When KPMG deployed AI agents to review the financial statements of a FTSE 250 company, the expectation was that the technology would assist with routine checks. Instead, something unexpected happened: the AI asked questions that were, by many accounts, more incisive than those posed by the human accountants. This pilot, part of a broader push by Big Four firms to integrate artificial intelligence into audit processes, offers a glimpse into a future where machines not only crunch numbers but also challenge assumptions.

The Pilot: AI Meets Financial Statements

The pilot involved using AI agents to scrutinize the financial statements of a FTSE 250 company, a mid-cap firm listed on the London Stock Exchange. The goal was to see if AI could enhance the audit by identifying anomalies, inconsistencies, or areas requiring further investigation. According to sources familiar with the pilot, the AI agents were able to generate a series of questions that probed the data more deeply than the human auditors had initially done. These questions ranged from queries about revenue recognition to the valuation of complex financial instruments.

What made the AI's questions stand out was their ability to connect disparate pieces of information and spot patterns that might have been overlooked. For instance, the AI might ask why a particular expense ratio had shifted significantly from one quarter to the next, or why a subsidiary's reported profits didn't align with the parent company's cash flow. Such questions are the bread and butter of audit skepticism, and the AI's performance suggests that it can be a valuable tool in fostering a more rigorous review.

Why AI's Questions Were Better

The AI's advantage likely stems from its ability to process vast amounts of data without fatigue and to apply consistent logic across all items. Unlike humans, who may become desensitized to routine numbers or skip over items that seem fine at first glance, AI can maintain a high level of scrutiny throughout. Moreover, AI can be trained on historical data and known fraud patterns, enabling it to flag areas that are statistically more likely to contain errors or manipulation.

But the pilot also revealed limitations. The AI's questions, while sharp, sometimes lacked context. Human auditors bring judgment and experience to the table, allowing them to prioritize which questions matter most and to interpret responses in light of broader business knowledge. In the pilot, human intervention was needed to refine the AI's output and to ensure that the questions were relevant and actionable. This underscores that AI is not a replacement for human auditors but rather a complement.

Implications for the Audit Profession

The pilot's findings come at a time when the audit industry is under pressure to improve quality and detect fraud more effectively. High-profile corporate collapses have raised questions about whether auditors are doing enough to challenge management. AI could be part of the solution, by providing a more systematic and thorough review of financial data. However, it also raises concerns about over-reliance on technology and the potential for AI to miss nuances that only human judgment can catch.

For KPMG and other Big Four firms, the pilot is a step toward integrating AI into their audit methodologies. The firm has been investing heavily in AI and data analytics, and this pilot is likely part of a broader strategy to enhance audit quality and efficiency. As AI tools become more sophisticated, we can expect them to take on more of the routine questioning, freeing up human auditors to focus on complex judgments and client relationships.

Challenges and Ethical Considerations

While the pilot's results are promising, there are hurdles to overcome. One is explainability: if an AI asks a question, auditors need to understand why it asked that question and what data drove it. This is crucial for audit documentation and for responding to regulators. Another challenge is data quality: AI is only as good as the data it's trained on, and financial data can be messy and inconsistent. Ensuring that AI systems are properly calibrated and validated will be essential.

There are also ethical considerations around accountability. If an AI misses a material misstatement, who is responsible? The firm, the auditor, or the developer of the AI? These questions will need to be addressed as AI becomes more embedded in audit processes. Regulatory bodies are likely to take a keen interest in how AI is used and whether it meets auditing standards.

The Road Ahead

The KPMG pilot is a sign of things to come. As AI continues to advance, it will likely become an integral part of the audit toolkit, helping auditors to ask better questions, identify risks more effectively, and ultimately improve audit quality. But it will not happen overnight. Firms will need to invest in training, infrastructure, and governance to ensure that AI is used responsibly and effectively.

For now, the pilot serves as a proof of concept: AI can indeed ask better questions, but it needs human partners to make sense of the answers. The future of audit is likely to be a collaboration between humans and machines, each bringing their unique strengths to the table. As the technology matures, the profession must adapt, ensuring that auditors are equipped to leverage AI while upholding the principles of integrity, objectivity, and professional skepticism.

Frequently Asked Questions

What exactly did the AI do in the KPMG pilot audit?

The AI agents reviewed the financial statements of a FTSE 250 company and generated questions for further investigation. These questions were designed to probe areas of potential risk or inconsistency, similar to the inquiries human auditors make.

Why were the AI's questions considered better than those of human accountants?

The AI was able to analyze large datasets consistently and without fatigue, spotting patterns and anomalies that humans might miss. Its questions were often more detailed and data-driven.

Does this mean AI will replace human auditors?

No. The pilot required human intervention to refine and contextualize the AI's questions. AI is seen as a tool to augment human judgment, not replace it.

What are the main challenges in using AI for auditing?

Challenges include ensuring explainability of AI-generated questions, maintaining data quality, and addressing accountability and ethical concerns. Regulatory acceptance is also a key factor.

How might this affect the future of the audit profession?

AI is likely to become a standard part of audit processes, enhancing quality and efficiency. Auditors will need to develop new skills to work alongside AI, focusing on higher-level judgment and client interaction.