AI training for teams: what to teach first, role by role.

Why generic AI training rarely sticks

People keep using AI when it saves them time on a task they do every week, and generic demos rarely show that. A session that writes a poem or plans a holiday proves the tool works; it does not show a sales coordinator how to answer a tricky question about lead times. Train on your team’s own workflows instead: last week’s customer emails, your price list and your standard operating procedures (SOPs), with sensitive details removed.

Australian government guidance for small business makes similar points: involve staff in researching and choosing AI tools, explain that the aim is to make their work easier rather than to replace them, and train people to use new tools effectively, safely and responsibly. Most teams start with generative AI (AI that creates text, images or code from a prompt), so that is where first training should focus.

What each role should learn first

Each role needs different first skills. Pick two or three tasks per role that happen often, take real time and are easy to check, and build the training around them.

Managers should go first. When the people who set priorities have used AI on their own work, their expectations are more realistic and they are better placed to judge what their teams produce.

Safe-use rules for everyone

Before anyone uses AI on company work, agree a short set of rules and repeat them in every session. These five cover the main risks for most small and mid-sized businesses.

• Company dataUse only the AI tools the business has approved for work data, and keep pricing, contracts and supplier terms out of personal accounts.
• Personal informationDo not enter personal information, particularly sensitive information, into publicly available generative AI tools. This is the OAIC’s recommendation, and the Australian Privacy Principles may apply to your business.
• Check every outputAI can give confident but wrong answers. Check facts, figures and names against a trustworthy source before relying on them.
• A person is accountableWhoever sends, signs or files the work owns it, whether or not AI drafted it.
• DisclosureTell customers when they are dealing with AI, such as a chat assistant, and follow any client rules on AI use.

A simple AI use policy

A one-page policy is enough to start, as long as it is clear and someone owns it. The National AI Centre publishes an AI policy guide and template. Its checklist asks whether a policy states what AI may and may not be used for, who approves higher-risk uses, what data staff may enter, when human oversight is required, how issues are reported and when the policy will be reviewed. A short outline:

• Purpose and owner: why the business uses AI and the senior person accountable for it.
• Approved tools: which tools staff may use for work, and how to request a new one.
• Data rules: what may and may not be entered, with examples from your own business.
• Uses: for example, drafting replies is allowed; sending prices without approval is not.
• Review and approval: which outputs need a second person, and who approves higher-risk uses.
• Reporting and training: how to report a problem, and the training each role completes.
• Review date: when the policy will be checked again as tools change.

Run a pilot after training

Training sticks when it leads straight into real use. Soon after the session, start a pilot on one workflow.

1. Choose one workflowPick a frequent task from the training, such as drafting replies to standard enquiries.
2. Record a baselineNote how long the task takes today and how often it needs rework.
3. Name a championOne person in the team collects questions, useful prompts and problems.
4. Run it for an agreed periodUse AI on real work for a few weeks, with the safe-use rules in place.
5. Check in weeklySpend fifteen minutes on what worked, what failed and what to change.
6. DecideCompare with the baseline, then extend to another workflow, adjust or stop.

Before choosing the workflow, run it through the AI readiness checklist. A workflow with no clear owner or messy inputs is a poor first pilot, however promising it looks.

How to measure adoption

Measure whether people use AI on real work and whether the work improves, not how many attended training. Attendance shows interest; it does not show change.

Where to keep learning

Training is not a one-off event, because AI tools and features change often. Some starting points:

• The Optivise AI Academy has learning resources, including courseware, tools and demos. Technical staff can also use the free Claude Certified Architect study guide and its mock exam.
• Eligible small and medium businesses in National Reconstruction Fund priority sectors can use the AI Adopt Centres, which offer free training courses, consultations and AI safety guidance.

Optivise AI, headquartered in Melbourne, includes role-based team training in its Business AI Training and Transformation service (Level 2), alongside workflow automation and integration, built around the team’s real workflows. Scope and fees are agreed in each proposal.