How AXA is using AI to rewrite the insurance model

Because “the best loss is the one that does not happen”

AXA Group senior leadership explain how the insurer is harnessing AI to transform the insurance model, moving from traditional risk pooling to proactive risk management and personalised client services.

AXA Group was founded in an era when risks were largely independent, statistically predictable and therefore straightforward to price. Pooling risk across millions of customers allowed insurers to spread exposure. That world has changed. Climate volatility, geopolitical instability, cyber threats and supply-chain shocks are increasingly systemic and interconnected.

For global insurers, using the past to determine what is likely to happen in the future leaves them at risk. AXA is adapting to this changed environment by placing artificial intelligence at the centre of a sweeping technological and cultural transformation.

At the company’s London office, senior technology leaders set out how AI is reshaping both operations and strategy. Matthieu Caillat, AXA Group Chief Technology & AI Officer and AXA Group Operations CEO, stressed that the ambition extends far beyond incremental efficiency gains. Rather than focusing solely on productivity, AXA is rethinking what it delivers to clients and how it delivers it.

“AI will shape the way we engage, interact, learn, work and live and it strongly impacts the way we deliver on our mission as AXA,” he said.

Inner sourcing

Dr Andreas Schertzinger, AXA Group Chief Data, AI & Innovation Officer, outlined the company’s layered AI strategy across predictive, generative and agentic systems. Predictive AI has long underpinned pricing, underwriting and customer analytics, with machine learning models trained on historical claims and behavioural data.

Generative AI, particularly LLMs, is now unlocking value from the 3.5 billion documents AXA manages globally, summarising content, extracting key information and automating document, image, voice and video processing at scale.

Agentic AI marks a further step, orchestrating workflows, reasoning across complex cases and suggesting next-best actions to employees. More than 60 agentic use cases are in testing or partial deployment at AXA, particularly in contact centres, underwriting and claims - areas characterised by heavy process loads but highly variable customer journeys.

“To prepare for scaling that that we are rolling out one common platform to power agentic AI across AXA,” Schertzinger said. This is based on open market standards, secure and compliant and is supported by leading technology partners. “

To accelerate innovation, AXA has adopted an “inner sourcing” model via a shared AI repository known as Share AI. Teams can publish solutions internally for others to adapt and improve.

It’s a model that has cut the time to market for new solutions from months to days.

Changing the nature of insurance

AI is also being embedded directly into client services. Pierre du Rostu, founder and CEO of AXA Digital Commercial Platform (DCP) explained how AXA Wildfire which launched in 2023, combines AI with geospatial analytics to assess topography, vegetation, wind patterns and human infrastructure in real time.

The system can provide up to 48 hours’ warning of elevated fire risk and offer tailored mitigation advice, reflecting AXA’s broader shift from payer of claims to proactive risk partner. The service extends to disaster management in the event of fires overwhelming defences, such as those in Los Angeles in January 2025.

Still, as du Rostu said, the objective is prevention.

“We guide our clients in practical action so that they can reduce risk. The best loss is the loss that does not happen.”

Agentic and generative AI use cases

Dr Andreas Schertzinger noted that submitting an insurance claim is rarely a linear or standardised process.

“With agentic AI, we can guide the customer through a personalised journey rather than trying to force them to standardise,” he explained. Instead of pushing customers into rigid process templates, AI agents can dynamically adjust questions, documentation requests and next steps based on context.

In underwriting, similar gains have been demonstrated. Traditionally, underwriters must navigate hundreds of frequently updated guidance documents to locate relevant rules and risk criteria. A pilot last year introduced a Retrieval Augmented Generation (RAG) solution that enables underwriters to query data directly. AXA’s Secure GPT then extracts, summarises and presents the most pertinent information.

According to Cali Wood, AXA UK & Ireland Head of Data and AI Strategy and Culture, the RAG pilot reduced the average time required to assimilate information from underwriting guides from 10 minutes to under three minutes per query. Moreover, 86% of participating colleagues rated the tool eight out of ten, reporting increased confidence that they had considered all relevant guidance before reaching a decision.

Sovereignty and sustainable transformation

Despite encouraging results from early deployments, AXA Group is taking a different approach to scaling than some of its competitors. Chubb, for example, recently announced a plan to reduce its workforce by 20% over the next three to four years as it accelerates AI adoption.

What makes AXA’s strategy interesting is its focus on controlled, sustainable AI integration rather than radical headcount reduction.

Matthieu Caillat acknowledged the financial and competitive pressures facing large insurers to automate quickly. However, he argued that speed should not come at the expense of resilience or control. A central concern is sovereignty - ensuring that client data remains protected and that AXA retains architectural independence.

“In terms of regulations on data residency, of course we make sure we comply,” he said. “We use a mix of partners and vendors and try to ensure we build resilience. We are architecting our estate so that we can switch from one to another.”

“It’s not perfect,” Caillat admitted. “We don’t always have a solution that is financially viable, but we are strengthening this approach to remain as sovereign as a company as we can in terms of who we work with. The best partners are changing. We need to be able to take the best of each vendor on the market.

“Locking in with a few partners is easier and faster. But that is not the path we will take.”

Workforce transformation is another area where AXA is striking a more nuanced tone than some competitors. “There will be a transformation of our workforce because that is a consequence of every transformation,” Caillat said. “But for us it is not man against machine. It’s about how AI augments our people.”

Rather than framing AI primarily as a cost-cutting tool, AXA is investing in workforce planning to help employees transition into emerging roles. Strikingly, the company continues to recruit across multiple functions, even as automation increases efficiency.

Caillat emphasised that continued recruitment is viewed positively not as a failure to automate fast enough. “This is about how you get people to the next job and what that job will look like in the future,” Caillat said. The pace of change, he added, demands a more proactive approach to identifying future skills and redeploying talent accordingly.

AXA’s digital leadership team set out a compelling case that retaining institutional expertise is a competitive advantage. “We are ensuring that the expertise that gives us a competitive edge remains in AXA,” Caillat said. By structuring architecture and partnerships carefully, and by using AI to capture and disseminate knowledge that might otherwise remain siloed, the company aims to enhance human judgement rather than replace it.

AXA’s AI strategy is client and culture led. The institution is playing a longer game than the many corporates who are slashing jobs and claiming to have automated them. Successful long-term adaptation to the changed world demands more than simply cutting or offshoring expertise and dressing it up as AI-led agility and dynamism.

AXA is adapting the insurance model itself away from pricing risks based on retrospective data, to dynamically anticipating and working with its clients to help prevent future losses. Crucially, it’s taking its human assets with it.