Only 4% of insurers fully trust AI agents despite growing interest

42% of insurers have yet to develop a formal strategy for deploying AI agents.

Only one in 10 of insurance organisations have partially or fully implemented AI agents, whilst another 20% are running pilot projects, according to a new report from Capgemini Research Institute

Meanwhile, 42% of insurers have yet to develop a formal strategy for deploying AI agents in their operations.

Capgemini also highlighted a growing interest in Agentic AI within the insurance sector, which forms part of a broader $450b opportunity. 

However, the study reveals that most insurers are still in the early stages of adoption, and many lack a clear implementation roadmap.

Sales (52%) and customer service and support (57%) are the top areas where insurers expect AI agents to manage at least one process or sub-process daily. 

Despite these expectations, trust in AI agents remains low. 

Only 4% of insurance organisations fully trust the technology. The overall trust level has declined from an average of 54% in 2024 to 47% in 2025.

Privacy (50%), bias (50%), and transparency (51%) are the top risks that concern insurance firms when it comes to AI agents. 

Looking ahead, 26% of insurers believe AI agents will augment human workers within one to three years, and 35% expect these agents to operate autonomously within human-supervised teams.

Programming and software development (67%) tops the list of required technical skills, while decision-making (60%) is the most critical soft skill. 

However, only 8% of insurance organisations have fully embedded ethical AI principles into their operations, well below the global average of 14%.