fair Work Commission follows its GenAI requirements.

What to do now Urgent: If you have a live dispute or Fair Work matter, check whether anyone has used AI to draft documents for it. From 20 October, that use must be disclosed and checked. Urgent: Tell managers today: no AI-written warnings, complaint responses or dismissal letters without a proper human review. Lower priority: Put a one-page AI use policy in place and add it to your induction and manager training.

Frequently Asked Questions

Can the Fair Work Commission ban AI in workplace emails?

In an anti-bullying case it can make orders to stop bullying, and in July 2026 it ordered two parties not to use AI tools to prepare correspondence to each other. That order applied to those parties, not to every workplace.

Is it illegal to use AI to write work emails?

No. But you’re responsible for what you send, and AI-written emails can become evidence in a dispute. Accuracy, tone and privacy all still apply.

Do I have to tell the Fair Work Commission if I used AI?

From 20 October 2026, yes. If you use generative AI to prepare an application, submission, witness statement or other document, you must disclose it, explain how it was used and confirm you’ve checked it.

Can I be ordered to pay costs because of AI errors?

Costs are usually not awarded in Fair Work matters, but they can be where a case is brought without reasonable cause or a party acts unreasonably. In the Stantec case, AI errors were part of the picture when the court ordered $45,000 in costs.

How should I respond to an employee complaint that looks AI-written?

Deal with the substance. Meet with the employee, ask what the key concerns are, and respond to those clearly. Ignoring or disengaging from a complaint can count against you.

Should my business have an AI policy?

Yes. A short, practical policy that covers approved tools, what can’t go into them and who is responsible for the output will prevent most of the problems in these cases.

Need an AI Policy That Doesn’t Read Like a Robot Wrote It?

HR Gurus will cut through the BS and give you a practical, one-page AI use policy for your team, plus clear rules for managers handling the sensitive stuff.

Book a call with HR Gurus today.

 

This article is general information only and is not legal advice. It is based on published decisions and reporting as at October 2026.