Giving staff an AI tool without teaching them how to use it produces uneven results and avoidable risk. A short, practical foundation lifts the whole team. Here is what we recommend every team learns first, and how to build from there.
1. How AI actually works, in plain terms
People use AI better when they understand its nature. Language models predict plausible text from patterns. They do not look facts up unless connected to a source, and they can be wrong while sounding certain. Staff should be able to tell the difference between a model, an assistant built on it and a search over approved documents.
2. Good prompting
Clear prompts state the task, audience, context, constraints and desired format, and provide examples. Iterating and asking the tool to show its reasoning or sources improves results. Teams benefit from a shared library of tested prompts for common tasks.
3. Checking outputs
Treat output as a draft from a capable but fallible assistant. Verify facts, figures, quotes and citations against primary sources. Read for tone and for anything that could mislead. The person who sends or publishes the content remains accountable.
4. Protecting information
Know which tools are approved, what data classifications may be entered and what must never be pasted into a tool, such as personal information, credentials and confidential client material. Understand what an assistant such as Microsoft 365 Copilot can see on your behalf, since it respects the permissions you already have. See our Copilot readiness guide.
5. Knowing when not to use AI
Decisions that significantly affect individuals, legal and safety-critical advice, and tasks where an error is costly need a person in control and often a different tool. Staff should know the escalation path and the organisation’s AI policy.
6. Spotting problems and reporting them
Make it easy to report a wrong, biased or leaking output. Early reports are the cheapest monitoring you have.
Role-specific next steps
- Managers and executives: accountability, risk tiers and the questions to ask of AI proposals. See the ten guardrails explained.
- Analysts and knowledge workers: advanced prompting, working with documents and data, and verification habits.
- Developers: building LLM applications, retrieval, evaluation and guardrails.
- Security and IT: access design, monitoring and incident response for AI.
Make the learning stick
Use the team’s real tasks and tools, schedule short follow-up clinics and measure adoption and quality, not attendance. A written acceptable-use policy keeps everyone aligned.
Book a session
EITSAA runs AI fundamentals, safe Copilot use and role-based workshops on-site in Adelaide or online, tailored to your tools and policies. Book a session or see Training & Team Consultations. For organisation-wide policy and controls, see AI Governance & Assurance.
This is general guidance, not legal advice.