Automated data collection delivered faster and more reliable trading insights.
A large Nordic company with employees all over the world wanted to make it easier to find their way around, coordinate the workday and access the right information.
The ambition was to bring it all together in one internal AI assistant that could work across countries and locations.
BCT helped design and build the infrastructure behind the solution.
The company's headquarters is large and complex. Every day, employees spend time finding their way, finding available meeting rooms and looking for information that is relevant to their particular role and location.
That was the kind of friction the assistant was meant to remove.
But the challenge wasn't only technical. The solution had to work across several countries with different requirements for AI models, data handling and compliance.
The project included:
The result was “ME AI” – an internal, voice-controlled AI assistant built for the small tasks that otherwise take unnecessary time during the workday.
Instead of introducing yet another system employees would have to learn, the assistant was built to make existing tasks easier.
Among other things, employees can book meeting rooms in seconds, find their way through the building via a 3D model, call elevators and get news that is relevant to their role and location.
From the start, the solution was built around three principles:
We built a RAG-based data pipeline that retrieves content from SharePoint via API, processes it and stores it as embeddings in a vector database.
This means the assistant's answers are based on the company's own, up-to-date knowledge – rather than generic answers from an AI model.
We built the backend with HONO and Prisma, so it could handle different tools, workflows and local configurations in one unified solution.
At the same time, the architecture was built to be extended as more features and integrations were added.
The interface was built in React and integrated into the company's existing digital environment.
The focus was on keeping the experience simple and intuitive, so employees could quickly start using the assistant without having to learn yet another complex system.
The solution was hosted on Azure and connected through the Vercel AI SDK.
At the same time, the company's data and core model setup were kept in its own infrastructure, so it could retain control over sensitive information.
One of the most complex parts of the project was getting the same AI assistant to work across countries with different legal requirements.
Employees in Billund, Monterrey and Shenzhen meet the same user experience. Behind the scenes, the system automatically adapts the choice of AI model and which features may be used, depending on the location.
This made it possible to create one coherent product across the organization without compromising on local requirements for compliance and data handling.
The solution made it possible to:
The project also showed that an internal AI assistant can be both useful in everyday work and built for the requirements that come with a large international organization.
Automated data collection delivered faster and more reliable trading insights.
We helped Hafslund Kraft move from manual finance processes to a modern, cloud-based solution.
Governance and security for health data at national scale
