How to build an AI strategy in New Zealand, starting with an AI readiness assessment

Most New Zealand businesses don’t need a thick AI strategy document. They need clear answers to three questions: where would AI save real time or money, are we ready to use it safely, and what should we do first?

The Government’s AI strategy, released in July 2025, is clear that the first step is yours: “it is up to businesses, not government, to see the opportunity that AI presents for them”. It also notes that decision-makers in small and medium businesses may understand AI in principle but struggle to identify specific uses for their own business.

This guide shows how to build an AI strategy in six steps, from a first list of problems to a roadmap with sensible governance.

Start with problems, not tools

It’s tempting to start with a tool: a Copilot licence, a chatbot, an agent someone demonstrated. But a tool only pays off if it solves a problem you actually have, and AI tools change quickly. Your problems change much more slowly.

MBIE’s Responsible AI Guidance for Businesses starts here too, with your “why”: be clear on what you want AI to achieve, and look for inefficient or repetitive, rule-based tasks.

To find them, list your pain points, as business.govt.nz suggests:

• What takes longer than it should?
• What feels repetitive?
• What could be automated?

Aim for five to ten answers, each a problem you can measure (how long quotes take to go out, how many hours month-end reporting takes), not a tool. Spend an hour on this with the people who do the work.

Assess your readiness across six areas

An AI readiness assessment shows how well placed your business is to act on those opportunities, and what to fix first. It covers six areas:

A low score doesn’t mean you can’t start. It shows where your roadmap should begin, such as documenting a process, tidying one system of record or agreeing which tools staff may use.

For a quick indication, our free AI readiness check asks 12 questions across these six areas and takes about three minutes. A fuller assessment adds interviews and a walk-through of how work really gets done.

If you run a small business, the Government’s AI Advisory Pilot follows a similar path, from an AI readiness assessment to a roadmap by an approved provider, and can co-fund part of the cost for eligible businesses. Our guide to the AI Advisory Pilot explains who qualifies.

Find and rank your use cases

Next, turn your problem list into specific use cases: the task, who does it, how often, the hours it takes and what a good result looks like. Score each from 1 (low) to 3 (high) on:

• Value: time saved, errors avoided, faster responses or extra revenue.
• Effort: the data cleaning, integration, new software and change for staff it needs.
• Risk: whether it involves personal information, customers, or decisions about people and money.

Start with high value, low effort and low risk, which matches business.govt.nz’s advice for a first pilot. A half-day workshop with the people involved is a good way to agree the scores.

Treat risk with care. MBIE’s guidance says that when AI is helping to make big decisions, “like those about money, health, or the law”, a person should always be involved. Use cases where AI drafts or summarises for a person to check are safer first steps than ones where it acts alone.

Some need no new software: check whether tools you already use, such as Microsoft 365 or Google Workspace, include AI features. To choose between an automation, a chatbot and an agent, see our guide to the difference between the three.

Plan a simple roadmap: first, next, later

A roadmap can fit on one page. Sort each use case, and each foundation to fix, into three groups:

• First (the next three months): one or two quick wins, plus the basics (an AI use policy, a short list of approved tools and a named owner).
• Next (three to six months): extend what worked, connect systems and tidy the data that bigger opportunities need.
• Later (six to twelve months): larger or riskier projects, such as customer-facing chatbots or AI agents, once the foundations and governance are in place.

Give each item an owner, a measure of success, a rough budget and a review date. The AI Advisory Pilot’s roadmaps outline opportunities, priority use cases, recommended tools, expected costs and benefits, and implementation steps: a good checklist for any roadmap.

Run each “first” item as a short pilot, comparing it with your usual process on time, quality and risk, as business.govt.nz suggests. Expand step by step, based on evidence, and review the roadmap quarterly.

Build the business case

The Government’s AI strategy notes that many businesses remain uncertain about the return on investment for their own circumstances. A short business case for each priority use case addresses that. Allow an hour or two for each, covering:

• The baseline: how long the task takes today, how often errors happen and how long customers wait. Measure it before you change anything.
• The benefit: hours saved multiplied by what those hours cost you, plus faster turnaround, fewer errors or capacity for more work.
• The full cost: setup, licences, usage charges, training time and ongoing support. MBIE’s guidance notes that AI tools have different cost drivers from other software, so check the total cost at high usage.
• The way out: whether you could change providers later and take your data with you.

Agree upfront what result would justify going further, and what would make you stop. If you’re budgeting for outside help, see what AI consulting costs in New Zealand.

Put governance and privacy in place

You don’t need a compliance department, but you do need the basics. The Government’s AI strategy takes “a light-touch and principles-based approach”, relying on existing laws such as the Privacy Act 2020 rather than new AI rules. The Privacy Commissioner’s starting point is that “the Privacy Act applies to everyone using AI tools in New Zealand”. In practice:

• Appoint a privacy officer. Every business that handles personal information must have one, and no special qualification is needed.
• Assess before you start. The Commissioner recommends a privacy impact assessment before using AI tools with personal information, updated regularly.
• Meet the Commissioner’s expectations for generative AI, which include senior leadership approval, transparency with the people affected, engaging with Māori, human review before acting on outputs, and making sure the tool doesn’t retain or disclose personal or confidential information.
• Know about IPP3A. Since 1 May 2026, if you collect personal information about a person from someone else, you must take reasonable steps to tell them, unless an exception applies. That matters if a tool gathers information about people indirectly.
• Give staff simple rules. MBIE’s guidance suggests a usage policy for generative AI, and business.govt.nz recommends a short list of approved tools with business accounts and a “do-not-paste” list. Typing, pasting or uploading information into an AI tool all count as disclosures.

It’s also good business. In the Privacy Commissioner’s 2026 survey, 67% of New Zealanders were concerned about government agencies and businesses using AI to make decisions about people using personal data, and 66% would consider changing service providers over poor privacy and security practices. Our free AI policy template covers the staff rules.

Common mistakes to avoid

• Buying tools before defining the problem. Choose the tool last.
• Trying to do everything at once. Start with one area, prove the gain, then expand.
• Having no owner. Name one person to approve tools and manage access, as business.govt.nz recommends. That reduces the risk of “shadow AI”: staff using unapproved tools without oversight.
• Skipping the baseline. Without it, you can’t show what AI saved.
• Putting personal information into free tools. MBIE notes that some free or public tools may share what you type or upload with outside parties. If in doubt, the Commissioner recommends not using AI tools to handle personal information.
• Removing people too early. Keep a person reviewing anything that affects customers, staff, pricing or finances.
• Forgetting training. MBIE’s guidance suggests foundational responsible AI training for all staff.
• Treating the strategy as finished. Tools and risks change, so review your plan regularly.

Next steps

If you’d like an independent view, our AI strategy and readiness service covers each step in this guide, from the readiness assessment to a roadmap and business case you can act on. To see where you stand first, try the free AI readiness check, or book a free 30-minute consultation.