Practical AI for marae, iwi and kaupapa-led teams

Kia ora e te whānau,

AI isn’t magic, and it’s definitely not a replacement for the relational mahi your people do every day. Used carefully, it can help with the admin around the mahi.

AI is everywhere at the moment.

Every second post makes it sound like you’re either already behind or about to be replaced by a robot. That isn’t a helpful way to talk to marae, iwi organisations, hauora providers, social services or small kaupapa-led businesses doing real work with real people.

Our view is pretty simple. AI can help with some admin. It can help you get off the blank page. It can summarise, tidy, draft and organise.

It can’t replace judgement, tikanga, trust, lived experience or the human relationships at the centre of your mahi.

‍

Where AI is useful now

AI is good at first drafts.

It can help turn meeting notes into action points. It can draft a pānui for you to edit. It can summarise a long report. It can reshape a policy into simpler language. It can help compare funder guidelines with your own notes. It can pull themes out of feedback if the data is safe to use.

That’s useful, especially when your team is small and the admin pile isn’t.

‍

Where to be careful

Don’t put private whānau information into random AI tools.

That includes health details, case notes, sensitive referrals, financial hardship, whakapapa information, legal issues, complaints, or anything a whānau member trusted you with in a closed setting.

If you wouldn’t paste it into a public website, don’t paste it into an AI chat unless you know exactly where that data goes, who can access it, and what your organisation has agreed to.

‍

Keep humans in charge

AI can draft. A person should decide.

That means a real person checks the facts, the tone, the privacy risk and the cultural fit before anything goes out. Especially grant applications, reports, website copy, social posts, policies, letters and anything that speaks on behalf of your kaupapa.

The voice still needs to be yours. The final call still needs to be yours.

‍

Good first uses

Start with low risk tasks.

Summarise public funder guidelines. Draft a checklist from a meeting transcript that doesn’t include sensitive details. Turn a long pānui into a short social caption. Create a first draft of a thank you email. Build a list of questions to ask before choosing a CRM. Tidy a messy set of internal notes into headings.

These tasks save time without handing over trust.

‍

Set simple rules before you start

Your team doesn’t need a 40 page AI policy before using the tools, but you do need some clear rules.

What information can go into AI? What information can’t? Who can use it? Which tool is approved? Who checks the output? How do you record when AI helped with published or funder-facing work?

Write those rules in normal language. Make them easy for kaimahi to follow. Then revisit them as your use grows.

‍

The Hono way

We’re interested in practical AI, not hype.

If AI helps a kaimahi get a first draft done faster, good. If it helps a manager understand a long document before a meeting, good. If it helps a small team spend less time staring at a blank page, good.

But the kaupapa stays with people.

AI should sit in the background, helping with the admin around the mahi. It should never be the thing carrying the relationship.

‍

Common pātai

How can kaupapa-led teams use AI safely?

Use it for low-risk admin like first drafts, summaries and tidying notes, keep private whānau information out of it, and have a person check the facts, tone, privacy risk and cultural fit before anything goes out.

‍

What should never go into an AI tool?

Health details, case notes, sensitive referrals, financial hardship, whakapapa information, legal issues, complaints, or anything a whānau member trusted you with in a closed setting, unless you know exactly where that data goes and who can access it.

‍

Do we need an AI policy before we start?

Not a 40 page one, but a few clear rules. What information can and can’t go in, who can use it, which tool is approved, and who checks the output. Write them in simple language and revisit them as your use grows.