Global – October 15, 2026 -- Only 19% of insurance executives believe their operating model is agile enough to support digital transformation, even though 71% say digitalization is already affecting it, according to CGI's 2026 Voice of Our Clients research.
Agentic AI moves insurers from managing interactions to orchestrating outcomes
Conversational AI can explain what a policyholder should do after storm damage. An agentic system could go further: authenticate the customer, check coverage, open the claim, collect information, arrange services, initiate permitted payments and send proactive updates. A human claims professional steps in when judgment or extra support is required.
Insurers could also use agents to identify coverage gaps, prevent policy lapses and anticipate risks before customers are affected. CGI's research ranks AI-enabled customer experience as a top priority, but says delivery depends on core modernization, automation and strong data foundations.
Delegated-authority limits should govern AI agents as they govern underwriters
Underwriters have limits, claims professionals hold settlement authorities, and payments above defined thresholds need additional approval. CGI argues each AI agent needs the same boundaries: what it can access, recommend, change, approve or initiate, and when it must stop and involve a human.
Regulators in Europe, the U.S. and Canada are raising accountability expectations
Key EU AI Act requirements, including transparency obligations, became applicable in August 2026. A growing number of U.S. states have adopted the NAIC Model Bulletin on AI use by insurers, and Canada's OSFI is strengthening expectations on AI and model risk.
CGI advises starting with simple, bounded use cases
No single provider will supply every capability an agentic model needs, which may span cloud infrastructure, foundation models, customer service platforms, core insurance systems, identity services, data platforms and cybersecurity controls. CGI has delivered client initiatives with platforms such as NiCE Cognigy, but says technology selection should follow the target operating model and desired business outcome.
Insurers should avoid starting with the most complex claims or the most autonomous agents. CGI recommends building reusable foundations for trusted delegation first, then expanding autonomy.