Kaleva Media – A Northern Finnish multichannel media company
Kaleva Media is a multichannel media company that provides locally relevant content, service models, and digital platforms. Its media business focuses on high-quality journalism, content products, and comprehensive advertising solutions across digital channels and printed media.
Kaleva Media recognized the need to stay at the forefront of technological development, and Skillwell became the natural choice as their AI partner. Our collaboration has long been built around AWS solutions, with Skillwell serving as a trusted and strategic partner in accelerating cloud-based development and modernization. This strong foundation made it natural to expand the partnership into the field of artificial intelligence.
“Skillwell has been a reliable, long-term partner for us in AWS matters. Extending our cooperation into AI solutions has been a natural step.”
- Heikki Luoma, Kaleva Media (Director of Digital Development)
Our first joint AI initiative was a generative AI–powered moderation solution that significantly improved the efficiency and accuracy of comment monitoring across Kaleva Media’s online services. The new moderation solution makes moderation decisions independently with over 99% accuracy.
Safe discussions are essential for a good user experience
In the era of online platforms and social media, comment moderation is more important than ever. Companies like Kaleva Media face challenges such as inappropriate or offensive user-generated content and other content-related risks.
Effective moderation not only protects a website’s reputation but also ensures that users can engage in discussions respectfully and enjoy content without disruptions. Kaleva Media took a step toward the future by embarking on the development of a modern, generative AI–based moderation system.
When the old moderation model is no longer enough
Kaleva Media’s previous moderation solution could no longer meet today’s needs: it struggled with understanding context, and correcting errors consumed unnecessary resources. The goal was to build a new, intelligent, and scalable system capable of accurately filtering inappropriate content while still enabling open and meaningful discussions.
An Intelligent and flexible moderation solution
Skillwell developed an automated comment moderation system that leverages advanced generative AI, powered through AWS Bedrock. This system helps Kaleva Media’s moderators ensure that all inappropriate content, whether abusive language or disruptive messages, is effectively filtered before publication.
The AI was specifically trained to automatically distinguish harmful content and improve its performance over time.
The system also takes into account the unique characteristics of Kaleva Media’s services. For example, comments posted on reader-generated content are moderated with different weights than comments on journalistic articles. This ensures the moderation is aligned with the content type and keeps discussions appropriate across all platforms.
Technology that grows with the needs
The AI-powered moderation system was seamlessly integrated into Kaleva Media’s existing infrastructure and consists of two main modules:
1. Moderation Engine
The first module handles the actual moderation. AWS Lambda triggers the generative AI via AWS Bedrock, enabling scalable and flexible moderation as comment volumes fluctuate.
2. Langfuse Feedback Module
The second module is a Langfuse-based feedback platform that collects data on AI moderation decisions. Kaleva Media’s moderators can review this data and evaluate whether the AI handled comments correctly. This feedback makes it possible to refine the moderation process—by adjusting prompts or rules—so the system’s accuracy continues to improve.
Better discussions and happier readers
Testing the new solution demonstrated a significant improvement in moderation accuracy. The intelligent system can evaluate every comment individually and understand the context, allowing freer discussion without the risk of over-filtering.
Because the system does not rely on rigid keyword filters, it can effectively distinguish between constructive conversation and inappropriate content. The system now filters inappropriate content independently with over 99% accuracy. As a result, users can participate in discussions more freely, while the system blocks only the comments that genuinely violate the guidelines.
Looking ahead – AI as part of the whole
Adopting new technologies like generative AI is crucial for any company wanting to remain competitive and respond to the challenges of an ever-changing digital environment. It is inspiring to see companies like Kaleva Media boldly adopt advanced solutions and develop their operations to meet future needs.
Through the moderation project, we established a shared direction in which AI will be strategically applied across multiple areas. The upcoming AI initiatives will focus on operational efficiency, quality assurance, and enhancing the reader experience even further. The goal is a continuous development path where ideas are validated quickly and taken into production when their impact is proven.
“We see this as just the beginning. AI will continue to support our business in many ways and further strengthen our high-quality journalism and the experience of our readers.”
- Heikki Luoma, Kaleva Media (Director of Digital Development)
