La IA entusiasma a las firmas de auditoría, quizás demasiado

The rapid adoption of AI technology by major auditing firms, such as KPMG and EY, has sparked enthusiasm but may go too far. While accounting is not the most glamorous profession, it has recently gained some excitement. Catherine Burnet, KPMG's audit director in the UK, explained to me last week how the technology is transforming auditors' roles, making the work much more interesting than in the past.

This enthusiasm is due to the fast adoption of AI by the Big Four firms, both globally and in the UK. AI agents filter transactions to detect fraud and errors, review junior auditors' work, and generally become more useful. The main transformation lies in AI's ability to analyze every one of the millions of transactions companies perform annually, rather than auditors only being able to examine a small number of them in detail.

Matthew Campbell, KPMG's UK audit technology director, compared it to analyzing the entire flow of a river instead of using a bucket to extract a sample. Auditing firms that face the risk of missing problems in large amounts of data are excited about being able to deploy an army of AI agents alongside humans. EY employs 85,000 auditors worldwide but conducts nearly double that number of company audits each year.

AI could prevent auditors from being overwhelmed. They could use some help. British firms have been under pressure to improve audit standards following a series of failures, including Carillion's collapse in 2018 (KPMG was fined £21 million in 2023 for its failures at Carillion). They have made progress with the help of technology, with both KPMG and EY now using their own cloud platforms for more comprehensive audits.

The sector is not the only one excited about AI; it is being used in all professional services, and there are hints that it is driving growth in productivity in the UK economy. Corporate clients use AI to organize and control their finances, and accountants don't want to be left behind. However, audit firms should not be carried away by the enthusiasm.

For one, AI does not change anything in terms of regulatory responsibility. An audit firm must not only support its audit opinions but also be able to explain the reasoning and justify the measures taken for each transaction examined or asset value verified. It cannot refer to a black box to the regulatory body. In a way, that is the goal of AI: to filter all operations and identify the cases of higher risk for a more thorough review; the most delicate work remains in the hands of humans.

But the AI model conducts an initial classification, not a comprehensive audit, and how can the firm be sure of its reasoning's soundness? It might be generating a series of false negatives by not detecting problems. This reveals an even greater difficulty: the human tendency to place too much trust in technology. The excessive dependence on value-at-risk models by banks, which underestimated the possibility of significant losses in a real estate crisis, contributed to the global financial crisis of 2008-2009.

AI-based audit models could function without problems for a long time and suddenly fall for an unheard type of transaction. Confidence in technology could also lead to a loss of human skills: junior auditors have always learned basic tasks that can now be assigned to agents. Richard Harrison, EY's UK digital audit leader, states that their principle is that AI is there to support us, not to replace us.

They strive to ensure that their AI agents guide auditors, rather than dictating answers. Both firms and the Financial Reporting Council (FRC), the UK's financial regulator, state that they want AI to help improve audit quality rather than focus on cost reduction by eliminating junior positions. However, the temptation is there: KPMG pressured its own auditor, Grant Thornton in the UK, to pass on savings from AI adoption to clients by reducing audit fees.

Another way to get a return on the AI investment is to convince clients to pay more for their results. Burnet says KPMG's technology improves audit quality and can provide companies with information about their data that they are unaware of themselves. This is fine as part of an audit, but British firms have regulatory limitations on non-audit services: they cannot simply turn AI into a new consulting line.

It is too early to know how all this will evolve, as audit firms are constantly creating new AI agents. The FRC has warned about the risks of hastily relying on untested technology, but the sector is firmly focused on AI. It is exciting times in a cautious sector; let's hope the enthusiasm doesn't get away from us.