What is AI? A plain English guide for New Zealand business.
What artificial intelligence is, the types you will meet at work, what it is not, and how New Zealand organisations are using it safely.
The short answer
Artificial intelligence (AI) is software that learns patterns from data so it can recognise, predict, decide or create.
For New Zealand businesses it shows up in tools like ChatGPT, Claude and Microsoft Copilot, and inside software you already use for accounting, CRM and email.
Most firms use it to draft, analyse, answer and automate everyday work.
Key facts.
- 91 percent of New Zealand organisations with 100 or more staff use AI, but only 15 percent are scaling it across the organisation.Datacom State of AI Index 2026, survey of 207 leaders, July to August 2026 · source
- New Zealand has no AI-specific law. The Government's AI Strategy takes a light-touch, principles-based approach that relies on existing law.MBIE, July 2025 · source
- The Privacy Commissioner's starting point is that the Privacy Act 2020 applies to everyone using AI tools in New Zealand.Office of the Privacy Commissioner, September 2023 · source
- The OECD defines an AI system as a machine-based system that infers from its inputs how to generate outputs such as predictions, content, recommendations or decisions.OECD, November 2023 · source
- Public service agencies have a published AI framework, with five principles, that private firms can use as a benchmark.Government Chief Digital Officer, 2025 · source
- Large language models generate likely-sounding text, so they can state errors confidently. Human review stays essential.AI New Zealand · why AI models hallucinate
What is AI, in plain English?
AI is software that works things out from examples instead of following rules a person wrote line by line.
Traditional software does exactly what it is told: if the invoice total is over $5,000, send it for approval. Artificial intelligence is different. It is shown thousands or billions of examples, finds the patterns in them, and uses those patterns on things it has not seen before. That is why it can read a messy email, spot a damaged fruit on a packing line, or draft a reply in your tone of voice.
The OECD's definition, which many governments use, puts it this way: an AI system takes inputs and works out how to produce outputs such as predictions, content, recommendations or decisions. The useful word there is infers. AI makes a best guess, which is why it is powerful and why it can be wrong.
For most New Zealand businesses, "AI" now means one of two things. The AI built quietly into software you already pay for, such as fraud checks in your bank, bank feed matching in your accounting software, or spam filtering in email. And the generative AI assistants your people use directly, like ChatGPT, Claude, Microsoft Copilot and Google Gemini. This guide covers both, and our AI glossary defines more than 100 terms if you want to go deeper.
How do the types of AI fit together?
Think of AI as a set of nested circles: each newer technique sits inside the one before it.
- Computer vision
- Speech recognition
- Natural language processing
- AI agents
The layers are techniques. The applications on the right draw on whichever layers suit the job.
The practical point: when a vendor says their product "has AI", ask which layer. A machine learning model that flags late payers is very different, in cost, risk and what it can do, from a large language model that writes your customer replies.
What are the main types of AI?
Eight types cover almost everything a New Zealand business will buy or use. The examples below are illustrative, by sector.
Machine learning
Software that learns patterns from past data to predict or classify, instead of following fixed rules.
- In a NZ business
- A retailer forecasting stock for summer; an accounting practice flagging unusual transactions.
- Where you meet it
- Built into accounting, banking and CRM platforms, usually switched on by default.
Deep learning
Machine learning that uses layered neural networks. It powers most modern AI, including everything below.
- In a NZ business
- The engine under the bonnet. Rarely bought on its own.
- Where you meet it
- Inside the tools in the rest of this list.
Computer vision
AI that interprets photos and video: spotting, counting, grading and reading.
- In a NZ business
- Grading fruit in a packhouse; reading supplier invoices into the accounts; checking site photos for safety gear on a construction job.
- Where you meet it
- Text recognition (OCR) in Microsoft 365, image input in Copilot and ChatGPT, and specialist industry systems.
Speech recognition
Turns speech into text, and from there into notes and actions.
- In a NZ business
- Meeting notes and action lists from Teams calls, captured with everyone's consent.
- Where you meet it
- Microsoft Teams transcription and recap, and note-taking apps such as Granola.
Natural language processing (NLP)
AI that reads and understands text: sorting it, pulling details out of it, and judging its tone.
- In a NZ business
- Triaging inbound email by topic and urgency, so the right person sees it first.
- Where you meet it
- Helpdesk and email tools, and the virtual property manager in our Aspire case story.
Large language models (LLMs)
AI trained on huge amounts of text to predict the next word, so it can read, write, summarise and reason across documents.
- In a NZ business
- Drafting proposals, answering customer email, summarising a 60 page policy before a board meeting.
- Tools to know
- ChatGPT, Claude, Microsoft Copilot, Google Gemini and Numa. Our side by side comparison of AI tools helps you choose.
