Case study
HomeAssur: An AI Case Study for Insurance
HomeAssur, a French insurer specialized in short-term rental coverage, trusted MarketerZ with two projects: an AI virtual assistant in 2025, then an enterprise AI agent for claims analysis in 2026. This AI case study for insurance shows our method in practice, from audit to production.
HomeAssur at a Glance: an AI Case Study for Insurance
HomeAssur is a French insurer that specializes in short-term rental coverage: furnished tourist accommodations, property management companies, and rental managers make up its policyholders. Its contracts cover damage caused by guests, lost rental income, and third-party liability.
Standard homeowner or landlord policies do not cover these specific risks. HomeAssur built a dedicated offering for this market, and the volume of claims and customer questions grew quickly.
MarketerZ has worked with HomeAssur since 2025, on two separate projects: a virtual assistant for the general public, then an enterprise AI agent for its internal teams.
2025: The Problem Behind the Jean-Claude Virtual Assistant
In 2025, the HomeAssur team was handling a high volume of repetitive questions: eligibility, coverage details, required documents, procedures to follow. The same answers came up again and again, on the same topics.
The knowledge needed to answer them was scattered across the website, contracts, the FAQ, and newsletters. Advisors had to track down the right information before they could reply, which slowed down every exchange.
On top of that came repetitive administrative tasks, like sending invoices or updating contact details. All of it was time not spent on personalized advice.
The Solution: a Chatbot Grounded in a Knowledge Base
MarketerZ designed Jean-Claude, an AI virtual assistant built for HomeAssur. The chatbot answers customer and prospect questions in natural language, available 24/7.
Jean-Claude draws on a knowledge base built from HomeAssur's guides, contracts, FAQ, and support tickets. RAG (retrieval-augmented generation) grounds every answer in this content, to limit the risk of AI hallucination.
A management interface lets the HomeAssur team review conversations and update the knowledge base, with no technical skills required. On the automation side, n8n handles invoice delivery and contact-detail updates. Quote requests are sent to Google Sheets and trigger an automatic quote email.
The Benefits the HomeAssur Team Saw
Without citing any figures, several qualitative benefits came out of this project:
- Immediate answers, around the clock, on eligibility, coverage, and procedures.
- Advisors refocused on complex cases, as repetitive requests were absorbed by the assistant.
- Centralized, traceable conversations, making it easier to follow up on customer requests.
- A stronger innovator image, in a market where short-term rental insurance is still a young segment.
2026: The Problem Behind Claims Analysis
Once a claim is filed, HomeAssur's claims handlers need to review several pieces of evidence: photos, invoices, estimates. Each file requires checking whether the claim is eligible under the contract and the applicable terms and conditions.
Handlers then need to estimate the payout, while keeping a clear record of what justifies each decision. This kind of document analysis is more complex than the questions a consumer-facing assistant handles: it involves precise business rules specific to each contract.
The Solution: a Traceable Enterprise AI Agent
To meet this need, MarketerZ built an enterprise AI agent dedicated to claims analysis, in production at HomeAssur since August 2026. The agent reads a file's evidence, checks eligibility against the contract, then produces a structured, well-reasoned pre-analysis with citations to the relevant clauses.
Each claim file has its own chat space, where a handler can ask the agent questions about the case at hand. An admin dashboard gives an overview of processed files and the active configuration.
Under the Hood
The agent runs on an open technical stack: LangGraph orchestrates the analysis pipeline, PostgreSQL and pgvector power the knowledge base, and FastAPI serves the API, with the Mistral API or local models via Ollama depending on confidentiality needs. The analysis pipeline is configurable, so it can adapt to HomeAssur's own business rules.
Before every production release, the agent is evaluated against reference cases (golden cases) that check the consistency of its analyses. A data loss prevention (DLP) policy governs sensitive information, and deployment can run on Docker Compose, on-premise or in the cloud. The agent integrates with HomeAssur's existing PHP-based back office.
The Benefits Claims Handlers Saw
- A structured, well-reasoned pre-analysis, with citations to the relevant contract clauses.
- Time saved for handlers, who start from a work base that is already built.
- More consistent, traceable decisions, from one file to the next.
- The final decision stays human: the agent prepares the analysis, the handler decides.
The agent's compliance with the EU AI Act was assessed: its use falls under a non-high-risk category, with transparency obligations to meet. Our approach to security and compliance applies to this kind of agent as well.
What These Two Projects Show About Our Method
These two projects, run a year apart for the same client, illustrate how we work. Every solution starts with a concrete problem identified together with the HomeAssur team, before any question of technology comes up.
The first project laid the groundwork: a centralized knowledge base, grounded in HomeAssur's real content. The second builds on that same logic, with tougher business requirements: contractual eligibility, traceability, evaluation before every release.
If you are considering an AI virtual assistant for your customer relationships, or an enterprise AI agent for your internal processes, our AI consulting method applies the same way: audit, prototype, integration, follow-up.
Have a Similar Project in Mind?
MarketerZ works with small businesses, mid-sized companies, and large SMEs, from needs assessment through production launch of a virtual assistant or an enterprise AI agent. Let's discuss your project in a free 30-minute consultation.
Discuss a Similar Project