BR and Ippen Digital Launch Joint Chatbot

The founders of the AI for Media Network have lifted their collaboration to the product level. The BreznBot answers user questions about the Munich Oktoberfest, drawing on content from BR, Münchner Merkur, tz, and oktoberfest.de. BR and Ippen Digital aim to pilot the concept of a “Trusted Content Pool” with this initiative.

Media consumption is constantly evolving, especially since the launch of ChatGPT, with chatbots becoming everyday companions. Users ask questions, and chatbots respond. In response to this trend, Bayerischer Rundfunk (BR) and Ippen Digital have introduced a joint product: Since September 8, the BreznBot has been answering user questions related to the Munich Oktoberfest, known as “Wiesn” in Bavaria. Questions include: What do I absolutely need to know about the Wiesn? What are some quirky facts about the Oktoberfest? What will be the trending dirndl in 2025?

The BreznBot can answer in various languages, for example English. Image: BR

RAG System with Content from different sources

The BreznBot utilizes a RAG system, enabling the bot to provide personalized responses from the Trusted Content Pool. For this purpose, the system collects articles and videos related to the Oktoberfest from sources such as BR, Münchner Merkur, tz, and oktoberfest.de—the official Munich city website for Oktoberfest. These were converted into semantic embeddings and stored in a vector database. For the videos, AI-generated image and text summaries were created and fed into the vector store. When a user poses a question, the RAG system searches the vector database for relevant information, which is then sent to the language model as additional context for answering the question. This allows the model to incorporate Oktoberfest content from the source pool into its response generation—and link to relevant information.

The BreznBot boasts a particularly extensive knowledge base, thanks to content from BR, Münchner Merkur, tz, and oktoberfest.de, making it one of the few examples of a Trusted Content Pool with verified information from various sources.

BR Director General: BreznBot as a New Form of Media Cooperation

BR and Ippen Digital jointly founded the AI for Media Network in 2024. The idea to deepen collaboration and expand it through a joint product emerged earlier this year. With the BreznBot project, BR and Ippen Digital aim to test the extent to which media organizations can collaboratively use journalistic content and tailor it with AI to the interests of users.

BR Director General Katja Wildermuth welcomes the project:

“Our collaboration with Ippen Digital sets a strong precedent for innovation-driven media development in Bavaria and demonstrates that public and private media organizations can jointly develop future-proof solutions—to the benefit of our audience and in the spirit of strong, trustworthy quality journalism.” The ‘BreznBot’ represents a pioneering project for a new form of media cooperation, which we will increasingly pursue in other areas in light of the rapid technological advancements through AI.”

Testing Audience Reception of Chatbots

The Munich Oktoberfest was chosen as the topic due to its nature as a time-limited event with high user interest. BR and Ippen Digital are eager to receive user feedback on the BreznBot to better assess the overall potential of AI applications like chatbots.

The development of the BreznBot was undertaken by BR’s AI + Automation Lab, an interdisciplinary team of journalists, software developers, and product developers, in collaboration with BR24. After only three months of development, the BreznBot was launched on September 8. Despite meticulous development, the system may sometimes make errors, so users are encouraged to verify important information. The BreznBot is available as a beta version during the Oktoberfest. Following this, BR and Ippen will jointly evaluate the usage data and user feedback.

If you’re curious, you can try out the BreznBot on the websites of BR, Münchner Merkur, and tz. It speaks English and other languages when asked to switch.