Generative AI
The umbrella for AI that creates new content. Text is the most common; these are the other five formats you will meet.
| Format | What it does | In a NZ business | Tools |
|---|---|---|---|
| Images | Creates or edits images from a description | Product shots, social graphics, concept boards | Adobe Firefly, ChatGPT images, Microsoft Designer |
| Video | Short clips from text or images | Social ads, explainer clips, storyboards | Google Veo, Runway |
| Music and audio | Music and sound from a prompt | Background tracks for videos and events | Suno, Udio |
| Voice | Text to speech and voice cloning | Training narration, phone menus, accessibility | ElevenLabs, Azure AI Speech |
| Avatars | AI presenters that speak a script on video | Training modules, multilingual explainers | Synthesia, HeyGen |
Want documents that look like yours rather than like AI? Our free toolkit shows how to make AI documents that match your brand. Rights in AI-generated images, music and video are unsettled, so set rules in your AI policy before you publish.
AI agents
Agentic AI takes steps on its own: it reads, decides and acts across your systems, with a person approving what matters.
- In a NZ business
- A virtual property manager drafting tenant replies from the company's own procedures, for a person to check and send.
- Tools to know
- Copilot Studio, Claude and Numa. Read more in our guide to AI agents and our CEO briefing on agents doing actual work.
What is not AI?
Plenty of useful automation is not AI, and plenty of products use the label loosely.
- Rules-based automation. Robotic process automation (RPA) and workflow tools follow steps a person defined. They are fast and predictable, but they do not learn. Often they are the right answer.
- Basic chatbots. A website bot that offers a menu of fixed answers is a decision tree, not AI. If it cannot handle a question it was not scripted for, it is not using a language model.
- "AI" as a marketing label. Ask the vendor what the AI actually does, what data it learns from, and where that data is processed. A clear answer is a good sign.
- A replacement for judgement. AI predicts and drafts. A person still decides, signs and is accountable.
How does AI work in New Zealand?
The tools are global, but the law, the data questions and the adoption pattern are local.
Adoption is wide but shallow
The Datacom State of AI Index 2026 found 91 percent of New Zealand organisations with 100 or more staff now use AI, up from 87 percent in 2025, but only 15 percent are scaling it across the organisation. There is no official Stats NZ figure yet: AI questions are being added to its business survey in 2026. Our own read of the national numbers is in AI in Aotearoa by the numbers and New Zealand AI adoption in 2025.
The law that applies
New Zealand has no AI-specific Act. The Government's AI Strategy takes a light-touch approach that relies on existing law. The most important of those for business is the Privacy Act 2020: the Privacy Commissioner's guidance on AI tools says the Act applies to everyone using AI tools, and that its principles apply at every stage, from the data a tool learns from to your prompts and its outputs. Consumer, employment and copyright law apply too. Our guide to AI and the Privacy Act goes through it in plain English.
Not legal advice. This page explains the general position. For decisions about your organisation, talk to a lawyer.
Where your data goes
Most AI tools process data offshore, usually in Australia, the United States or Europe. Business plans of ChatGPT, Claude, Copilot and Gemini say they do not train their models on your business data by default; free personal accounts work differently. Check the plan, not just the product, and read our explainer on trust concerns with popular AI assistants.
Creative rights
Who owns AI-generated images, music and video is unsettled in New Zealand. The Copyright Act 1994 has a provision for computer-generated works, naming the person who made the arrangements as the author, but it has not been tested against modern generative AI. Treat it as a policy decision, not a given.
How are NZ businesses using AI?
The common wins are unglamorous: drafting, answering, summarising and sorting.
Client case story · Property management
Aspire Property Management: a virtual property manager
Aspire built an AI assistant into its own portal that drafts email and WhatsApp replies from its procedures, the Residential Tenancies Act and its insurance policies, for property managers to check and send. Read the Aspire story.
Professional services
Proposals and reports
Turning meeting notes and data into a first-draft proposal or report, which a senior person then edits.
Sales teams
Call coaching
Reviewing call transcripts against a sales method. Our free toolkit shows how to coach sales calls with ChatGPT, Claude or Copilot.
Marketing
Content that stays on brand
Drafting and adapting posts, newsletters and web copy from a shared brand guide.
Operations
Knowledge on tap
Answering staff questions from the company's own procedures instead of a colleague's memory.
For more, with time saved per task across the use cases we have scoped with clients, see AI examples from New Zealand businesses.
Our view
Everyone is adopting AI. The winners are adapting. Adoption makes the work you already do faster. Adaptation redesigns how the work is done. Almost all the effort in New Zealand is going into the first, and almost all the value sits in the second.
That distance, between the 91 percent of larger organisations using AI and the 15 percent scaling it, is what we call the Adaptation Gap. It is rarely a tooling or training problem. It is an operating model problem: AI does not fix a broken process, it speeds it up.
Two things follow. Governance is an accelerant, not a brake, because nobody can safely say yes to AI in a real workflow without a policy and a data boundary. And "just experimenting" is the riskiest place to sit, because company information is already flowing into tools nobody is watching. Our AI Enablement Playbook sets out the six steps we use.
