Guide · AI for law firms
Automate the work around the law, not the legal judgement: client intake, first drafts from your own templates, sorting incoming documents and email, and research a lawyer then checks. A New Zealand firm cut its time per client from 12 hours to 3 this way. The Privacy Act 2020 doesn't stop you using AI, but the firm stays responsible for client information whichever tool touches it, so where the data goes and who signs off matter as much as the tool.
Use AI for the work that surrounds legal judgement, and keep the judgement, and the sign-off, with a lawyer. In a typical small practice that means client intake, first drafts from the firm's own templates, the documents and emails that pour in on every matter, and the first pass of research.
The Privacy Act 2020 doesn't stand in the way of any of that. What it does is make the firm answerable for client information whichever tool handles it. So the first question isn't which AI is cleverest. It's where the data goes, who can see it, and who checks the output.
One New Zealand firm went from 12 hours of work per client to 3, after losing a big part of its legal team. The firm is anonymised, as it is in the published case study.
| Task | Before | After | What was built |
|---|---|---|---|
| Client intake | 40 minutes per client | 15 minutes | Online intake forms, automated scheduling, call transcripts summarised. |
| Affidavits | 4 hours per client | Under 1 hour | Drafts generated from the recorded intake call and the firm's own templates, then client review. |
| Legal submissions | Up to 6 hours | About 1 hour | Draft generation, suggested case law, template formatting. |
| Documents and email | Read and filed by hand | Classified and filed automatically | Email processing, document classification, alerts on anything missing. |
Across the firm that's more than 50 hours a month back. The work took about eight weeks: four of discovery and planning, three of building, and one of testing and training. The firm could keep taking on clients with a smaller team, which was the point.
Anything that is advice, anything filed, and anything sent to a client in the firm's name needs a lawyer's eyes before it goes.
The balance matters both ways. Too cautious and you spend more time reviewing than you save. Too relaxed and you run the risk of real damage. In the firms I talk to it's normally the second one I'd worry about.
The Privacy Act doesn't name AI, but its information privacy principles apply to every tool that touches client information, and the firm stays responsible. In plain terms:
On top of that sit a lawyer's own duties of confidence and the protection of privileged material. None of this is a reason not to use AI. It's a reason to choose the set-up deliberately, before staff choose it for you.
Six questions cover most of what a firm needs to know before client information goes anywhere near an AI tool.
| Ask | A good answer |
|---|---|
| Is our data used to train the model? | A plain no, backed by the provider's business terms. |
| Which country is it stored and processed in? | A named country and provider, not "the cloud". |
| Who can see what? | Access by role and matter, so a junior can't browse every file and staff can't read each other's chats. |
| How long is it kept? | A retention period you set, and a way to delete on request. |
| Is there a log? | Every document read and every draft produced, with who approved it. |
| What if we stop using you? | The data, the code and the accounts are already the firm's. |
For anything I build now, the default is AWS in Sydney, in a cloud account the firm owns, with access controls and a full log. If staff are already pasting client details into their own ChatGPT accounts, start with this guide on personal AI accounts.
Start with the one step that eats the most time or causes the most rework, measure it, and automate that before anything else. In most practices it's intake or the first draft of a standard document. Both are low risk, because a lawyer reviews the result anyway.
| The job | Fixed price, as of October 2026 |
|---|---|
| Intake, scheduling and email handling across the tools you already run | NZ$500 to $5,000 |
| Reading, drafting and filing documents and email | NZ$1,000 to $8,000 |
| An internal assistant that answers from the firm's own precedents and files | NZ$3,000 to $12,000 |
| Hosting, monitoring and support | From NZ$99 a month |
Prices exclude GST. AI usage is billed at cost on top and is usually a few dollars to a few tens of dollars a month for a small firm.
Yes. Nothing in New Zealand law bans it, and the New Zealand Law Society and the courts have both published guidance on using generative AI. The duties don't change: client confidentiality, competence, and responsibility for anything filed or sent. In practice that means business-grade tools that don't train on your data, no client details in free personal accounts, and a lawyer checking every output.
No. It requires the firm to keep personal information secure, accurate and used only for the purpose it was collected for, wherever it is processed. A cloud provider that only stores or processes data on the firm's behalf is generally treated as holding it for the firm, so the firm stays responsible. Know which country the data sits in, what the provider's terms allow, and how a breach would be reported to you.
It can, if it is asked to cite from memory. That's the main risk in legal research, and there have been court cases overseas where lawyers filed submissions with invented citations. The safer set-up has the AI suggest authorities from a defined source the firm trusts, links every citation back to that source, and leaves checking each one to a lawyer before anything is filed.
The Privacy Act requires you to tell people why you collect their information and who will receive it, so a short line in your privacy statement and engagement letter covering the tools you use is sensible. The courts' guidance for lawyers puts the responsibility for accuracy on the lawyer rather than requiring a disclosure in every case. If in doubt, check the current guidance from the Law Society.
As of October 2026, automation of intake, documents and email from Ez-AI is NZ$1,000 to $8,000 as a fixed price for most jobs, and an internal assistant that answers from the firm's own documents is NZ$3,000 to $12,000. Hosting, monitoring and support start at NZ$99 a month. Prices exclude GST and discovery is free.
Walk me through one matter from first call to filed document on a free 30-minute call. I'll tell you which steps are worth automating, which aren't, and what it would cost as a fixed price. Discovery is always free.
Book a free call