← Insights

How to start using AI in your New Zealand business, safely

Start with small, low-risk jobs, keep a person in charge of the results, and be careful what information goes in.

By Milo Coldren · Published 4 October 2026 · 2 min read

This is general information, not legal advice.

The short answer

Pick one or two repetitive tasks where mistakes are cheap, write down a few simple rules, and have a person check what the AI produces. Then expand slowly. New Zealand has useful guidance to lean on: MBIE publishes voluntary Responsible AI guidance for businesses,1 and the Privacy Commissioner has guidance on AI and the Privacy Act.2

What the guidance says, in short

MBIE's guidance is voluntary and written for businesses of all sizes.13 A law firm's summary of it highlights advice that applies to almost any small business:3

  • Treat AI as a tool, not a decision-maker. Check outputs for errors, made-up facts, copyright problems and bias.
  • Avoid putting sensitive or proprietary information into generative AI tools.
  • Keep records of where and how AI is used in your business.
  • Make sure staff understand your rules for using it.

The Privacy Commissioner is more direct about personal information. It recommends a privacy impact assessment before you start, says to keep personal information secure including the prompts you enter, and says that if in doubt you should not use AI tools to handle personal information.2

Where to start

  1. Choose low-risk tasks. Drafting first versions of emails, summarising your own documents, tidying meeting notes and preparing report drafts are good places to begin, because a person can review the result before it goes anywhere.
  2. Keep personal and confidential information out of public tools. Use dummy details when testing, and check what a tool does with the data you give it before using it with customer information.
  3. Write a one-page rule sheet. What tools are allowed, what information must never go in, who checks the output and where AI use is recorded.
  4. Keep a person in the loop. The AI drafts, a human approves. Do not let it send or decide on its own until you trust it for that specific job.
  5. Train the team. Show people the rules and a few good examples. Most problems come from people guessing.
  6. Measure the result. Track the time saved and the errors caught for a few weeks, then decide what to expand.

Jobs AI does well in a small business

  • Scoring and sorting incoming enquiries, with a person making the final call.
  • Drafting replies to common questions for someone to approve.
  • Turning raw numbers into a plain-English weekly report.
  • First drafts of proposals built from your own past work.

Some of these touch customer information, so they need the care above. This is exactly the kind of setup I help with on the AI side: choosing the first jobs, building the system, writing the rules and training your team.

References

  1. Responsible AI guidance for businesses. Ministry of Business, Innovation and Employment (MBIE). Accessed 5 October 2026.
  2. Artificial Intelligence and the Information Privacy Principles. Office of the Privacy Commissioner. Accessed 5 October 2026.
  3. New Zealand issues long-awaited guidance for businesses on Artificial Intelligence (30 July 2025). AJ Park. Accessed 5 October 2026.

Related guides.

Want help with this? Say hello →