What are the risks of using AI?
Four risks matter most, and each has a practical guardrail.
- Accuracy. Language models can hallucinate: state something false with confidence. Keep a person in the loop for anything that leaves the building, and read why AI models hallucinate and how to prevent it.
- Privacy. Personal information in a prompt is still personal information. Use business plans, set a data boundary, and do a privacy impact assessment before AI touches personal information, as the Privacy Commissioner recommends.
- Security. Agents that read email and documents can be tricked by hidden instructions, known as prompt injection. Prompts are not controls: limit what an agent can access and approve what it does.
- Bias and rights. Models reflect their training data, and rights in generated content are unsettled. Check outputs about people, and agree rules for images, voices and likenesses before you use them.
The free AI Champions Committee resource pack gives you a governance structure, and our master prompt drafts a starting AI policy in minutes.
Where should a NZ business start?
Start one stage up from where you are, not five.
- Find out where you stand. The free AI Benchmark is 40 questions, takes about 20 minutes, and shows where you stand and the next move.
- Name an AI Leader and set a policy. One person owns adoption, and a short policy says what is and is not allowed.
- Pick two or three real tasks. Repetitive, time-consuming work with a clear owner. Tie each to a number you already track.
- Give the AI your context. Our free setup checklists for Claude and Copilot show how to load your procedures, tone and templates once, for everyone.
- Train the team, hands on. A Lunch n Learn gets a team started, and the AI at Work Foundations course builds the everyday skills.
Frequently asked questions
What is the difference between AI and generative AI?
AI is the whole field of software that learns from data to recognise, predict or decide. Generative AI is one part of it: models that create new content, such as text, images, audio or video. Every generative AI tool is AI, but most AI in business software, like fraud detection or demand forecasting, is not generative.
Is ChatGPT the same as AI?
No. ChatGPT is one product, built on OpenAI's large language models. It is a type of generative AI. Claude, Microsoft Copilot and Google Gemini are similar assistants from other companies, and much of the AI businesses use sits inside other software and has nothing to do with chat.
Is it legal to use AI in a New Zealand business?
Yes. New Zealand has no AI-specific law, but existing law applies, especially the Privacy Act 2020. If you put personal information into an AI tool, you are responsible for it under the information privacy principles. This is general information, not legal advice.
Will AI replace jobs in my business?
In our experience roles change more than they disappear. AI takes on drafting, sorting and summarising, and people move to checking, deciding and handling the exceptions. The businesses that get the most from it redesign the work around that split rather than cutting people first.
Is our data safe in AI tools?
It depends on the plan more than the product. Business and enterprise plans of the major assistants commit not to train on your data by default and offer admin controls. Free personal accounts often do not. Set a policy on which tools staff may use, and with what information.
How much does AI cost for a small business?
Most business AI assistants are sold per user per month, and prices change often, so check each vendor's pricing page. The larger cost is usually time: setting the tools up with your context and training people to use them well.
Where can I learn the AI terms used on this page?
Our AI glossary defines more than 100 terms in plain English, from algorithm to zero-shot learning, each with a link you can share.
Sources
- State of AI Index 2026, Datacom (fieldwork 28 July to 3 August 2026; accessed 19 October 2026)
- State of AI Index 2025, Datacom (fieldwork July 2025; accessed 19 October 2026)
- Artificial intelligence (AI), Office of the Privacy Commissioner (September 2023; accessed 19 October 2026)
- Privacy Act 2020, Parliamentary Counsel Office (version as at 1 October 2026; accessed 19 October 2026)
- New Zealand's Strategy for Artificial Intelligence: Investing with Confidence, MBIE (July 2025; accessed 19 October 2026)
- Public Service AI Framework, Government Chief Digital Officer (January 2025; accessed 19 October 2026)
- Updates to the OECD's definition of an AI system, OECD.AI (November 2023; accessed 19 October 2026)
- Copyright Act 1994, section 5(2)(a), Parliamentary Counsel Office (version as at 13 November 2025; accessed 19 October 2026)
- About the Business Operations Survey, Stats NZ (April 2026; accessed 19 October 2026)
- Recap in Microsoft Teams, Microsoft (product page; accessed 19 October 2026)
- Copilot Studio, Microsoft (product page; accessed 19 October 2026)
- Gemini for Google Workspace, Google (product page; accessed 19 October 2026)
- Veo, Google DeepMind (product page; accessed 19 October 2026)
- Firefly, Adobe (product page; accessed 19 October 2026)
- ElevenLabs for business, ElevenLabs (product page; accessed 19 October 2026)
- Synthesia, Synthesia (product page; accessed 19 October 2026)
Related topics
Keep going.
Your next step
See where your organisation stands on AI.
The free NZ AI Benchmark takes about 20 minutes and shows the one move that will make the biggest difference next.
Changelog: 19 October 2026, first published.