How AI helps French Health Insurance Claims?

Introduction

Lola Health is a French mutual insurance provider dedicated to women’s health. Their mission is to simplify access to care by offering guidance and support throughout the medical journey. As their services expanded, so did the diversity of user needs especially when it came to navigating care pathways.

To better support their members, Lola Health partnered with Rakam to build an AI conversational assistant capable of guiding users toward appropriate care options based on their profile, history, and stated preferences. In a highly sensitive medical context, this required a careful balance of relevance, safety, and full regulatory compliance.

Navigating healthcare can be overwhelming. Users often have questions about specialists, treatment paths, coverage, or next steps yet these questions are personal, sensitive, and sometimes urgent. Lola Health needed a system that could provide clear orientation without compromising privacy or trust.

They faced key constraints:

• Users needed personalized guidance based on their context and medical pathways

• The assistant had to operate in a highly regulated health environment

• Recommendations needed to be traceable, explainable, and safe

• The system had to protect confidential medical information at every stage

“In health, guidance is not just a feature. It’s a matter of trust, empathy, and responsibility.”

The Vision

Lola Health wanted a conversational interface that would feel intuitive, supportive, and aligned with real care needs not a generic chatbot or a simple FAQ. The assistant needed to understand user profiles, recognize patterns in their navigation, and respect their declared preferences.

At the same time, Lola Health was determined to meet upcoming European regulatory standards from day one, especially given that the assistant falls under a high-risk classification in the AI Act.

Objectives included:

• Offer relevant, contextual orientation through natural conversation

• Ensure that every recommendation is audit-ready and explainable

• Protect user privacy with strong safeguards and restricted data flows

• Build an AI system that institutions can trust not just users

The Solution

Rakam collaborated with Lola Health to design a conversational AI assistant and a recommendation engine built specifically for healthcare navigation.

1. A Care-Sensitive Conversational Interface

The assistant interprets user questions, understands their care context, and guides them toward the next steps in their care pathway.

• Uses profile, preferences, and navigation history to adapt answers

• Provides clear orientation (specialists, care steps, documentation)

• Avoids medical diagnosis or prescriptive language, focusing on guidance

• Responds with empathetic, non-directive tone aligned with women’s health needs

“The goal was not to automate decisions. It was to help users feel guided, informed, and supported.”

2. A Recommendation Engine Built for Safety

In a medical context, every recommendation must be consistent, traceable, and defensible. The engine was designed accordingly.

• Each suggestion includes a clear reasoning path

• Sensitive categories trigger additional safeguards or fallback messages

• All logic is stored for auditability and future review

• Recommendations adapt to regulatory expectations around safety and clarity

3. A Compliance First Development Approach

Because the system falls under high-risk AI classification, the project integrated compliance work from the beginning.

This included:

• Model benchmarking to select an ethically aligned base model

• Formal evaluation procedures rooted in the AI Act

• GDPR-specific test sets to detect data exposure risks

• Response optimization to ensure neutral, medically safe outputs

• Full logging of decision paths and model behavior

“Compliance wasn’t the last step. It structured the entire project.”

The Outcome

The conversational assistant is now an integral part of Lola Health’s service experience, helping users find clarity and direction in moments where support matters most.

• Users receive trusted, contextual guidance for their health pathway

• Lola Health ensures traceable and secure recommendation flows

• The system meets the expectations of upcoming European regulations

• The assistant enhances accessibility without replacing human expertise

“We built a support tool that reflects the sensitivity of care and the trust our members place in us.”

This project reaffirmed Lola Health’s commitment to delivering responsible digital health services, while showing how AI can improve care navigation when built with safety and empathy at its core.