The Hidden Risk of AI in Tax

Artificial intelligence is changing the way we work, and the tax profession is no exception. Information that once took time to research can now be accessed in seconds, and AI tools can be useful when trying to search for legislation, explore a tax issue or identify potential areas of concern.

However, having access to information and knowing how to apply it correctly are two very different things. This is particularly true in VAT, where the right answer often depends on the facts, the commercial arrangements and the wider circumstances of a transaction.

As AI becomes more widely used, I believe it is worth considering what this means for the future of VAT advice and, in particular, the value of experience.

Recent discussion within the tax profession around the growing “mentorship gap” highlights an important issue. Expertise is not built simply by accessing answers but is developed through experience, challenge and learning from those who have dealt with complex situations before.

The Changing Nature of Tax Expertise

A recent report on the mentorship gap in tax highlights concerns that changing working practices and advances in technology may affect how future tax professionals develop their skills.

Traditionally, much of the learning took place through working onsite alongside experienced colleagues. Junior advisers could see how more senior professionals approached difficult questions, challenged assumptions and dealt with situations where the legislation did not provide an immediately obvious answer.

These skills are difficult to develop from information alone and even more so in the age of remote working or even hybrid working.

For example, determining the correct VAT treatment of a transaction may require much more than identifying the relevant section of legislation. It can involve understanding the contractual arrangements, establishing what is actually being supplied, considering the parties involved, assessing whether the VAT treatment reflects the commercial reality and understanding practical risks.

These are skills that develop over time. They come from dealing with different situations, learning from mistakes and seeing how seemingly similar transactions can produce very different VAT outcomes.

The Practical Implications for taxpayers

For businesses, AI can be a useful starting point when researching VAT questions. It can explain general principles, summarise legislation and help identify issues that may need further consideration.

However, one of the risks with AI is that it will often provide an answer even when it does not have enough information to reach a reliable conclusion. It can present incorrect information with confidence, misinterpret legislation or even invent references and legal provisions that do not exist. This is commonly referred to as hallucination. The difficulty is that the answer may sound entirely plausible, making it difficult to recognise when something is wrong, particularly if the person using it does not have a strong understanding of VAT.

The problem arises when a general explanation is treated as a definitive answer to a specific transaction.

VAT is full of areas where the detail matters. The correct treatment may depend on various factors. Even two transactions that appear almost identical on the surface may require different VAT treatment once the underlying facts are examined.

There is also the question of how tax authorities may view the position if it is subsequently reviewed or does not take practical experience on the issue into consideration because in some cases it won’t have this information. A VAT treatment that appears reasonable in theory may be difficult to defend if the supporting documentation is incomplete or the commercial arrangements do not support the position taken.

This is where relying solely on an AI-generated response can create a real risk. An answer may sound convincing, refer to relevant legislation and appear to address the question directly, but that does not mean it is correct or appropriate for the circumstances. Without someone who understands VAT checking the reasoning, verifying the sources and considering the wider facts, an error can easily go unnoticed until it becomes a problem for the business.

Professional Judgement Makes the Difference

 In my experience, good VAT advice is not simply about knowing the legislation or being able to find the relevant guidance.

It is about asking the right questions before reaching a conclusion and being able to apply it in practical terms.

What is the business actually doing? What do the contracts say? Are the facts complete? Is there another interpretation of the legislation? Has a relevant exception been overlooked? And, importantly, how comfortable are we that the position can be supported if it is challenged?

Sometimes, the most valuable part of the advice is identifying that there is not enough information to give a definitive answer. In other cases, it is recognising that a particular VAT treatment carries a risk that the business needs to understand before proceeding.

That judgement is built through experience. It develops from dealing with real transactions, discussing difficult cases with colleagues, reviewing previous decisions and understanding how VAT principles operate in practice.

It is also what helps an adviser distinguish between an answer that is technically plausible and one that is appropriate for the business concerned.

AI Has a Place, but It Needs to Be Used Properly

I do not see AI as something the tax profession should ignore. Used properly, it can be a valuable support tool. It can assist with research, but professional experience is needed to assess the information and understand its implications.

The key is understanding its limitations and ensuring that the output is checked against reliable sources, current legislation and expertise.

AI should support the VAT team’s work but not replace the thinking that goes into reaching a conclusion.

There is also a longer-term consideration for the profession. If junior advisers increasingly rely on technology to provide answers without being involved in the process of analysing and resolving difficult issues, there is a risk that some of the practical learning traditionally gained through experience could be lost.

Developing the next generation of tax professionals means giving them opportunities to think critically, question assumptions and learn from experienced colleagues. Technology can assist with that process, but it cannot remove the need for it.

Final Thoughts

As AI becomes more accessible, businesses will understandably expect faster answers to their tax questions. But speed should not be confused with certainty, particularly when the consequences of getting a VAT treatment wrong can be significant.

In my opinion, the real value of an experienced VAT consultant lies not just in knowing the rules, but in understanding when those rules apply, what could go wrong and what needs to be practically considered before a business makes a decision.

AI can help find information. Experience helps determine what that information means in the context of a real business.

And when a VAT decision needs to stand up to scrutiny, that distinction matters.