Challenge
Securex, a leading provider of social administration and human resources services, operates across multiple business units, each addressing distinct operational needs. The Lex4You department serves as a hub for specialised legal advice, offering HR news, labour law updates, and critical regulatory information to businesses and entrepreneurs. Central to its operations are Collective Labour Agreements (CLAs) - legal instruments negotiated between employer groups and employee organisations, outlining workplace rights, obligations, and benefits.
Every year, Securex's legal experts meticulously analyse new and updated CLAs to extract key insights on topics such as purchasing power, additional benefits, premiums, transportation costs, and working hours. These analyses form the foundation of employment agreements available on the Lex4You portal, which is used by customer care agents and legal advisors to assist businesses in navigating labour laws.
However, this process presented several key challenges:
- Time-intensive manual processing: legal professionals spent extensive time in reviewing and summarising CLAs, making updates a slow process.
- Inefficient search & retrieval: Customer care teams relied on long, complex documents, making it difficult to quickly retrieve relevant information
- Lack of AI-enhanced summarisation: without automation, there was no efficient way to extract only the changes from previous years, which is the primary interest of businesses consulting these agreements
Securex required a system that could automate CLA document summarisation, and more specifically, the summary of changes in CLAs. The system should maintain high accuracy, comply with legal terminology and produce structured outputs, ensuring customer support and internal workflows.
Solution
To address these challenges, Faktion developed a Legal AI system tailored for Securex AI Lab. The system was designed to:
- Automatically analyse CLAs and identify differences between previous and current versions.
- Generate concise, structured summaries that highlight key updates.
- Enhance searchability by extracting relevant keywords and metadata.
- Seamlessly integrate with the Lex4You portal, making summaries readily accessible to customer care representatives and legal advisors.
The LLM model was tweaked to:
- Detect legislative changes based on key indicators such as ‘increase’, ‘adjusted’, ‘modified’, etc. or date-specific references
- Ignore irrelevant sections and focus on substantial updates
- Maintain consistency in terminology and formatting, aligned with Securex’s legal standards
Approach
The implementation followed a structured methodology, combining data analysis, iterative model development, and human validation:
- Data analysis & processing: Identify key sections within CLAs and determine relevant content. Faktion also applied text pre-processing techniques to extract structured data from .docx files while filtering out irrelevant sections.
- Summarisation model development: By using zero-shot and few-shot summarisation techniques, Faktion generated structured summaries. Faktion also designed a change-detection mechanism to compare documents and highlight updates, and integrated keyword extraction techniques to enhance search functionality
- Human validation and continuous improvement: Faktion conducted multiple iterations of AI-generated summaries, incorporating feedback from Securex’s legal advisors and refining prompts and instructions to improve clarity, brevity, and domain-specific terminology accuracy. Moreover, Faktion addressed discrepancies between the Dutch and French versions to ensure consistency across languages
- Final integration & evaluation: By integrating the tool into the Lex4You platform, Faktion made it accessible for Securex’s customer care and advisory teams.
Outcome
The summarisation system has significantly improved efficiency, accuracy, and accessibility within the Lex4You department. Key results include:
- 100% accuracy in content identification.
- 90% compliance with Securex-specific legal terminology.
- 50% reduction in summarisation and update processing time.
Which ultimately led to:
- Faster response times.
- Improved searchability.
- Higher accuracy & consistency.
- Reduced manual effort.
Looking Ahead
Following the success of this project, Securex plans to further enhance the AI-powered Lex4You summarisation system by:
- Expanding AI processing to additional legal documents beyond CLAs.
- Refining multi-language capabilities for better alignment between Dutch and French summaries.
- Exploring full integration into Securex's MyH&S platform, ensuring legal and compliance information is centralised and easily accessible.