Fokus Ella AI
Elisa and Sevendos Building AI-Assisted Customer Service
Ella AI is a real-time, AI-powered assistant developed for Elisa’s Fokus service portal. It is designed to support business customers with reporting, service management, and anomaly detection. The solution provides personalized assistance and helps users make informed decisions by delivering relevant information at the right time, reducing the need for manual support.
The project was driven by a clear goal: to improve the customer experience and move toward a smarter, more self-directed service model.
Ella AI was developed in collaboration with Sevendos. As part of the partnership, Sevendos handled the user interface development, including the creation of an admin interface where all AI configurations could be managed. This gave the AI developers more room to focus on core functionality and made it possible to work in fast, iterative cycles.
Why Ella AI was needed
Elisa saw that their business customers needed a clearer view of their data usage, costs, and service insights, all in one place. The current interface had plenty of information, but customers found it difficult to find relevant content or interpret it quickly. As a result, the support team was burdened with simple questions that could be automated with the right AI solution.
Putting the pieces together
- Proof of concept: We started by building a functional internal AI assistant that understands the content of the Fokus service and is able to answer user questions about how the services are used.
- Personalised experience: The solution provides a user-specific view and interface to reporting data. This allows the chatbot to highlight metrics and insights that are directly relevant to that particular user.
- Clear navigation model: We created an explicit model of the portal’s structure for the AI, so that it can understand what information is available in each view.
- Admin interface: We moved all configuration (like metrics and instructions) into a separate admin interface, which the project coordinator and tester can use independently.
- Hands-on development: A Sevendos developer built the interface to support flexible and fast updates, while also developing new configurable features as part of the iterative work.
- Broad testing: We tested Ella with different customer profiles and used feedback to assess usability and scalability.
Testing has shown that AI is able to surface user-specific, relevant information quickly and accurately. Ella lightens the load on customer support, makes services easier to use, and gives users the information they need. We're expecting a noticeable drop in manual support requests as customers begin to rely on Ella for everyday help.
Joni-Petteri Paavilainen
Customer Reporting Manager at Elisa
Insights worth stealing
We learned early on that AI doesn’t automatically understand how services, content, and documentation relate to one another. While we expected it to grasp these connections on its own, it became clear that they needed to be explicitly modeled—especially relationships between navigation and guidance content. This was largely due to tacit knowledge held by people that hadn't been formally documented.
Through the development of Ella AI, we also uncovered a broader customer challenge: users were genuinely struggling to navigate the portal and locate relevant information. The AI highlighted this gap in a way that traditional feedback methods hadn’t.
Separating configuration data, such as metrics, views, and guidance, into a dedicated admin interface proved essential. It allowed non-developers to manage the system independently and gave developers more space to focus on core functionality. This significantly improved both the pace of development and long-term maintainability.
Based on this, we recommend the following:
- Model structure explicitly. Don't assume the AI will infer it—map out navigation and content clearly from the start.
- Document tacit knowledge. What’s obvious to humans isn’t always visible to systems. Make that knowledge part of the design process.
- Build for maintainability. Move configuration into a user-friendly interface early in development.
- Use clear roles. A role-based setup—with developers, testers, and coordinators working side by side—works best when configuration is accessible and separated from code.
Looking ahead
One of the challenges we anticipate is uneven adoption. Some customers may take longer to start using the AI assistant, which could limit its impact without additional support. At the same time, those who adopt it quickly may begin to request more detailed content and user-specific data.
To manage this, the rollout will be gradual. We’re prepared to temporarily slow the pace of scaling to ensure that the solution can meet broader customer needs, such as partially replacing existing reports.
Our long-term goal is to significantly reduce the amount of manual work required at the customer interface. Once Ella AI reaches its intended level of coverage and usability, it can take over routine support tasks and free up resources for more value-adding work.
Sevendos consultant Tommi Bäckgren, who’s been closely involved in the Ella AI development project, shared positive reflections on the progress so far. According to him, the project has stayed on track without any major hurdles, and development is set to continue into next year. Tommi highlighted how the team has prioritized reliability, particularly when it comes to financial bots — with all calculations handled directly in code and navigation carefully structured. He noted that customer satisfaction is being actively monitored, and the system is equipped to respond to real-time changes. Even in cases where service providers fail to report outages, built-in monitoring mechanisms ensure continuity. Tommi also mentioned that the team is already considering future enhancements, including the exciting possibility of introducing voice control to the service.
Interested in exploring similar projects together?
I'm happy to share ideas with you!
