AI Mistakes Australian Managers Should Avoid

AI can make managers more effective — but it can also make them less effective if it is used in ways that undermine the human judgment and authentic relationships that good management requires. The mistakes in this guide are not hypothetical. They are patterns that are showing up among managers who are enthusiastic about AI but have not yet developed the judgment to use it well.

The common thread is using AI to avoid the difficult, human parts of management rather than to support them. Management is fundamentally a human activity. AI can handle the administrative overhead, but it cannot replace the judgment, empathy, and accountability that the role requires.

When AI Makes Management Worse

AI can make managers more effective — but it can also make them less effective if it is used in ways that undermine the human judgment and authentic relationships that good management requires. The mistakes in this guide are not hypothetical. They are patterns that are showing up among managers who are enthusiastic about AI but have not yet developed the judgment to use it well.

The common thread is using AI to avoid the difficult, human parts of management rather than to support them. Management is fundamentally a human activity. AI can handle the administrative overhead, but it cannot replace the judgment, empathy, and accountability that the role requires.

Mistake 1: Using AI to Avoid Difficult Conversations

The most damaging way a manager can misuse AI is by using it to substitute for difficult conversations rather than to prepare for them. This shows up in several forms: sending an AI-drafted message instead of having a face-to-face conversation, using AI-generated feedback as a substitute for a genuine performance discussion, or hiding behind the formality of a written AI document to avoid the discomfort of direct engagement.

Difficult conversations are difficult for a reason — they require the manager to be honest, to be present, and to engage with the other person's response in real time. AI cannot do any of these things. When managers use AI to avoid this work, the result is typically worse outcomes for the team member, damaged trust, and unresolved issues that compound over time.

The right use of AI in this context is to prepare for the conversation — to develop your thinking, anticipate responses, and find the right language — and then to have the conversation yourself.

Mistake 2: Letting AI Write Performance Documentation Without Your Input

Performance documentation — review notes, improvement plans, formal warnings — has legal and organisational significance. It needs to accurately reflect your observations and judgments, be specific about the behaviours and outcomes in question, and be consistent with the conversations you have had with the person.

AI-generated performance documentation that is not grounded in your specific observations is a significant risk. Generic language that could apply to anyone, assessments that are not based on actual incidents, and documentation that does not match the conversations you have had can all create problems — both for the person receiving it and for any formal process it might support.

AI can help you write performance documentation more clearly and consistently, but only if you provide the specific substance — the actual observations, the specific incidents, the concrete examples. The documentation should reflect your judgment, expressed clearly with AI's help — not AI's judgment expressed in your name.

Mistake 3: Sharing Team Members' Personal Information with AI Tools

Managers handle sensitive information about their team members — performance issues, personal circumstances, health matters, salary details, and other information that is both confidential and, in many cases, legally protected. Using AI tools to process this information without understanding how the tool handles data is a significant compliance risk.

Most consumer AI tools retain inputs and may use them for training. Entering a team member's name, performance history, or personal circumstances into a consumer AI tool could constitute a privacy breach under Australian privacy law, and could expose both the manager and the organisation to legal liability.

The rule is straightforward: do not enter identifying information about specific individuals into AI tools unless you are using an enterprise tool with appropriate data protection and your organisation has a clear policy permitting this use. When using AI for people management tasks, describe the situation in general terms rather than identifying the specific person.

Mistake 4: Over-Relying on AI for Hiring Decisions

AI tools that screen resumes, score candidates, or generate interview assessments can introduce bias and create legal risk if they are used without adequate human oversight. In Australia, employment decisions must comply with anti-discrimination law, and using AI tools that produce discriminatory outcomes — even unintentionally — can expose the organisation to legal liability.

AI can be a useful tool in the hiring process for tasks like drafting job descriptions, developing interview questions, and summarising candidate notes. But the assessment of candidates and the hiring decision itself require human judgment and human accountability. Managers who use AI to make or heavily influence hiring decisions without adequate oversight are taking on risk that is difficult to manage after the fact.

Mistake 5: Not Being Transparent with Your Team About AI Use

Many managers are using AI to draft communications, prepare for meetings, and handle administrative work without telling their teams. In most cases, this is not a problem — there is no obligation to disclose that you used AI to help draft a team update. But there are situations where transparency matters.

If you are using AI to generate feedback that you present as your own observations, or using AI to produce assessments that team members believe reflect your personal judgment, the lack of transparency can damage trust when it comes to light. The standard to apply is whether the team member would feel misled if they knew how the communication was produced. If the answer is yes, more transparency is warranted.