Be Market Be Market Be Digital Be Online Be Online Be Cloud Be Cloud Be Secure Be Secure Be Connect Be Connect Bea Bea Be AI Be AI Be Watch Be Watch Be Screen Be Screen Pay Now

// AI

AI in a New Zealand accounting practice: what actually works in 2026, and what does not

The four AI uses that pay back in a NZ accounting practice this year, the three that waste a partner’s time, and the 90-day plan that keeps client data where the Privacy Act and your professional body expect it.

By the Be Digital team 3 min read

The short version

  • The uses that pay back in this sector are Client correspondence drafting, Workpaper and meeting summaries and Year-end query lists.
  • The ones that do not: Letting AI file anything. — see why below.
  • Start with one workflow, a 30-day pilot, and the data permissions fixed first.
  • Governance is a one-page policy, not a project.

Accounting practices run on documents, deadlines and correspondence, which makes them one of the best fits for AI in New Zealand — and one of the easiest places to get it wrong, because client data is involved in every task. Here is what we see working in practices from four partners to forty.

What actually works

Use What it does What it needs
Client correspondence drafting Copilot drafts replies to client emails from the thread and the file, in your tone, for a human to check and send Microsoft 365 Business Standard/Premium + Copilot; permissions reviewed
Workpaper and meeting summaries Summarises long email chains, Teams meetings and PDF bundles into the points that matter for the file Teams transcription on; documents in SharePoint
Year-end query lists Builds the list of missing items per client from last year’s file and this year’s ledger Clean SharePoint structure per client
Bank rec and anomaly flags Xero/MYOB native AI suggests matches and flags unusual transactions Already in the software; just switch on and review

What does not work yet (and why)

  • Letting AI file anything. Every output to IRD or a client needs a human sign-off. The tools are drafters, not preparers.
  • Consumer chat tools for client work. The convenience is real and the Privacy Act exposure is total. Use the tenant-bound tools or nothing.
  • “Ask the AI about tax law.” It will answer confidently and sometimes wrongly about NZ specifics. Use it to draft from sources you supply, not to be the source.

What it costs, and what it gives back

Cost. Copilot licences for the three to six people who write most, on top of Microsoft 365 Business Standard or Premium; a one-off permissions and file clean-up; a couple of hours of training per person. For a six-partner practice that is a low five-figure annual commitment once the pilot has proved it.

Return. Practices we work with report two to four hours a week back per fee-earner on correspondence and summaries in the first quarter, and a noticeably shorter year-end query cycle. The return shows up as capacity — the same team handling more clients at busy season — rather than as a line on the P&L.

Risk if skipped. Staff use consumer tools anyway, with client data, and nobody knows.

The 90-day plan

  1. Weeks 1–2: Permissions review on SharePoint; move any client files off the old server; write the one-page policy.
  2. Weeks 3–6: Copilot pilot with the two people who write the most client correspondence. Time the tasks.
  3. Weeks 7–10: Switch on the native AI in Xero/MYOB for every client; train the team on review, not generation.
  4. Weeks 11–13: Decide licences per role; add a line to engagement letters about AI use; schedule a quarterly review.

Governance, in one page

Before any of this: a one-page AI policy saying which tools are approved, what may never be pasted into them (client identifiers, health, financial and legal detail unless the tool is inside your Microsoft tenant), and who owns it. We cover the page itself in a separate article, and Be AI aligns the work to ISO/IEC 42001 and the NIST AI RMF so the answer to “is this governed?” is yes.

Every use above runs inside DigitalShield, on the Microsoft 365 you already pay for, with the permissions and backups done first. If you want the pilot scoped for your practice, see the industry plan or talk to a Be AI consultant — the readiness check is free.

Questions people ask

Can we put client financial data into ChatGPT?

Not into a consumer tool. Client data belongs in tools that run inside your own Microsoft 365 tenant with enterprise terms u2014 Microsoft 365 Copilot or an approved assistant u2014 never in a free chat window. Your one-page AI policy should say exactly that.

Will AI replace junior accountants?

It is replacing the parts of the junior role that were data entry and first-draft summaries, and making the review and advisory parts more valuable. Practices using it well are hiring the same number of people and doing more with them.

What about Xero and MYOBu2019s own AI features?

Use them u2014 bank reconciliation suggestions and anomaly flags are mature and sit inside tools with proper terms. The gap they do not fill is correspondence, workpaper summaries and advisory drafting, which is where Copilot earns its licence.

How do we handle IRD, CA ANZ and Privacy Act obligations?

Keep client information inside tenant-bound tools, log what AI was used for on an engagement, keep a human sign-off on every output that goes to a client or IRD, and write it down as policy. That satisfies the practical expectations of all three.

// More insights

Keep reading.

// Ready when you are

Want this done, not just explained?

Tell us what your business runs on today. We will map what DigitalShield would cover, what it would cost per user, and what we would fix first — free, and in plain language.

  • Free, no obligation
  • You keep the findings
  • A real engineer, not a call centre