AI Integration in Insurance: Reshaping Governance and Strategic Efforts

The integration of AI agents with legacy systems is reshaping governance in the insurance industry.

AI agents are increasingly being employed in insurance for tasks such as summarizing submissions, triaging claims correspondence, and managing service inquiries. While these technologies offer transformational potential, they are often being implemented alongside existing core policy administration systems without fundamentally altering them. This approach poses unique challenges in governance and reveals the dichotomy between AI’s theoretical potential and its real-world application. IT departments face a shift in governance roles as they adjust to the enveloping of AI around these unchanged core systems.

Economic and Structural Implications

The promise of AI in insurance rests on strong theoretical foundations, supported by the industry's reliance on defined rules and structured transactions. According to the Boston Consulting Group, modernizing core IT systems is a significant investment, requiring up to 4 percent of non-life gross written premium as capital expenditure. McKinsey's projections further support the productivity enhancements AI could bring, estimating improvements from 20 to 50 percent in areas like discovery and reverse engineering, and up to 90 percent in testing and reconciliation. However, McKinsey emphasizes that modernization is not simply about rewriting code. It involves deciphering undocumented business rules, data conversion, and consistency checks—tasks where AI can assist but not replace human oversight.

Current Adoption and Challenges

Despite optimistic projections, actual implementations reveal a slower pace of adoption. Capgemini's 2026 report indicates that only 10 percent of property and casualty insurers are currently leaders in AI adoption, with the majority still in exploratory stages. The hesitation is grounded in several challenges that AI integration faces.

• Data Truthfulness: AI accuracy is contingent on comprehensive and current system records, which are often lacking.
• Reversibility: Insurance transactions affect multiple areas, and reversing them within outdated systems can be difficult.
• Ownership Issues: System development teams frequently move on, leaving gaps in monitoring and support.

Proactive Strategies and Regulatory Developments

To manage these challenges, insurers are adopting asymmetric access strategies. AI agents can broadly read data but only write through controlled and logged gateways, minimizing transaction duplication and integrating reconciliation into the process design. This strategic approach aligns with maintaining core systems as the definitive record.

Regulatory landscapes are also evolving in response to AI’s growing role. Several states have embraced the NAIC model for AI in insurance, and others are developing proprietary regulations. The NAIC’s trial of an AI risk-evaluation tool highlights the necessity for insurers to maintain clear records of AI interactions within core systems.

The Path Forward

The insurance sector is at a pivotal juncture where the interface between AI agents and core systems must be thoughtfully designed. Insurers that successfully develop these interfaces can delay full system migrations, aligning short-term actions with long-term strategic goals. By preparing their infrastructure for AI’s impact, carriers can adeptly navigate the technological evolution of the industry.