Challenge

Securex, a leading provider of HR and legal services, operates the Lex4You portal, which serves as a primary knowledge hub for employers, HR professionals, and legal advisors. The platform contains extensive documentation on labour law, social security, and tax regulations. However, navigating this wealth of information manually posed significant challenges:

  • Time-consuming search process: Customer care representatives and legal advisors had to manually browse through large volumes of documents to find relevant information
  • Inconsistent query results: Users often struggled to retrieve precise and contextually relevant answers  due to fragmented search capabilities
  • Scalability issues: as Securex’s Lex4You database is constantly evolving, the need for an intelligent, scalable solution to streamline legal information retrieval has become increasingly urgent

Securex required an AI-powered solution capable of delivering quick, reliable, and contextually appropriate accurate responses to user queries.

Solution

To address these challenges, Faktion developed a GenAI-powered Copilot Assistant for the Securex AI Lab, with an elaborate Retrieval-Augmented Generation (RAG) system under the hood. This solution combines:

  • Integrated user experience: a smart chat interface that allows users to interact naturally with the AI, receiving instant responses along with source citations
  • Advanced information retrieval: AI-driven search capabilities that locate and extract the most relevant information from the Lex4You portal
  • Context-aware AI response: LLM technology that synthesises retrieved content into coherent, accurate answers

By integrating retrieval and LLM-driven content generation, the system significantly enhanced Securex’ legal information delivery process.

Approach

This need was addressed within the partnership between Faktion and Securex’s AI Lab.

  1. PoC development & integration: The AI Copilot has been built using a RAG-based architecture, enabling the Lex4You database to be integrated with a vector search engine, allowing the system to locate relevant content, based on the user query, efficiently. To enhance accuracy, the system processed queries using a specialised prompt engineering framework tailored for legal text interpretation.
  2. Smart agent & user experience: a chat-based interface has been developed, providing a seamless and intuitive way for Securex employees to interact with the AI. The Copilot has been designed to personalise responses based on user roles (e.g. legal advisors vs HR professionals). Additionally, the system includes reference linking, allowing users to verify AI-generated responses with direct access to original legal sources.
  3. Evaluation & optimisation: to measure accuracy and effectiveness, the Copilot underwent extensive testing using both domain expertise and Securex's internal legal FAQs. The generated responses of the system were benchmarked against human-generated answers using RAGAs evaluation metrics, including faithfulness, answer correctness, and context relevancy. The search engine was tested across multiple retrieval methods, ensuring that the Copilot selected the best possible answers for each query.

Outcome

The Lex4You Copilot Assistant can now deliver substantial improvements in efficiency and accuracy across Securex's legal advisory workflows:

  • Substantial reduction in search time
  • Higher accuracy & reliability: responses are being evaluated for faithfulness and correctness, ensuring legally sound answers
  • Enhanced knowledge utilisation: Securex’s vast legal database can now become fully accessible, which would make it easier for customer care teams to provide precise and up-to-date advice
  • Improved user experience: the smart agent UI can now provide intuitive access to legal information, with direct links to relevant source documents

The prototype has shown that Securex is able to scale its knowledge delivery, ensuring that HR professionals and legal advisors receive the right information at the right time. Upon the validation of this AI-powered legal copilot prototype, further enhancements and improvements should be developed, including expanding the knowledge domains and multi-language support.

By combining AI-driven retrieval and contextual knowledge generation, Securex and Faktion have demonstrated that accessing, processing, and delivering legal information can be much more quicker and more accurate, in an intuitive way. This will enable faster and more informed decision-making in HR and labor law contexts.