We use AI every day. It makes our team faster and it catches things we would otherwise miss. It also changes where your information goes, and you should know how

We have told other businesses to be wary of any finance provider that can't say which AI tools it uses, where the data sits, or who checks the output. Hold us to the same test. Here are our answers

The short version

AI helps with the work. People do the work, and people own it

Nothing AI produces reaches you, or the ATO, until a qualified person on our team has checked it. We are a registered BAS agent. The Tax Practitioners Board's Code of Professional Conduct applies to everything we do, whether a tool was involved or not. If something is wrong, it is on us. The Digit Promise has no AI exception

Where AI helps, and where it stops

Most of the AI you will come across is already inside the apps that run your finances. Xero suggests which account a bank line belongs to. Receipt capture tools read a supplier invoice and pull out the amount, the GST and the date. We treat all of it as a suggestion. When Xero guesses that a $480 Bunnings charge is repairs and maintenance, sometimes it is, and sometimes it is a new tool that belongs on the balance sheet. Only someone who knows your business can tell, so someone makes that call

Then there is the general AI we use ourselves, which is Claude, made by Anthropic. It drafts emails, weekly updates, the commentary in your monthly reports, and summaries of long threads. It looks for trends in your numbers so your accountant can spend the time on what they mean. We also connect it to our own client and project management systems so it can help us prepare for your work and keep track of what is outstanding. And we use it to write our internal procedures and build our own tools

What AI doesn't do - lodge. Nothing goes to the ATO without peer review and a team leader's sign off, and both of those are people. AI doesn't give you advice. It doesn't decide anything about your pay runs, your employees' entitlements or your obligations on its own

AI has limits, and we would rather say so. It can be confidently wrong. It can be months out of date. It can miss the one detail that changes the answer. So everything it produces is a draft until a person checks it

Your information

Two kinds of AI, two levels of risk

The AI built into Xero, your payroll app and the rest works on data those apps already hold - transactions, payroll, employee details, bank accounts, tax file numbers. It is protected by each provider's own security, and every provider we use has to meet our Supplier Code of Conduct

General AI is where most of the risk sits, so we are stricter there -

  • One approved platform, Claude, on paid business plans we manage centrally. No personal accounts and no free tools for client work
  • Your information is never used to train AI models. That is a condition of the plans we use
  • Tax file numbers, bank account and BSB details, passwords and logins do not go into general AI. Other personal information is limited to what the task needs, and de-identified where we can
  • Technical guardrails back up the written policy. They limit what personal information can be sent to AI and help stop client data being extracted from our systems

Your information is stored on systems hosted in Australia. When we use Claude, what we send it is processed by Anthropic in the United States under contractual confidentiality terms. We prefer to process in Australia and do so wherever that option exists. Our Privacy Policy lists every provider and where each one processes data

Some of our team are based in the Philippines. They work inside our Australian systems, under the same AI rules, the same confidentiality obligations and the same Australian supervision as everyone else. Your data is not copied there

Why there is no opt out

We don't offer an AI free version of our service. The AI is inside the accounting software itself, so there is no switch to flip. It is also how we do quality work at a fee that makes sense. Running two ways of working would make us slower and less consistent, not safer

What you get instead is everything on this page, and a conversation whenever you want one. If something here worries you, tell us. We would rather explain it than have you wondering

How we pick the tools

Four questions before anything is approved -

  1. Where does the data go, and who can see it?
  2. What does it do when it doesn't know the answer?
  3. What would the vendor do under pressure, commercial or otherwise?
  4. Does it fit our values as a certified B Corp?

Approved tools sit under a written AI policy the whole team is trained on. AI assisted work goes through the same review, supervision and record keeping as everything else we do (see Professional Standards). We review the tools, the policy and this page at least every six months, and sooner if something changes

We follow the Tax Practitioners Board's guidance on AI and the Code of Professional Conduct (TPB(GS) 55/2026), the Australian Privacy Principles, and the National AI Centre's Guidance for AI Adoption

Questions?

Email ai@digit.business or call 1300 344 480. Privacy complaints follow the process in our Privacy Policy. Complaints about our work go through Complaints and Feedback

Want the longer story? Read The AI decision nobody prepared you for and how we built our data guardrail

Last updated - September 2026