AI training for staff in New Zealand: what to cover and how to make it stick

Your team can probably use AI at work already. Microsoft includes Copilot Chat with eligible Microsoft 365 licences, Google has made its Gemini app a core service in most Workspace editions, and ChatGPT is a browser tab away.

Access alone doesn’t change how people work, though. Without guidance, a few enthusiasts use AI for everything, others try it once on a vague request and go back to old habits, and someone may paste customer details into a personal account. The Government’s AI strategy lists “workers who know how to use AI productively and responsibly” among New Zealand’s skills gaps.

This guide covers what good AI training for staff includes, the formats and time involved, the privacy rules that apply, how to measure results and what funding exists.

What good AI training covers

Good training answers the questions that stop people using AI well: what can I put into it, how do I ask, how do I know the answer is right, and what is it good for in my job?

Safe use and what not to share

Start with plain rules: which tools are approved, which account to use, and what never goes in. The business.govt.nz guidance suggests a short “do-not-paste” list (personal information, customer details, financial records, anything confidential) and points out that typing, pasting or uploading information into an AI tool all count as disclosures.

Show people why the account matters. Google asks people using its Gemini app with a personal account not to enter confidential information they wouldn’t want a human reviewer to see. For qualifying Google Workspace accounts, it says content isn’t human reviewed or used for AI model training outside your domain without permission.

Prompting with real work

Ask each person to bring two or three tasks from their own week and practise on those. Teach a simple habit: give the tool the task, the context, the audience, the format you want and an example of good output, then refine the answer back and forth.

Checking outputs

The Privacy Commissioner warns that generative AI tools “often produce very confident errors of fact or logic”. Teach a routine: confirm facts, numbers, names and dates, open any source the tool cites, and never send AI output to a customer unread. MBIE’s guidance calls the habit of accepting outputs as correct without critical evaluation “automation bias”, so build error-spotting into the exercises.

Role-specific use cases

Group people by role and agree a few uses for each: quotes and customer replies for sales and service staff, meeting summaries for managers, job reports from site notes for field teams. Save the prompts that work in a shared library so good habits spread.

Pair training with a short policy and the privacy basics

Training without rules leaves people guessing, and rules without training get ignored. MBIE’s Responsible AI Guidance for Businesses suggests foundational responsible AI training for all staff, covering allowed business uses and AI’s limitations, followed by regular training. It adds that usage policies can help tell staff which tools are endorsed. Our AI policy template gives you a starting point to adapt; teach it in the same session.

Cover these privacy basics:

• The Privacy Act 2020 applies. The Privacy Commissioner’s starting point is that it “applies to everyone using AI tools in New Zealand”, and you need to understand enough about how a tool works to be confident you are upholding the information privacy principles.
• Personal and confidential information stays out unless it has been explicitly confirmed that the provider doesn’t retain or disclose it. That is one of the Commissioner’s expectations for generative AI, along with having a person review outputs before acting on them.
• Staff know who to tell. Every organisation must have a privacy officer, and slip-ups should reach them quickly: if a privacy breach has caused or is likely to cause serious harm, you must notify the Commissioner and affected people as soon as you are practically able.

Customers care about this. In the Commissioner’s 2026 privacy survey, 66% of New Zealanders said they would consider changing service providers over poor privacy and security practices. The Commissioner’s free 30-minute Privacy ABC e-learning module makes useful pre-work.

Formats that work, and the time they take

Allow roughly a day per person over the first two months, spread out. The follow-up matters most: by then people have tried AI on real work and arrive with specific questions.

Choose champions for enthusiasm, not technical skill, and give them time to help colleagues and collect good prompts.

Free self-paced options can cover the basics: business.govt.nz points to the AI Forum’s e-learning modules and the Ministry of Social Development’s Digital Passport courses, and Microsoft publishes Copilot training modules on Microsoft Learn. MBIE notes that courses like these can be particularly effective when paired with guidance on your own business context.

Bring people with you

People will wonder what AI means for their jobs, so raise it first. If the aim is to take tedious work off people’s plates rather than cut roles, say so plainly and stick to it. If roles may change, don’t promise otherwise: Employment New Zealand’s good faith guidance says that where business decisions might result in job losses, employers should give employees relevant information about the proposals and a chance to comment before any final decision.

Then make adoption easy:

• Ask the team which tasks they would most like to hand over, and start there.
• Have managers go first and share their prompts, including ones that didn’t work.
• Pair confident users with cautious ones.
• Treat mistakes reported early as lessons, not misconduct, so problems surface while they are small.
• Agree what happens to the time saved: faster service, clearing backlogs or work that keeps getting put off.

Measure adoption and time saved

Decide what success looks like before the first session, or you will end up counting logins. Track three things:

1. Adoption: are people using the approved tools each week? Admin reports can show this. Microsoft’s Copilot usage report, for example, shows enabled users, active users and the active users rate over 7, 28, 90 or 180 days.
2. Time saved on real tasks: before training, time three or four common tasks per role, then repeat four to six weeks later, noting quality and any problems with privacy, tone or accuracy, as business.govt.nz suggests.
3. Confidence: a short survey before and after (using the tools, knowing what not to share, checking outputs) shows where to focus follow-ups.

Spot-check self-reported savings, and count the near misses people report: they show the policy is being used.

Is there funding for AI training?

Funding for staff AI training is limited. Two Regional Business Partner Network schemes come close, with conditions:

• AI Advisory Pilot. Co-funds up to 50% of an AI roadmap (up to $2,500 excl. GST) and of putting it in place (up to $12,500 excl. GST), which can include training staff and supporting change. It doesn’t fund stand-alone AI training. Businesses need 50 or fewer full-time equivalent employees, GST registration and an NZBN, must be actively trading and registered with the network, and work with an approved provider. Places aren’t guaranteed. It runs until 30 June 2027; our AI Advisory Pilot guide has the steps.
• Management Capability Development Fund. Co-funds approved management training from registered providers: up to 50% of one-to-one or 75% of group training, up to $5,000 a year (excl. GST). You need fewer than 50 full-time equivalent employees, an NZBN and a discovery session with a growth advisor, and must be a trading, privately owned business or a Māori trust or incorporation (GST registration is needed for one-to-one training). AI is a listed subject for group training, but the fund is for owners and managers: team and individual employee training, and product training such as Microsoft Office, aren’t eligible.

Talk to a growth advisor before committing to a provider.

Common mistakes

• Buying licences and hoping. Access isn’t adoption. Plan training before the rollout.
• Generic examples. Use your team’s own tasks, not someone else’s demo.
• One session and done. Without follow-ups, new habits fade.
• Banning AI with no alternative. It risks pushing people onto unapproved tools and personal accounts you can’t see, which business.govt.nz calls shadow AI.
• Skipping permissions before Copilot training. Microsoft Copilot (formerly Microsoft 365 Copilot) only surfaces organisational data each user already has permission to view, Microsoft says, so anything overshared in SharePoint, Teams or elsewhere in Microsoft 365 can turn up in answers. Tidy sharing first.

Where to start

Start with one team, a handful of real tasks and a short policy, then build on what works. Our AI training for teams runs hands-on AI workshops on your team’s own work, with the policy and follow-ups that make new habits stick. To see where you stand first, try the free AI readiness check or book a free consultation.