Challenge

Securex, a leader in HR and social administration services, operates the Lex4You portal, providing legal and HR-related information to businesses and professionals. To ensure accessibility, all content is published in Dutch, French, German, and English. However, the existing translation process presented several challenges:

  • Time-intensive and costly: the combination of machine and human expert translation by specialised external translation agencies, requires significant efforts and budget tied linearly to the translation volume, making content updates slow and resource-heavy.
  • Inconsistencies in terminology: ensuring uniformity in domain-specific terminology was difficult across different languages.
  • Scalability issues: as Securex continued expanding, the growing volume of legal and HR documents made manual translation unsustainable.

Securex envisioned building an internal scalable, automated translation service combined with LLM-based text optimisation, that would ensure high-quality translations reducing operational costs and improving efficiency.

Solution

To address these challenges, Faktion collaborated with Securex AI Lab to develop this AI Translation Service. The solution leverages the state-of-the-art AI model:

  • DeepL and GPT hybrid translation: Combining the precision of DeepL with the adaptability of GPT models to refine translations.
  • Lex4You-specific glossary integration: Ensuring accurate translations of HR and legal terminology.
  • Automated XLIFF-based workflow: Direct integration with Securex' web content management system, allowing seamless content translation and publishing.
  • Post-processing AI refinement: Proofreading with AI to improve the readability and accuracy of translations.

The Proof-of-Concept (PoC) demonstrated significant efficiency gains while maintaining high output quality, prompting Securex to fully operationalise the system for large-scale deployment and usage.

Approach

The project followed a structured multi-phase implementation:

  1. PoC development & evaluation: The AI system has been set up using Securex’s HR and legal datasets, with a focus on ensuring accuracy and fluency in translations. The focus for the PoC involved Dutch-German translations. Multiple translation approaches were compared, including DeepL, GPT models, a combination of DeepL with GPT refinement, and an initial GPT model translation with GPT refinement. To validate performance, Securex' native-speaking legal experts conducted evaluations, ultimately confirming that DeepL combined with GPT refinement offered the best trade-off between speed, cost, and accuracy.
  2. Integration & automation: The AI translation service has been integrated directly within Securex' web content management system, allowing automated translation requests to be triggered whenever new content was uploaded. An XLIFF-based translation pipeline was implemented to ensure formatting, metadata, and structural integrity were preserved throughout the translation process. By automating the workflow, Securex can significantly reduce manual intervention while maintaining high translation quality.
  3. Productisation & scaling: The translation service has been expanded beyond the Dutch-German translations, to support additional languages, including French and English. A refined glossary enforcement system was introduced to enhance terminology consistency across all translations. Furthermore, CI/CD pipelines have been implemented to enable continuous deployment of model improvements and AI updates, ensuring the system remains adaptive to evolving linguistic requirements.

Outcome

The AI translation service has been successfully scaled across Securex, delivering significant benefits:

  • 70% faster translation turnaround
  • 50% cost reduction
  • 99,9% terminology consistency
  • Seamless web content management system integration

This transition from a Pilot within the Securex AI Lab to a company-wide fully operational AI Translation service showcases how Securex is embracing AI-driven automation, in a two-step approach, to enhance service delivery. By automating translation workflows. Securex and Faktion have set a new standard in AI-driven multilingual content management, ensuring efficient and scalable communication across HR and legal domains.