Artificial Intelligence in Insurance Auditing

Consistent integration of Artificial intelligence enhances insurance auditing and redefines it as a strategic partner, providing management and the board with richer insights than previously possible. The question is not whether AI will change the industry – this is already happening. The question is whether your audit firm will shape this transformation or merely react to it.

AI is transforming fundamentally how financial statement audits are conducted in the insurance industry. Full population analysis, instead of sampling, is becoming feasible; anomaly detection is increasingly automated; and analytical tasks that once demanded significant manual effort can now be executed in real time.

This goes beyond a simple efficiency narrative. AI expands the scope of audit coverage – and thereby enhances the quality of the audit assurance provided. Firms that deliberately build AI capabilities and combine them with rigorous governance will achieve enhanced audit outcomes for their clients whilst simultaneously freeing up time for more valuable professional judgement.

The following three use cases are currently employed in our audit insurance practice.

1. Automated voucher testing
   AI models extract relevant data points from supporting documents and organise them into a structured format for seamless, audit-proof integration with financial accounting. This automated pre-processing significantly reduces the need for time- and resource-intensive manual data extraction, which is typically limited to spot checks and involves substantial documentation effort that restricts comprehensive analysis of large data volumes. By enabling automated full audits, these models enable a much deeper level of inspection than sample-based methods, uncovering anomalies and subtle details that manual processing might fail to detect. At the same time, human review remains essential to assess the findings from AI and focus on the professional evaluation of complex anomalies, thereby enhancing overall audit quality.
   
   
2. IBNR and provision testing
   By automating loss triangle generation, detailed IBNR (Incurred But Not Reported) calculations at the highest level of granularity, and extensive sensitivity and driver analysis across all claims data, full recalculation becomes feasible and achieves audit standards with enhanced depth. This approach enables objective, data-driven validation of complex provision valuations, resulting in greater audit thoroughness and improved quality compared to traditional, manual portfolio reviews that are time-consuming and often lack comprehensive data verification.
   
   
3. Regulatory compliance testing
   By leveraging AI to extract relevant regulatory texts and automatically verify internal policies consistently against FINMA and accounting requirements, a complete and seamless audit trail can be established. Automation significantly reduces the manual effort involved in reconciliations, mitigating the risk of compliance gaps. As a result, auditors are able to concentrate on higher-risk matters, enabling more in-depth audit execution and delivering greater value to the client.

Governance: The Indispensable Foundation

At Deloitte, our responsible AI implementation in audit engagements follows a structured, staged approach – beginning with proof-of-concept tests with established success criteria, advancing through pilot phases with expert validation, and leading to full deployment. Sensitive client data is processed exclusively in approved, secure environments, with data stored in Switzerland, as required by relevant data protection and regulatory frameworks.Each AI-assisted audit step includes defined human-in-the-loop control points, ensuring that automated interim results are subject to human review and validation before being incorporated into subsequent stages.

AI models must be sufficiently explainable, ensuring the outputs are fully anchored in audit documentation and subject to review by competent professionals. Our Audit Quality and Risk Management Team is involved throughout each release and validation process. All must understand this fundamental truth: AI serves as a tool, human judgement and accountability remain paramount.

AI and Human Expertise: A Powerful Partnership

The true power of AI in audit emerges not from automation alone, but from the synergy between intelligent systems and human expertise. AI excels at identifying patterns, processing complete datasets, and flagging potential anomalies with consistency and speed that humans cannot match. However, it is the auditor's professional judgement that determines whether these flagged items represent genuine risks or benign variations. This collaborative model, where AI handles the heavy lifting of data analysis and auditors focus on interpretation and evaluation, creates a more robust audit than either could achieve independently. By combining machine precision with human insight, we elevate audit quality to new levels.

Coming soon: Beyond the three use cases outlined in this article, several important themes warrant ongoing attention as AI becomes more embedded in audit practice. How do we maintain audit quality and consistency as AI systems scale across different client engagements and jurisdictions? What governance structures ensure that AI models remain fit-for-purpose as business environments and regulatory requirements evolve? How can audit firms balance the efficiency gains from automation with the need for robust human oversight and professional scepticism? These questions – around scalability, governance, and the human-AI interface – represent the frontier of responsible AI adoption in audit and will continue to influence how firms design and deploy AI-enabled audit solutions.
 

Authors

• Denise Wipf, Partner, Audit & Assurance
• Dr. Madan Sathe, Partner, Forensic & Financial Crime
• Sandy Goebler, Director, Audit & Assurance
• Gabriel Maerchy, Manager, Audit & Assurance