About the Project

The Gesamtverband der Deutschen Versicherungswirtschaft (GDV) is the umbrella organisation for Germany's private insurance industry, representing over 400 member companies that collectively manage more than 500 million insurance contracts.

GDV needed a way to help their staff and members quickly find accurate, up-to-date policy information across a vast and complex document landscape. N3XTCODER was engaged to build an AI-powered knowledge assistant that could navigate this complexity.

Our Services

• AI Solution Architecture
• RAG Pipeline Development
• Azure AI Integration
• Access Management Setup
• Evaluation Pipeline Design
• Agile Development

The Challenge

GDV faced several knowledge management challenges:

• Document sprawl: Tens of thousands of documents spread across their Typo3 members' portal, internal SharePoints, and government-published legislation in PDF form

• Version complexity: Different versions of documents with different validity periods over time, making it difficult to know which information was current

• Manual research burden: Staff relied on manual research and human confirmation of document validity, which was time-consuming and error-prone

• Scale: Supporting 400+ member organisations dealing with policy questions across 500 million insurance contracts

The Solution

We built an MVP AI Knowledge Assistant using Microsoft Azure that brings together information from all sources into a single, intelligent interface.

Key features:

• Natural language queries: Staff can ask questions in plain language and get relevant answers

• Date-aware prioritisation: The system understands document validity periods and prioritises current information

• Source citations: Every answer includes a preview of the source document and a link to the original

• Conversation history: Users can refine their queries in a natural conversational flow

• Feedback system: Good/bad ratings and comments help continuously improve results

• Evaluation pipeline: A prompt library enables systematic testing and quality assurance

• Access management: Microsoft Entra integration ensures secure, role-based access

Tech Stack

• LLM: Microsoft Azure OpenAI (GPT-4o)

• RAG Index: Azure AI Search

• Access Management: Microsoft Entra

• Hosting: Microsoft Azure Sweden Central (EU-compliant, green data centre)

How We Collaborated

We used agile development methodology with regular sprints and demos. The project team consisted of a GDV Product Owner working closely with an N3XTCODER AI Solution Developer, with support from a Technical Product Owner.

This lean team structure enabled fast iteration and direct feedback loops, allowing us to deliver a working MVP in just 3 months.

The Outcome

The AI Knowledge Assistant is now helping GDV staff and members navigate their complex document landscape:

• Faster information retrieval: Questions that previously required manual research across multiple systems can now be answered in seconds

• Source transparency: Every answer links back to the original document, maintaining trust and enabling verification

• Date-aware accuracy: The system automatically prioritises current information, reducing the risk of acting on outdated policies

• Foundation for growth: The MVP architecture provides a solid base for expanding capabilities and document sources