AI can help tax professionals work faster, but the thinking, accountability and training of the people in our profession must stay firmly human. In this article, William Young FCPA GAICD, founder of Praxio AI, explains why 2026 should be the year tax firms move from AI experimentation to accountable workflows.
This article is sponsored content in partnership with Praxio AI, an Australian company building practical, trusted AI for public practice accountants and tax professionals.
A year ago, tax professionals were asking, “Can AI help my firm?” Now the question has shifted to: “How do we implement AI safely, consistently and in a way that improves the quality of our work?”
Over the past year, I’ve presented to many accountants and tax professionals. Through building Praxio AI – Tax Assistant, I’ve also spoken with practitioners about where AI helps, where it creates risk and what firms need to do differently in 2026. I’ve seen that shift happen in real time.
AI is moving from simply answering questions to doing the work. AI Agents can now take a client scenario, research it and return a polished draft. Clients are also arriving with AI-generated answers they want their adviser to confirm.
Three key beliefs inform my approach to AI and how the tax community can interact with it.
1. Humans must control the critical thinking
AI is very good at producing a first draft. But a draft is not advice.
In tax practice, the practitioner still frames the question, tests the facts, identifies missing information, weighs competing interpretations and decides what is appropriate for the client. A client can generate an answer using AI. A tax professional knows how to question it.
The real pain point is often not that the practitioner has no idea of the answer. Many know the concept, or likely direction. The time is spent finding the provision, ruling or ATO guidance that supports the position and can be kept on file. AI can help here, but only when its citations are accurate and its source links can be checked. An unverified reference is a risk, not a shortcut.
2. We cannot outsource accountability, duty of care or review
A polished AI answer is not the same as a supportable tax answer.
Recent examples of AI going wrong point to the same lesson. In some cases, AI-generated content has included false legal citations, misleading outputs or unsupported analysis that was not properly checked before being used in professional work. The issue is usually not that AI was used. The issue is that it was used without enough verification, review or governance.
The recent TPB Guidance Statement 55/2026 on AI reinforces that practitioners remain responsible for the services they provide. AI output should be assessed, verified and supplemented by professional judgement before it is relied on.
The practical response does not need to be a 50-page AI policy. Start with a simple AI register. Define approved and prohibited use cases. Set rules for client data and redaction. Then set review rules based on risk. Client advice and complex tax positions should require source checking and senior review. Review and improve the policy.
Before using AI, ask:
- What is the task?
- What is the risk if the output is wrong?
- What information am I entering?
- What sources support the answer?
- Who reviews it?
- What evidence is kept on file?
The aim is to move from “AI gave me an answer” to “I know how the answer was checked, reviewed and supported.”
3. We must invest in people, especially the next generation
AI should accelerate learning, not replace it. If junior staff use AI to skip the reasoning process, firms may create a long-term capability problem.
Use AI as a supervised assistant. Ask juniors to form a view before they prompt. Review the AI’s sources together. Discuss missing facts, assumptions and alternative views.
The tools matter here too. Features such as clarifying questions can help users think through the facts, assumptions and missing information, which is also valuable for staff training. Whatever the tool, the practitioner should still be able to see what the answer rests on, with accurate citations and source links that can be checked.
Training tax professionals has always involved mentoring, review and judgement. That responsibility remains ours.
Putting this into practice
General AI tools are powerful, but technical tax work needs something more disciplined: credible sources, tax-specific workflows, clarifying questions, citations that can be checked, and outputs that support practitioner review.
Praxio AI – Tax Assistant was built around those needs, following the workflow of a tax professional: identify the issue, research credible Australian tax sources, analyse the position, consider alternatives and risks, provide a recommendation, draft clearly and keep a review trail. Citations and source links let the practitioner check the answer, the review trail supports accountability, and clarifying questions prompt users to think through the facts. A second-pass review step challenges the initial analysis and checks for gaps and alternative interpretations before the draft reaches the practitioner. These AI review steps support, rather than replace, professional review and judgement.
For firms implementing AI in 2026, my suggestion is simple: start small, set clear rules, protect client information, review every output and improve as you go.
Starting small can be as easy as taking one real research question and running it through a tool like this. Praxio AI – Tax Assistant offers a free trial, so you can see how the sources are cited, check them yourself and judge whether the output would stand up to your firm’s review process.
Start a free trial of Praxio AI - Tax Assistant: https://app.praxio-ai.com.au/signup
This article is sponsored content in partnership with Praxio AI. The views expressed are those of the author, and the mention of Praxio AI - Tax Assistant or any other product does not constitute an endorsement by The Tax Institute. Readers should make their own enquiries and seek professional advice before relying on any product or service.