Go back
HDI Global Built Enterprise-Grade AI Applications with Kern AI
HDI Global Netherlands, a leading industrial insurer, partnered with Kern AI to fast-track the development of secure and scalable Generative AI applications across multiple domains and use cases. Confronted with fragmented knowledge, technical skill gaps, and rising internal demand for AI-driven solutions, HDI adopted Kern AI’s data-centric platform to build a suite of LLM-powered tools.
HDI Global Netherlands
HDI Global Netherlands is a specialty & corporate insurer based in The Netherlands, with its home office in Germany, HDI is part of the Talanx Group. It provides tailored insurance solutions for complex and niche risks across sectors like marine, energy, and cyber.
The collaboration empowered business units to embed their domain expertise directly into the LLM applications, enabling the AI to better understand internal processes and deliver more relevant results than standard out-of-the-box solutions. By accelerating access to knowledge, reducing manual effort, and improving consistency in responses, HDI is now well-positioned to scale Generative AI adoption across multiple lines of business — all while maintaining full control and compliance.
Application & Use Case Overview
Together with HDI, we’re developing a range of innovative applications across diverse use cases, including:
SCOT - A company-wide assistant powered by cutting-edge language models. It adapts to each task, offering deep, fast, or complex answers, and can process documents and browse web pages in real time.
Cyber Agent - The Cyber Agent is an AI tool that extracts key insights from cyber insurance questionnaires. Instead of manual analysis, employees can upload files and let the AI handle the rest—freeing them to focus on decisions and customer support.
Product and HR Assistant - The Product and HR Assistants provide expert guidance on HDI's products and processes, making internal information easier to find and understand.
Challenge / Problem
HDI faced several recurring challenges, particularly around knowledge access, efficiency, and internal processes. To address these, they implemented SCOT, which allows employees to quickly access reliable information by dynamically selecting the most suitable large language model (LLM).
This significantly reduces the time spent searching through documents or websites. Additionally, the Cyber Agent automates data extraction from cyber insurance questionnaires, eliminating the need for manual analysis and reducing the risk of errors. By removing repetitive, low-value tasks such as data entry and document lookup, employees are able to focus more on critical decision-making and customer engagement.
Furthermore, the Product and HR Assistants streamline internal processes by making product knowledge and procedural guidelines readily available, enabling teams to navigate complex workflows with greater clarity and speed.
Solution Overview
Platform Architecture
HDI applications leveraged Kern AI’s full-stack data-centric AI suite:
- Data Management - All data was processed, chunked, and stored in our data management system, ensuring it could be efficiently accessed by a large language model (LLM).
- Information Retrieval - We developed a custom information retrieval pipeline designed to add context to user queries by accurately extracting relevant knowledge from the valuable documents of the company.
- LLM Integration – We benchmarked leading large language models to identify the best-performing option and seamlessly integrated it into the retrieval process. Those models can be updated continuously as a drop-in replacement for newly released and improved models.
- Custom UI – We delivered a custom user interface fitting to the corporate design of HDI that offers modern, secure, and intuitive access to the application for domain experts.
- Web Integration - Using internal data is not always enough. Using the web searching capabilities of Kern AI, the latest and most up-to-date information can be used in the blink of an eye.
Collaboration Model
During the development phase, Kern AI worked directly with domain experts to ensure that HDI’s custom logic and processes were embedded directly into the solution for advanced customization. The final setup enabled non-technical domain experts to easily access a powerful LLM-powered application that understands how they work. While providing technical users with the building blocks to continuously optimize and scale the solution, directly on our platform.
Impact & Results
SCOT – Company-Wide Assistant
- Reduced time spent searching for internal information by 30%
- Achieved a 75% employee adoption rate within the first 6 months
- Enabled faster cross-team collaboration through instant access to shared knowledge
Cyber Agent
- Cut down questionnaire analysis time by 80%
- Increased data extraction accuracy, reducing follow-up clarifications by 50%
- Improved risk evaluation through consistent data structuring and interpretation
Product & HR Assistants
- Lowered internal support requests by 60%
- Improved onboarding speed and information access for new employees
- Boosted employee satisfaction through faster answers and reduced internal friction
Tremendous flexibility in building use cases
With KernAI, we gained tremendous flexibility in building use cases and utilizing their APIs, all while operating within a safe and secure environment.
Remy van Wijk
HDI Global Netherlands
Why HDI partnered with Kern AI
HDI selected Kern AI because of:
- Data-First Approach . Our solution enabled high-quality data inputs for LLMs, resulting in improved accuracy and performance.
- Enterprise flexibility - Easily integrated with existing enterprise tool.
- Customizability – The solution was built with modular components, allowing HDI to adapt and scale the application as business needs evolve.
- Security & compliance – The solution was built with enterprise-grade data privacy and access control, meeting strict compliance requirements.
- Transparency & control - Admins could trace and refine model behaviour based on input/output monitoring.