The role of AI as a powerful ally to ensure compliance and mitigate financial crime risks in insurance

Generative and agentic AI can materially reduce the burden of compliance in insurance while improving quality, consistency and auditability, provided the right data, process and governance foundations are in place. With the evolution of generative AI, expectations are high that compliance activities in insurance will become more efficient and more effective in quality and consistency. Although many machine learning applications are still being evaluated, deep learning is the next step forward in AI curve, and generative AI embedded in agentic frameworks is emerging as a game changer.

Agentic AI is the next evolutionary step, combining Robotic Process Automation (RPA) style orchestration with AI driven decision support for the human between process steps. This means processes are not only automated (as much as possible) but also guided by AI that evaluates context between steps and informs the human to make an informed decision. In simple terms, agentic AI coordinates tasks and provides recommendations using AI/GenAI throughout the workflow. Early movers report 30–60% faster cycle times, 20–40% lower compliance effort, and improved consistency and auditability, though capturing value requires changes across data, IT, processes and people/culture.

Our research into the Swiss financial services industry suggests strong potential for efficiency gains and cost reduction in regulatory compliance, in monitoring transactions for suspicious activity, and in other anti‑financial‑crime work led by Money Laundering Reporting Officers (MLRO). In Switzerland, insurers are upgrading their financial‑crime controls to stay cost‑effective and to use recent advances in AI. The latter is in fact increasingly being leveraged in vendor selection, benchmarking and calibration as ways to improve monitoring of life insurance Anti-Money Laundering (AML).

The insurance industry has made clear strides forward: Know Your Customer (KYC), ongoing transaction monitoring and regular internal training are now standard across many carriers. Technical screening for suspicious payments to beneficiaries and sanctions breaches are increasingly embedded into solutions, and most insurers operate a risk‑based approach with enhanced monitoring for higher‑risk customers, geographies and distribution channels. Yet a gap often persists between regulatory requirements and practical reality. Traditional controls focus on straightforward cash flows, while complex product features and cross‑border arrangements are harder to capture. Loopholes can occur due to indirect distribution through brokers, opaque beneficiary arrangements and misaligned systems between countries or between parent companies and subsidiaries – making oversight more difficult when policies are sold and paid out abroad.

Here we highlight five use cases with significant potential to improve compliance efficiency:

1. Compliance training: Generative AI enables interactive, podcast‑style learning solutions, applicable both to professionals working in compliance functions and to firm-wide trainings. Fed with current policies and guidance, these sessions allow to improve engagement, offer life Question and Answer options and up-to-date messaging functionalities.
2. Control testing: The Retrieval‑Augmented Generation (RAG) AI method maps regulatory rules to a firm’s internal policies, procedures and controls, identifies gaps (e.g., a rule with no matching control or a partial match), and proposes fixes, speeding design effectiveness reviews and preserving a clear audit trail from rule to control to test.
3. Negative news screening: Beyond keyword searches, generative AI retrieves, clusters, de‑duplicates and summarises relevant adverse media, tags risk themes and assembles evidence packs, providing an effective starting point to meet screening requirements.
4. Anti Money Laundering (AML) transaction monitoring: Machine Learning (ML) reduces false positives while maintaining an audit trail and explainability for alert scoring. Generative AI also standardises case summaries, improving consistency and quality in high‑volume false‑positive documentation and speeding up first‑line reviews.
5. AI governance and AI risk management: To operate safely, maintain an end‑to‑end framework covering model inventory and classification, approvals, validation and monitoring, explainability, human‑in‑the‑loop controls and AI literacy, while aligning to regulatory expectations (e.g., EU AI Act) and model risk standards.

AI offers insurers a credible path to more effective and efficient compliance, particularly where traditional controls struggle with single‑premium and surrenderable life policies, investment‑linked products and cross‑border or broker‑led distribution. Realising the potential requires investment in data and IT, process redesign and cultural change, plus strong AI governance with clear human oversight. With leadership commitment and disciplined execution, insurers can enhance transparency over the origin and use of funds, materially reduce false positives and cycle times, and strengthen resilience against money laundering, fraud and sanctions risks – delivering better outcomes for customers, regulators and themselves.

Authors

• Dr. Marcel Thom, Partner, Insurance Lead, Switzerland
• Dr. Madan Sathe, Partner, Forensic & Financial Crime
• Rebecca Roj, Senior Consultant, Swiss Technology & Transformation