Black Friday for AI services: 20% off any one-off project. Three spots, sign and pay by 30 Nov. See the prices

Industries · Professional services · Accounting

AI for accountants and accounting firms.

Start with one job. Prove AI pays. Keep your clients’ data safe.

For Sydney accounting, tax and bookkeeping practices with 5 to 100 staff, where the filing, recaps and follow-ups after client calls still happen by hand.

Your path with AI as a practice: Audit. Build. Train. Partner.

The Korbai Method: find the one admin job where AI pays in your practice, prove it on your own clients’ work, and build from there. A practitioner checks every figure, every piece of advice and anything a client will read.

Not now Advice, lodgements or AML risk calls made by AI

Free 30-minute call: What happens next

Start here

  1. Talk: 30 minutes on how your business runs. No prep needed.
  2. Plan: I come back with the first step, and what it costs.
  3. Decide: go ahead or don’t. If AI won’t pay, I’ll tell you.
“It never felt like a sales pitch, it felt like a conversation with someone who knew what he was talking about.”
David Murphy Director & Residential Sales, Murphy Residential

01 Audit: Find the job that pays first

Fixed fee, fixed window

Generally the best first step. We follow the work from a new client to a lodged return, find the admin that eats the evenings, and prove the number on the one worth handing over.

What you get

A clear report and a decision you can act on

More on the Audit

Not now Advice, lodgements or AML risk calls made by AI

02 Build: Prove one workflow

First workflow live in weeks

We build the workflow that won the Audit, or the one you came to us for, in the practice software and Microsoft 365 you already run. One thing at a time, with client data set up deliberately.

What you get

One workflow, running live

More on AI automation

03 Train: Your team runs it

Checked again after 30 days

Your partners, managers and admin team learn to run it on their own client work, and to check every draft before it’s relied on.

What you get

Everyone owning one job that now runs with AI

More on AI training

04 Partner: Keep it on track

Ongoing, with regular check-ins

Ongoing advisory. AI tools and the rules on client data both move, and most practices don’t need a head of AI on payroll. They need one on call.

What you get

Shared accountability and a name on the outcome

More on ongoing advisory
  • One proven step at a time
  • Your clients’ data stays safe
  • Where AI won’t pay, I’ll tell you

What can AI do for accountants?

It takes the admin around client work off your team: filing call records and drafting the recap, chasing client documents, preparing onboarding and AML checks, and checking a file before it goes to review. AI drafts and a practitioner signs off, so advice, lodgements and AML decisions stay with your people. For the wider picture, including the TPB’s guidance, read our guide to AI in accounting.

After every client call

What happens after a client call, once it’s built?

The call’s record lands in the right client folder in your own document system, a recap is drafted for you to check and send, and the actions wait as dated tasks. It runs on the recorder, document system and task list you already use.

Example

Assistant Your call with Jo Kestrel has ended. File the record under client code KES01?

You Yes, KES01.

Assistant Filed. The recap email is in your drafts, and three tasks are waiting for your OK.

  • You confirm the client code

    It suggests the folder. Nothing is filed until a person confirms it.

  • The recap waits for you

    It drafts the client email. You read it, change it and send it.

  • Tasks with an owner and a date

    Each action becomes a task in your list, ready for you to approve.

  • Only what the job needs

    The recorder and the automation see the call, not your practice management data or anyone’s TFN.

We scope, build and look after it through our AI automation service.

Where AI pays

Where else does AI pay in an accounting practice?

Many practices are being handed AI before anyone has shown the team how to check it. These are the jobs we’d set up next, each with a person signing off.

Client documents, chased

  • A checklist per client of what’s still missing for the job.
  • Reminders go out on the schedule your manager sets.
  • Your manager sets up each list and decides who gets a reminder.

Onboarding and AML checks

  • The intake, ID request and documents are prepared up to the decision.
  • A suggested risk rating comes with its reasons.
  • A person makes the risk call and signs it off.

Checks before review

  • Your own review checklist, run over the file before the reviewer opens it.
  • Anything the checklist expects but the file doesn’t hold is flagged.
  • The reviewer still checks the work and signs it off.

Monthly packs and client updates

  • Drafted from your own templates and the client’s figures.
  • Only in a business account approved for client data.
  • The practitioner checks every figure before it goes out.

Start with the AI you already pay for

  • Microsoft 365 and your practice software may already do part of the job.
  • We set that up first, and only build where the work runs across systems.
  • Your team learns to use it with us, on their own client work.

Work out your own hours with the AI time savings calculator

How a build runs

What happens in the first weeks of a build?

Before anything new goes live, it’s set up, measured and tested on your own jobs.

