Davies has warned that the growing use of autonomous artificial intelligence agents across underwriting, claims management and customer servicing could significantly increase conduct risk for insurers in the UK and Ireland. The warning forms part of the first report in the Davies AI Impact Series, which examines compliance and governance challenges associated with agentic AI systems. (Captive Insurance Times)
Unlike conventional AI models, AI agents can make decisions, adjust their behaviour and operate with a greater degree of independence. Davies said this increased autonomy can create additional challenges around governance, transparency, accountability and customer protection. (Captive Insurance Times)
The report stresses that insurers do not face a regulatory vacuum when deploying AI. Existing frameworks, including the UK Financial Conduct Authority’s Consumer Duty, UK and EU GDPR requirements and the Central Bank of Ireland’s Consumer Protection Code 2025, continue to apply to AI-driven decisions. However, insurers are increasingly expected to demonstrate stronger controls and oversight over how automated decisions are made. (Captive Insurance Times)
Areas identified as particularly exposed include pricing, underwriting, policy eligibility, claims assessment and fraud detection. Autonomous decision loops could potentially lead to unintended outcomes, such as aggressive pricing or systematic claim denials. Continuous-learning systems could also reinforce biases if models repeatedly learn from their own outputs. (Captive Insurance Times)
Davies also highlighted accountability as a major governance issue. Insurers should establish clear senior-level ownership for AI decisions, maintain audit trails recording system inputs and outputs, provide meaningful human oversight and ensure that third-party AI providers are governed to the same standards as internally developed systems. Outsourcing technology, the report notes, does not transfer regulatory accountability. (Captive Insurance Times)
The report also points to the EU AI Act as an emerging benchmark for responsible AI governance, particularly for insurers operating across jurisdictions. Pricing, underwriting and claims-related applications may attract heightened requirements concerning documentation, transparency and human intervention. (Captive Insurance Times)
Another major recommendation is moving away from one-time compliance assessments. Because AI agents can change after deployment through retraining, data drift and evolving operating conditions, insurers need continuous monitoring throughout the technology lifecycle. (Captive Insurance Times)
Davies has proposed a 13-point AI compliance readiness checklist covering areas including accountability, auditability, risk classification, human oversight, bias testing, data governance, vendor risk, monitoring, employee training, incident management, customer redress and operational resilience. The framework is built around five pillars: accountability, transparency, fairness, continuous monitoring and operational resilience. (Captive Insurance Times)
The report concludes that insurers capable of consistently demonstrating fair, transparent and accountable AI-driven outcomes could turn effective AI governance into a competitive advantage rather than treating it solely as a regulatory obligation. (Captive Insurance Times)
