Despite the explosion of AI in insurance, 70 to 80% of projects remain at the POC stage. What distinguishes the actors who succeed in industrializing them?
From enthusiasm to blockage
Contrary to popular belief, the problem does not come from AI itself, but from the environment in which it is deployed. Projects are still too often approached from an essentially technological angle, focused on models rather than business uses. The result is a disconnect between IT, data and operational teams, hampering large-scale adoption.
Added to this is a major difficulty: data management. In many organizations, information is still unstructured and dispersed in documentary silos, between historical EDMs, archives and messaging systems. If we add to this phenomenon the technological debt of old and poorly interoperable systems… then we better understand the nature of the bottleneck.
Finally, scaling up comes up against sector requirements: compliance, security, traceability. Without clear governance, performance indicators, or risk management, particularly in terms of bias, projects remain embryonic.
Rationalization of the information base, a real key to success
Insurers who have successfully scaled their AI projects have approached the subject through a different prism: they do not start with AI, but with data. Because insurance is, above all, a documentary sector. Contracts, claims files, customer exchanges: value relies on the ability to effectively use this content.
However, legacy architectures are today showing their limits. Traditional ECM, often rigid, and the proliferation of repositories slow down access to information and its use.
The key therefore lies in the use of technological solutions often themselves based on AI to automate the implementation of a unified, structured and accessible information base. These platforms make it possible to intelligently centralize content, contextualize it and make it usable in real time by AI tools. In other words, using AI solutions to transform a fragmented documentary heritage into an “AI-ready” strategic asset.
Scaling up: structuring choices
The players who take the plunge rely on clear choices. They first favor high-value business use cases, allowing rapid and measurable gains, such as the automation of claims processing or contract analysis.
They also rely on technologies capable of modernizing their ECM without interruption of service. But above all, they adopt intelligent information management, which makes it possible to enhance existing data and build an evolving base, in line with the rapid developments in AI.
Finally, they integrate governance and compliance issues from the outset: data traceability, explainability of models, compliance with regulatory frameworks (GDPR, Solvency II, etc.). So many essential conditions for sustainable industrialization.
When it comes down to it, industry leaders aren’t just doing AI. They rely on the tools that allow them to create the conditions for its large-scale deployment. By relying on enhanced control of their information assets, they make their organization truly “AI-ready”.
For others, the risk is clear: remaining at the POC stage for a long time, while the pioneers are already transforming the trial.
