AI in Insurance: What AXA Health & Zurich Got Right (Innovate London 2026)
C · Weak signal A 23-minute panel at a vendor's customer conference with practitioners from AXA Health and Zurich UK; a few concrete practices, but no measured outcomes and the only figures are speaker estimates and relayed analyst claims.
Summary based on the full transcript·Original: about 23 minutes to watch
In short
A panel at Smart Communications' Innovate London 2026 event (uploaded 29 October 2026), with Claire, a communications architecture manager at AXA Health, and David Hobson, platform lead at Zurich UK DevOps. AXA Health describes a catalogue of vetted data packages that marketing teams use without waiting for an insights team, and a central hub that guides use cases through legal and security review. Zurich describes a developer-built agent that consolidates document template specifications and generates test data. Both say data quality and people matter more than the model.
Key numbers
- about 54approved AI data products in AXA Health's internal AI marketplace
- a weekdeveloper analysis time for template specifications that Zurich UK says now takes hours or minutes (speaker estimate)
- maybe 10%Zurich speaker's estimate of the model's share in AI success, against process, data and people
As reported by the source.
English·Video·opens youtube.com
What it reports
- AXA Health runs an internal "AI marketplace" of about 54 data products (customer profiles, claims data, communication and channel preferences), checked for accuracy and approved for use in AI models. Teams now self-serve on who to contact, when, with which message and channel; the speaker reports faster decisions and better campaign results without figures.
- AXA's AI hub vets use cases for a real problem, prioritises them and walks them through legal, architecture, data protection and IT security review. It was piloted and is being scaled.
- AXA call centres use AI call wrap-ups in place of manual notes; agents were nervous at first and later said they could focus on the customer.
- Zurich UK held enablement sessions for developers to find slow manual processes first. One agent reviews legacy template specifications and proposes shared content; analysis that took a developer about a week now takes hours or minutes. The same agent builds a data schema from a business specification and generates XML test cases for template logic variants.
- Zurich's tools were built by existing developers, not a separate AI team; some staff with no coding background built applications with AI help.
- The moderator opens with analyst claims on data quality (a Gartner prediction that 60% of AI projects will be abandoned without AI-ready data); these are relayed, not insurer data.
How they got it
Speaker accounts from a moderated panel, taken from auto-generated captions. No method, baselines or measurements.
Limits
- Hosted by the vendor whose platform both speakers use.
- All outcomes are anecdotes or speaker estimates.
- The AXA speaker's surname is not given; one caption names a different company where AXA is likely meant.
- Auto-captions may contain errors.
Why it matters for companies
Two practices are worth noting for regulated firms. AXA Health's model of pre-approved, AI-readable data packages addresses a common block: business teams waiting on an insights queue, and compliance teams unsure what data a model may see. Zurich's case shows developers aiming agents at their own slowest chores, specification analysis and test-data generation, rather than at customer-facing features. Neither claim is measured, so treat them as patterns to test, not results.
Summary based on the full transcript. Read the original for the full argument and data.Watch the original on youtube.com