What does the future look like for auditors?

We know computer programs based on AI can already handle financial accounting and prepare tax returns that are traditionally handled by auditing firms. On paper, it looks like the accountant's role could seriously be at risk in the next five or ten years.  

However, perhaps it's less about dualism between man and machine and more about interdependence. For example, at Mazars, we are already using AI because we believe the technology has great potential. For us, innovation in audit is about leveraging AI's opportunities and automating as many routine audit functions as possible. This frees up the accountant's time for the most important thing: working with clients face to face.  

AI will always reach certain limits, and that's where human expertise is needed. For example, there will be situations that the machine is unfamiliar with and, therefore, cannot evaluate and classify – this is precisely where professionally trained auditors will be needed to intervene. AI's strength is drawing conclusions from historical data and events and applying them to the here and now. However, sometimes we have to make decisions on unprecedented events which AI can't compute because it lacks the historical statistical basis and the relevant data.  

How a lack of historical data can impact AI  

If a client has a legal dispute, an accountant's role is to assess whether the client has correctly made the appropriate provision. Let's say the legal dispute involves electric cars. Suppose the battery in every third car from an electric vehicle manufacturer explodes. Electric cars haven't been around long enough for AI to have learned what a repair costs, the likelihood of potential damage to the environment or how high the risk of injury is to people. In this case, a responsible person is needed to make an estimate based on commercial expertise.  

The auditor's task is to understand this estimate and perhaps challenge it to find out which factors have been taken into account, what hasn't been considered, or what may not have been considered enough so we can support the client. In this example, historical data isn't available for AI to verify the estimate. Equally, how is data based on legal proceedings in, say, the United States relevant in another country? AI is only as good as the data it has access to. 

Is there a case for AI offering a better outcome? 

Indeed, the question of proportionality plays a role in auditing. However, when it comes to paying fees, having access to people with expertise who can discuss a crucial issue from different angles and then come to a common conclusion gives clients more confidence. In contrast, AI may give you an improperly trained model or incorrect data, producing values that have no relation to reality. It's the intellectual achievement that is decisive here. 

In terms of outcomes, it depends on the product. Let's say we ask AI to design a thermos mug. We can describe its specifications in detail because millions of comparable products have been sold, which means AI has a lot of data to work with. There is, therefore, a high probability that AI will create an optimal product that a human designer can't do as well. However, when dealing with unknown facts faced in auditing, we need experts with the capacity for abstraction. In this scenario, a client would place more trust in auditors with well-founded opinions that AI can't offer. It's all about trust. 

What are AI's strengths in auditing, and where is it best employed? 

Summarising texts and uploading scanned documents to see that a receipt, for example, meets the requirements of VAT law is something that AI does very successfully. So, the impact of AI on balance sheet preparation and routine accounting tasks is already making a positive difference. This will only improve. However, what AI can't do is check that a document has actually been sent. Even at this routine level, there needs to be an element of human interaction and oversight.  

Human involvement becomes even more necessary when complex financial arrangements in a network of companies are involved. The ability to support and challenge clients when required is a vital part of the auditor's role that AI can't replicate. As auditors look to the future, harnessing AI's capabilities to help streamline routine tasks will be essential as the regulatory and compliance landscape becomes more complex.