AI for New Zealand accounting firms.

Where AI is useful in an accounting practice

AI pays off in a practice wherever staff move information between clients, inboxes and the ledger by hand: reading what clients send, matching it to a checklist, coding bills, spotting what looks wrong and turning numbers into words. The ledger work itself is increasingly automated inside Xero and MYOB, so the opportunity for most firms sits in the gaps between those systems.

Use is already widespread. In a survey commissioned by Xero and published on 12 October 2026, 78% of 526 senior NZ accounting and bookkeeping professionals said their practice had tried AI and 49% used it in day-to-day work, while 94% did not expect AI to reduce team numbers in the next year. It is a vendor survey, part panel and part Xero customers, so read it as direction rather than a census. One finding is worth acting on: 89% of practices using AI every day said their core processes were well documented, against 30% of practices not using it.

Accounting workflow map

Worked guides cover the first three in detail: collecting client documents for year end, coding invoices and reviewing exceptions and preparing month-end reports.

The rules and professional bodies that apply

There is no AI-specific rule for accountants in New Zealand, and we found no AI guidance from the External Reporting Board. Existing obligations do the work.

• Privacy Act 2020. Practice files hold salaries, IRD numbers, family trust details and health-related expense claims. If an AI provider only processes that data for you, the Privacy Commissioner says your organisation remains fully responsible; if the provider uses it for its own purposes, such as training, sharing it may be a disclosure. The Commissioner’s fallback advice is blunt: “If in doubt, we recommend you do not use AI tools to handle personal information.”
• AML/CFT Act 2009. An accounting practice is a designated non-financial business or profession when it carries out specified activities, such as acting as a formation agent or providing a registered office. Reporting entities must conduct customer due diligence (s 11) and keep identity and verification records for at least five years after the relationship ends (s 50), under the Act as at 1 July 2026. The Department of Internal Affairs became the sole AML/CFT supervisor on 1 July 2026.
• Professional standards. CA ANZ launched an AI Fluency Playbook for members in 2025, reported by Chartered Accountants Worldwide (CA ANZ’s own page was not reachable when we checked). Your body’s confidentiality and due-care principles apply to AI-assisted work exactly as they do to a junior’s work.

Take advice from your professional body or a lawyer on AML/CFT scope and engagement terms.

Xero, MYOB and the rest of the practice stack

Use the native features for what they do well and build only for the work between systems, which is usually document processing across email, portals and PDFs. As listed on 19 October 2026:

• Xero JAX captures bills and receipts, sends invoices and reconciles transactions, with each action logged for review. Xero claims 97% reconciliation accuracy. Even at that rate, a client with 1,000 bank lines a month leaves roughly 30 for a person. Xero also offers a read-only connector for Claude.
• MYOB Iris offers Ask Iris, a question-and-answer feature available with Lite, Pro and AccountRight, with AI business insights in beta.
• Practice tools. Xero Practice Manager or MYOB Practice for jobs and deadlines, a capture tool for receipts, and Microsoft 365 or Google Workspace for email and workpapers.

Our notes on Xero and MYOB cover the APIs and permission scopes, and the custom versus off-the-shelf comparison helps decide when a native feature is enough.

A first project: year-end client document collection

Document chasing is the safest place to start. Nothing is filed, the checklist already exists, and the measure is obvious.

1. Pick a client segment. Start with 40 to 60 simple files, such as rental property owners or sole traders, where the year-end checklist is stable.
2. Write the checklist as rules. List each document by client type, with what makes it acceptable: period covered, account number, signed or unsigned.
3. Let AI read and draft. The system reads what arrives, ticks items off, and drafts the chaser. Staff approve every send for the first season.
4. Audit a sample. Check one in ten files marked complete against the documents themselves.
5. Compare with last year. Track days to complete file and staff time spent chasing.

Illustrative example: a practice prepares 400 individual and small business returns, each needing about 2.5 chasers that take 10 minutes to write and check. That is 400 × 2.5 × 10 = 10,000 minutes, about 167 hours a season. If staff spend 2 minutes approving each drafted chaser instead, the total falls to about 33 hours, a saving near 133 hours before setup and audit time. Our AI workflow automation service builds this kind of loop, and the lesson on calculating time saved by AI shows how to measure it honestly.

Where AI fails in accounting work

Workflows we map in this sector

• Invoice coding and exception reviewSupplier bills coded to account, tax rate and tracking, with only the exceptions sent to a person.
• Month end reporting preparationClose checks run overnight and variance commentary drafted from the ledger, ready for an accountant to edit.
• Client document collectionYear-end lists built from last year’s file, uploads checked for completeness, reminders that name what is missing.

Written by the Levy Street team in Wellington. Last updated 19 October 2026.