Building AI That Actually Scales in Insurance

Alejandro Martínez · 2026-05-13

France FinTech recently gathered insurers, insurtechs, and experts at Accenture's offices to tackle what's become the industry's central challenge: scaling AI beyond the lab. It's no longer about running experiments. The real test is industrializing, integrating, and transforming existing models into systems that work at scale. One clear theme emerged throughout the morning: understanding where the market really stands, identifying the blockers, illustrating success through real examples, and exploring shifts already happening in distribution channels. ### Moving from "It Works" to "It Lasts" Insurance has moved past the AI discovery phase. Experiments are everywhere, and some players have pulled ahead. But the edge now isn't about launching proof-of-concepts (POCs). It's about making them stick over time. As David Sardas, Director at Accenture, put it at the start: "The POC proves it works. Scaling proves it lasts." Lasting means operating in production, controlling costs, staying compliant, and keeping systems robust year after year. That's where the real trouble starts. The barriers are clear: fragmented data that's tough to use, complex and messy IT systems, high compliance and security demands, and sovereignty issues still being figured out. AI adoption follows a clear path: first individual use, then team assistance, then partial execution in systems, and eventually fuller automation. Right now, most companies are stuck in that tricky transition from assistance to execution. ### It's More About People Than Tech Quickly, the conversation shifted from technology to organization. The main hurdle isn't technical. Scaling is really an organizational challenge. Valentin Bardet, Head of Customer Success FR/BE at Shift Technology, drove this home: "Resistance to change is the main blocker to scaling." Everyone agreed. A POC can work in a controlled environment, but rolling out at scale means bringing teams along, tweaking processes, and letting go of old habits. Noureddine Bekrar, Co-founder and CTO at Leocare, added another angle: "Scaling is when AI naturally fits into teams' work. You stop asking if you're using it. It's just part of your day." In other words, AI becomes invisible, not something you test but something woven into daily operations. This transformation means tackling often-overlooked issues: fear of job loss, lack of understanding of the tools, and ethical questions. Some companies find structured measures like AI charters, dedicated committees, and usage rules essential for building trust. ### Building the Foundation Early While the human side is key, technical foundations are non-negotiable. On data, the consensus is loud and clear: it's the bedrock of everything. Quality, structure, accessibility, and governance directly determine how well you can industrialize AI. Bekrar reminded us: "70% of the work is data. It's the least visible part, but it determines what you can do next. Without a solid base, nothing else holds." Here's what that foundation typically includes: - Clean, well-organized data pipelines - Strong governance frameworks - Clear ownership and accountability for data quality - Scalable infrastructure that can handle growth ### The Path Forward The message from the event was simple but powerful: scaling AI in insurance isn't just a tech project. It's a transformation that touches people, processes, and data. Companies that succeed will be those that invest in all three, not just the flashy algorithms. For US professionals watching European developments, this shift offers lessons. The same challenges--fragmented data, legacy systems, change resistance--are universal. But the urgency is growing. Those who build for scale now will lead the next wave.