Global insurance company AXA has embedded artificial intelligence across its core insurance operations in a change that will have a major impact on software testing, quality assurance and digital resilience within the highly regulated insurance sector.
While underwriting and pricing have long relied on historical patterns and actuarial judgment, AXA’s technology leadership outlined a strategy built around predictive, generative, and agentic AI systems.
“AI is shaping the way we engage, interact, learn, work and live, and has a profound impact on how we achieve our mission as AXA,” Matthieu Caillat, Chief Technology & AI Officer at AXA Group, explained in a recent case study.
For banking and insurance QA and software testing teams, this change is much more than experimentation and must be closely monitored. In fact, it took the team at AXA more than four years to test, pilot, and deploy new AI-powered capabilities.
Predictive models and agent AI workflows have brought new demands for validation, accuracy testing, robustness, and compliance assurance, especially as these tools move closer to core decision-making systems.
To celebrate the launch, Caillat convened AXA’s technology, data, and AI teams as well as some of the company’s business teams. “We are on a bold journey to shape the future of AI in insurance,” Cailla declared at a rally in Paris earlier this month.
He added: “The focus now is on accelerating end-to-end transformation, impactful execution, scaling across the group and delivering tangible value.”
Agent AI use cases
AXA reported more than 60 agent AI use cases in testing or partial implementation, spanning underwriting, contact centers, and claims processing.
For test teams, deployments of this scale mean continuous assessment, structured QA oversight, and control over how AI-driven automation is performed on live financial workflows.
Dr. Andreas Scherzinger, AXA Group Chief Data, AI and Innovation Officer, said the insurer’s ambition is to move beyond automation to “augmented intelligence”, where human expertise teams generative and predictive models.
He told the IT-focused website Computing that AXA “designs and deploys AI solutions across the value chain,” stressing that the goal is to support and enhance human judgment, not replace it, and that its goal has a direct impact on how models are tested, validated, and managed.
A central focus for the test team was tools that support dynamic extraction, summarization, and presentation of guidance materials.
AXA highlighted that it uses Search Augmented Generation (RAG) to help insurers sift through large reference documents.
The pilot reduced average investigation time “from 10 minutes to less than 3 minutes per query,” with 86% of users rating the tool 8 out of 10 and expressing confidence that they had considered all relevant guidance.
“We are on a bold journey to shape the future of AI in insurance.”
– Matthew Kaiya
To QA specialists, that performance metric resembled end-user validation data more than headline claims.
The metrics of speed improvement and end-user reliability are similar to key validation data for QA teams, but they also reveal important testing requirements. This means ensuring that summaries are accurate, outputs are robust under a variety of conditions, and model decisions comply with regulatory standards.
In a computing case study, Schertzinger also pointed to AXA’s approach to building a common AI platform that is “open, secure, and compliant,” emphasizing that governance and secure integration are core to the company’s strategy.
For QA and testing professionals, this ecosystem has highlighted the need for rigorous integrated testing where AI components interact with established underwriting engines, policy systems, and regulatory controls.
broader strategy
AXA leaders framed this rollout as part of a broader platform strategy. The insurer said it was building one common foundation to “power agent AI across AXA,” designed around “open market standards” and aimed at maintaining “security and compliance.”
For the software testing team, that direction brought with it familiar corporate demands. Integration with underwriting and claims environments still required functional and integration testing, especially when AI tools interact with traditional policy systems and regulated decision-making chains.
Document summarization and workflow automation also required quality control to avoid misclassification, hallucinatory output, or deviations from regulatory guidelines. These risks have become more acute as AI systems replace manual review procedures.
AXA’s approach emphasized that the introduction of AI in insurance is not just a matter of innovation. This became a testing and resiliency challenge, with QA teams playing a central role in validating fairness, accuracy, and compliance while ensuring that AI systems operated reliably under real supervisory oversight.
Interested in becoming a QA Financial subscriber?
It’s completely free
* Receive our weekly newsletter every Wednesday * Receive priority invitations to forum events *
Register here now
Regulation and compliance
Looking for news about the regulations and compliance requirements that drive the development of software quality engineering in financial companies? Visit our dedicated site Click here for the Regulations and Compliance page.
read more
Watch now
QA Financial Podcast
