Reclaiming insurance margin: From silent drift to structural control in the AI era

By Abhishek Dwivedi, Senior Vice President and Global Head of Insurance, Xebia

GenAI is no longer experimental in insurance. From quote generation to claims triage and document intelligence, enterprise-grade pilots have become embedded workflows. However, the business impact remains inconsistent. Most insurers have crossed the automation threshold without capturing proportional margin gains. This isn’t a failure of tools. It’s a failure of orchestration design. Systems have evolved to handle volume and variance but not fiscal consequences.

Where margin actually slips

Margin erosion is rarely driven by weak underwriting or poor pricing anymore. It begins with unowned overrides, ambiguous escalations, and the absence of traceability across decision points. In many cases, AI systems operate with high throughput but zero economic feedback. A GenAI assistant may generate 200 quotes an hour, but if 30 percent are later overridden for priority advisors with no audit linkage, the system accelerates ...