  1. Start with what you already pay for

    • We set up and try the features already in your software, against pass marks you agree.
    • Anything that doesn’t earn its keep is switched back off.
    • A five-minute setting gets fixed on the spot, and we tell you that’s all it was.
  2. Count before we quote the build

    • We measure the job on your own records, not anyone’s memory.
    • For a practice, that might be how many client calls a week end with a record still to file by hand.
    • Each number makes the build’s price smaller, firmer or unnecessary.
  3. Test it where it can’t do harm

    • First on made-up records.
    • Then on real work, writing nothing, with its answers checked against your team’s.
    • Then a trial with your staff watching, before it’s kept.
  4. You stay in charge

    • Each build is agreed on its own, and you can stop after any one.
    • Nothing reaches a client unless one of you has approved it.
    • Accounts stay in your business’s name, and nobody shares a password.

Book a discovery call

Client stories

What clients say

“I came to Paul barely scratching the surface of AI and fairly sceptical as a journalist. In a brief session he cut through the noise - no jargon, no hype - and showed me practical ways to make it a really productive research tool. The tip to use it to dial in directly to my go-to information sources was worth the session alone.”
Sophie Barton · Freelance Journalist and Editor

More client stories

Where to be careful

What do we set up carefully in an accounting practice?

Advice, lodgements and AML decisions stay with your people. The rules below already apply to the client information AI touches, so the first workflow is built around them. General information, not tax or legal advice.

  • Client permission

    For tax practitioners, entering client information into an AI tool can count as disclosing it to a third party, depending on how the tool is set up, and that needs the client’s permission.

    • A signed engagement letter can carry the permission.
    • We list the tools and where each stores client data, so the letter can name them.

    Source: TPB(GS) 55/2026, July 2026

  • Tax file numbers

    The TFN Rule requires reasonable steps to protect TFN information from misuse and unauthorised access, and to limit access to the people who need it for tax purposes.

    • Automations see only the fields the job needs.
    • TFNs never go into a tool the practice hasn’t approved for client data.

    Source: Privacy (Tax File Number) Rule 2015, rule 11

  • Checked at every step

    The TPB expects AI output to be verified and reviewed for accuracy at each step of the workflow, with those steps documented.

    • Every draft is checked before anyone relies on it.
    • The check is recorded on the file, with who did it.

    Source: TPB(GS) 55/2026

  • Where client data is kept

    Under APES 110, keeping client information confidential covers protecting it as it’s collected, used, transferred, stored and destroyed.

    • We check where each tool stores data before client information goes in.
    • If a vendor’s page doesn’t say, we ask before the tool is used.

    Source: APES 110 Code of Ethics, paragraph 114.1 A1

Already using AI on your jobs? Check one use in about five minutes

Before you call

Not ready to talk yet?

Try this yourself this week

Find what you’ve promised

  • In Copilot, or another AI your practice has approved that can read your work email, ask: “What did I say I’d do in email this week that isn’t done yet?”
  • Check each item against the email it came from before you act on it.
  • If it earns its keep, save the prompt and run it each morning.

Our guide on where AI saves an accounting practice time, and where it doesn’t covers the TPB’s guidance, AML/CTF and engagement letters in full. For which assistant plan keeps client details private, read is it safe to put client information into AI tools.

FAQ

Questions we get asked

Will AI lodge anything or send advice to our clients?

No. Advice, lodgements and AML decisions stay with your people. AI drafts the file note, the email or the pack, and a practitioner checks it before it’s relied on or sent.

Our practice software and Microsoft 365 already have AI. Why would we need more?

You may not. We set up and try what you already pay for first, and tell you if that’s enough. We only build where the work runs across systems, such as the call recorder, the document system and the task list.

Do we need AI that runs on our own servers to keep client data safe?

For a practice your size, generally not: it costs too much to set up and run for the return. A business account with the right data settings, checked before client information goes in, is usually the answer.

Our IT provider is cautious about AI. Will you work with them?

Yes. The question is the cost of getting it wrong against the cost of getting it right, and we sit between “don’t touch it” and “anything goes”. Access and security calls are made with your IT provider involved.

Do you need logins to our systems?

Not to work out what to build: we see how the work runs by screen share. If a build needs access, it’s set up in your practice’s name, and our extra login comes off when testing ends.

Will the December Privacy Act change affect our onboarding software?

It might. From 10 December 2026, if the Privacy Act covers your practice, your privacy policy must describe decisions software makes, or does something substantially and directly related to making, using personal information, where they could significantly affect someone. A risk rating that shapes whether you take on a client may count.

Try the free AI and privacy check

Will our team be trained to use it?

Yes. Training comes with each build, on your team’s own client work. If the tools are already in place and it’s the habits that are missing, training can run on its own.

More on AI training

What does it cost?

Each build is quoted once it’s scoped, so you see the cost before you decide, and you can stop after any one. If you start with the Audit, it’s a fixed fee for a fixed window.

Get started

Which job would you hand over first?

Book a discovery call. Half an hour, no obligation. I come back with the first step and what it costs.

AI Readiness Score

Before you go, how ready is your business for AI?

Twelve questions, three minutes, scored on the spot. No email needed to see your result.

Score your business