Why does a practice need an AI use policy?
The TPB's guidance statement on AI, issued on 22 July 2026, says tax practitioners remain responsible for the services they provide when using AI and should not rely on AI output as a substitute for their own analysis. It also points to the Determination's requirement to establish, document and enforce a system of quality management. An AI use policy is one of the documented procedures that system can include. Without one, each staff member decides alone what goes into which tool, and the practice cannot show how AI use is controlled.
Which tools should the policy approve?
List each approved tool by name, the account type staff must use and what it may be used for. Separate business accounts set up by the practice from personal or free accounts, because data retention and model training settings can differ between them. Note the AI features built into software the practice already uses, such as the ledger, the practice management system or the email client, since those may process client data without staff opening a separate tool.
- Tool name and account type, with practice business accounts only.
- Approved uses, such as drafting, summarising or formula help.
- Whether client information may be entered.
- Who approved the tool, and when it is next reviewed.
What client information can go into AI tools?
The OAIC recommends, as best practice, that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools. The TPB says registered agents must obtain each client's permission before disclosing client information to a third party, which can include entering it into AI tools depending on how they are configured, and that permission can come through a signed engagement letter, signed consent or other communication. Give staff one simple rule: no client information in any tool that is not approved for it, and approved tools used only within what the engagement letter permits.
How should staff check AI outputs?
The TPB notes that AI models may hallucinate or give inaccurate information, and says practitioners should verify AI-generated content at each step of the workflow and document those steps. Turn that into a rule staff can follow: every AI output that reaches a client, a ledger, a working paper or a lodgement is read and checked by a person with the skills to judge it. Calculations are recomputed. References to legislation or ATO guidance are checked against the source. The reviewer's name goes on the file.
What records should staff keep?
Section 30 of the Determination requires practitioners to keep records that correctly record the tax agent services provided to each client. Where AI was used in that work, the record should show which tool was used, for what, and who reviewed the result. Keep this light: a note on the job or working paper can cover most tasks. For automated workflows, the system's own log can serve as the record if it shows the input, the output and the review.
How do you introduce the policy to staff?
Run a short session that walks through the policy with examples from your own work, using fictional client details. Show an approved tool with its settings, a prompt that would break the client information rule, and an output with an error that a reviewer should catch. Ask staff which tools they already use, since that list is the starting point for approvals. Review the policy when the TPB or OAIC updates its guidance, when someone requests a new tool, and at least once a year.
