The Evolution of Tax Research in Australia
For decades, the foundation of tax research was volumes of legislation and commentary founds on bookcases. Then came the digital age and search engines like Google became the go-to first step. Today, we stand on the cusp of another evolution: artificial intelligence.
The rise of AI promises to transform how we navigate the complexities of tax law, but as with any new technology, not all tools are created equal. Understanding the difference between general and specialised AI tax research is now crucial for maintaining efficiency and accuracy in a modern practice.
This article explores the evolution of the research process by comparing three distinct methods:
- The traditional Google search;
- The use of general-purpose AI like ChatGPT; and
- The application of a specialised AI designed for tax professionals.
The Traditional Method: Navigating the Maze of Google Search
For many years, Google was the default starting point for queries. When faced with a technical question, such as the tax deductibility of software development costs, an accountant’s first instinct was often to type a few keywords into the search bar.
This approach can be fraught with inefficiency. For example, a search for "software development costs tax deduction Australia" can yield a mixed bag of results. You might find an ATO page, but it could be about a tangentially related topic like in-house software for internal business use, which isn’t relevant if the software is developed for sale to customers. You may also see links to blogs from unknown authors, marketing content from software companies and guidance from overseas jurisdictions.
The core problem is the time spent filtering signal from noise. An accountant must critically evaluate every source, cross-reference information and hunt for the specific ATO ruling or legislative clause that provides a definitive answer. This process can take hours of non-billable time, turning a simple query into a frustrating and prolonged investigation.
A Step Forward with Pitfalls: General-Purpose AI (Like ChatGPT)
The arrival of large language models like ChatGPT marked a significant leap forward. Instead of just providing links, these tools can synthesise information and provide a direct, conversational answer.
Posing the same question about software development costs to ChatGPT may yields a more coherent response than a Google search. It might correctly identify that the expenditure is capital in nature and could be treated as a depreciating asset.
Despite this advancement, general-purpose AI tools can have serious drawbacks for technical tax research. Their knowledge is based on the vast, unfiltered Internet, which can lead to three critical problems:
- Fabricated Sources: In their effort to be helpful, these AIs can "hallucinate" and invent sources. They might cite a tax ruling with a plausible-sounding reference number that simply does not exist. In other cases, they might provide links to real rulings that are completely off-topic, such as one concerning trust income or a breach of an employment contract. This sends the researcher on a wild goose chase and completely undermines the credibility of the answer.
- Lack of Substantiation: The AI might provide a conclusion without any links to its sources. It might state that software copyright is deductible over 25 years, but offer no direct path to the ATO guidance or legislation that confirms this, leaving the accountant to manually search for substantiation.
- A Non-Specialised Interface: These platforms are designed for general use. They do not present information in a way that aligns with an accountant's workflow. The lack of a side-by-side view of the answer and its source documents can make verification a clunky, multi-step process.
While faster than Google, a general-purpose AI can introduce a dangerous level of unreliability into the research process.
The New Frontier: Why Specialised AI is the Future of Tax Research
Recognising the limitations of general tools, a new category of specialised AI has emerged. These platforms, such as SavvyWise, are built from the ground up for a specific professional domain.
In the SavvyWise vs ChatGPT comparison, the difference is clear: a specialised tool is not trained on the entire internet, but on a curated and verified knowledge base of Australian tax legislation, ATO rulings and other authoritative sources. This focused approach solves the core problems of its generalist counterparts.
For example, when asked about software development costs, a specialised tax AI can provide a precise and, most importantly, a substantiated answer. It can correctly identify the expenditure as capital, points towards its treatment as an intangible depreciating asset (copyright) and provide direct links to relevant private binding rulings and tax rulings that support this position.
The benefits are transformative:
- Accuracy and Reliability: By drawing only from authoritative sources, the risk of hallucination or citing low-quality information is virtually eliminated. The answers are grounded in Australian law.
- Efficiency: The interface is designed for accountants. Answers are presented clearly, with sources linked in a user friendly side bar.
A Practical Comparison: Putting AI Tax Research to the Test
Watch this video to see specialised AI-assisted tax research in practice as compared to Google and ChatGPT.
Conclusion: Embracing Specialised Tools for a Modern Practice
While general-purpose AI is an impressive technology, its inherent unreliability and lack of specialisation make it a risky proposition for professional tax research. The future lies in purpose-built AI that acts as an expert tax assistant with verifiable, high-quality source information. By leveraging specialised AI tax research tools such as SavvyWise, accounting firms can enhance their efficiency and accuracy.