AI dictation for NZ lawyers: what to know before you adopt it
AI dictation tools can save NZ lawyers significant time on attendance notes and file notes. Here is what to evaluate before adopting one, including privacy and data residency considerations.
AI dictation tools have improved significantly in the last two years. For NZ lawyers, the practical case is straightforward: attendance notes, file notes, and first drafts of correspondence that used to take 20 minutes to type can now be produced from a voice recording in a fraction of the time.
The question is not whether the technology works. It is which tools are appropriate for a NZ law firm environment, and what you need to check before rolling one out.
What AI dictation actually does
Traditional dictation tools, like older versions of Dragon, converted speech to text with no understanding of context. AI dictation tools go further. They can transcribe, summarise, identify speakers, and produce structured output from a recording.
In a legal context, this means a lawyer can record a client meeting, a phone call, or a verbal summary into their phone, and receive back a structured attendance note ready for review and filing. The time saving is real, particularly for fee earners billing time that would otherwise go unrecorded.
Tools worth evaluating
Microsoft 365 Copilot integrates directly with Teams, Outlook, and Word. For firms already on Microsoft 365, it is the lowest-friction option. Copilot can transcribe Teams calls, summarise meetings, and draft documents from prompts. Data is designed to stay within your Microsoft 365 tenant, which is a better starting position than most third-party tools. That said, see the privacy section below before treating it as risk-free.
Otter.ai is a standalone transcription and summarisation tool. It is widely used and capable, but data residency is worth checking for NZ law firm use. Client information recorded and processed through a US-based service raises NZ Privacy Act 2020 questions depending on what is in the recording.
Nuance Dragon Legal Anywhere (now part of Microsoft) is a cloud-based option built specifically for legal terminology. It has a strong reputation for accuracy with legal language and runs on Microsoft's cloud infrastructure, which simplifies the data residency question for firms already in the Microsoft ecosystem.
Philips SpeechLive is a cloud dictation platform with a long track record in professional legal and healthcare environments. It supports recording via smartphone, desktop software, or dedicated Philips hardware, and routes audio to automatic transcription or to a transcriptionist for manual turnaround. Data is processed and stored on European servers, which is a cleaner starting position on data residency than US-hosted alternatives. Worth evaluating for firms that want a purpose-built dictation workflow, or those already using Philips pocket memo recorders who want to move to a cloud-based system.
Other tools: A growing number of PMS vendors and third parties are building AI dictation and note-taking features directly into their products. If your PMS has a native option, evaluate that first before adding a separate tool.
NZ Privacy Act 2020 considerations
This is where most NZ law firms underinvest in their evaluation.
Client meetings contain privileged information. When you record a meeting and send it to a cloud-based AI service for transcription, that data is leaving your environment. The questions to answer before adopting any tool:
- Where is the data processed and stored?
- Does the vendor use your data to train their models?
- What are the retention policies for recordings and transcripts?
- Does this constitute a disclosure under the NZ Privacy Act 2020?
Microsoft 365 Copilot is architecturally designed to keep data within your Microsoft 365 tenant, and for most firms it does. But two issues are worth knowing about before treating it as straightforward.
The first is a documented incident. In early 2024, a server-side logic error in Microsoft 365 Copilot (tracked as advisory CW1226324) caused the AI to index and process emails explicitly marked as "Confidential," bypassing the sensitivity labels that should have blocked access. The bug affected Sent Items and Drafts for a period of weeks before Microsoft deployed a fix. Affected tenants were notified, but organisations had no way to know their confidential emails had been processed during that window. Microsoft has since patched the issue, but it demonstrated that tenant boundary protections can fail and that vendor advisories are not always surfaced through standard monitoring.
The second is structural. Copilot operates on whatever permissions a user already holds in Microsoft 365. It does not apply independent judgment about whether access is appropriate. Most Microsoft 365 tenants have accumulated years of permission drift: documents shared broadly, folders inherited from old projects, SharePoint sites that everyone can read. Before Copilot, that over-permissioning was a latent risk. Copilot makes it active by surfacing documents a user technically has access to but would never have found manually. In a law firm, that could mean client files from matters a fee earner was not involved in appearing in AI-generated responses.
These are not reasons to avoid Copilot. They are reasons to do permission and sensitivity label hygiene before deploying it. Confirm your DLP policies and sensitivity labels are correctly configured and test that they actually block Copilot access as intended. Cloud-based third-party tools require more scrutiny still, but Copilot is not zero-effort from a governance standpoint.
Legal terminology accuracy
General-purpose AI dictation tools are trained on broad datasets. Legal terminology, NZ-specific terms, and firm-specific naming conventions may be transcribed incorrectly. Any AI-produced attendance note should be reviewed before filing. Build that step into your workflow from the start rather than treating it as optional.
What to do before rolling out a tool
Before adopting any AI dictation tool across a firm:
- Check data residency and processing location
- Review the vendor's privacy policy specifically for model training opt-outs
- Confirm whether client consent is needed for recorded meetings under your firm's obligations
- Pilot with internal meetings before using with client calls
- Set a review step before any AI-produced note is filed or sent
Getting independent advice
Valley IT helps NZ law firms evaluate and deploy AI and productivity tools within a Microsoft 365 environment. We can review your current setup and advise on which tools fit your security and privacy requirements.
Book a free consultation if your firm is looking at AI dictation and wants an independent view before committing to a tool.
This post reflects publicly available information and IT practitioner experience as at May 2026. AI tool capabilities and privacy policies change frequently. Verify current data residency and privacy terms directly with each vendor before adopting any tool.