The Bundesliga transforms fan experiences with AI using Amazon Nova

Operated by the Deutsche Fußball Liga (DFL), the Bundesliga’s overarching goal is to be the most fan-centric league in the world. This means providing match content that connects with all audiences, no matter where they’re located. With such a large global presence, including over 500 million fans in Asia alone, the Bundesliga needed to increase the number of languages in which its content is available while also speeding up distribution.

“Being the most fan-centric league in the world means that we can provide our content and our products not only in German but in several languages within minutes. This is where generative AI is a key player for us,” says Alexander Krey, head of product at DFL Digital Sports.

Previously, Bundesliga editors followed a time-consuming manual workflow translating content into each language one at a time, and locating the right image or video to accompany the text. To accelerate the localization process and keep pace with growing content demands, the Bundesliga turned to generative AI. “Using generative AI on AWS gives us the tools to quickly scale our content in several languages, making the lives of our editors much easier than before,” says Krey.

To find the best solution for scaling and speeding up its content production, the Bundesliga began evaluating different foundation models. Having already worked within Amazon Bedrock—a service for building generative AI applications and agents—for some time, the Bundesliga used Amazon’s Working Backwards process, starting with specific use cases and working backwards to uncover how it could achieve its desired outcomes.

Using this process, the Bundesliga identified an opportunity to use Amazon Nova foundation models to enhance its localization capabilities with frontier intelligence. Embarking on a proof-of-concept phase using Amazon Nova, the Bundesliga team was impressed with its speed, quality, and cost efficiency. “Where Amazon Nova really started shining for us was the step from the proof of concept to production-ready tools,” says Krey. “The Amazon Nova models deliver their results fast. This makes it simple for us to start with a small use case and fine-tune from there.”

A team of experts from DFL and AWS started building three use cases powered by Amazon Nova models for the Bundesliga: AI Live Ticker (AILT), Content OS, and Intelligent Generation of Metadata (IGM). AILT is a real-time commentary system, using Amazon Nova models to produce live written match insights in multiple languages and varying tones—for example, casual or professional. The second use case, Content OS, streamlines postgame content production with AI-powered, transcription, translation, voice-over, subtitling, and distribution. Finally, IGM uses Amazon Nova to scan and extract metadata from the more than 50 years of content in the Bundesliga archive, making it readily searchable with natural language processing. “Amazon Nova empowers us to rapidly scale our content across multiple languages, significantly streamlining our editorial workflows,” says Krey.

Using Amazon Nova foundation models to strengthen its localization and content production, the Bundesliga is now set to communicate with users in their native language. The league aims to roll out the solution to broadcasters to establish a deeper connection between fans and their favorite teams. This elevated customer experience provides a greater sense of community among fans and contributes to the global growth of the Bundesliga. “Our main and biggest target group is still the German market, but through generative AI, we have the chance to scale content in several different languages,” says Krey. “This means that we can build up communities in other regions like North America.”

In addition to transforming the audience experience, the Bundesliga has accelerated its content production dramatically. processing time for video localization has been reduced by 75%—and match insights can now be communicated to viewers rapidly. “We can make faster decisions based on more metadata, and the time it takes to produce features is much faster than before,” says Krey. “We couldn’t do this without generative AI or Amazon Nova.”

With their time diverted from manual translation, Bundesliga editors can now focus on producing richer, more creative match content. And although human intervention is still an important part of the localization process, the Bundesliga is seeing a noticeable improvement in output quality. “Everything we push to our live environment is double-checked by an editor, but for the majority of the content, we can just give it the green light,” says Krey. “We’re definitely seeing an improvement in the quality of the output.”

As a result, the Bundesliga is seeing cost reductions for content production workflows. The cost of localization per match, per language has decreased by over 99 percent, from roughly 9.5 US dollars to 0.05 US dollars. What’s more, using Amazon Nova for Content OS has delivered a 3.5x cost reduction in large language model inference compared with similar models.

Going forward, the Bundesliga is scaling its use of Amazon Nova to cover new AI use cases and is planning to expand localization into South America and Asia. Along with piloting new integrations, the business plans to use Amazon Nova to improve fan-facing products that are already in place. “Whether it’s localization, personalized content, or tools to help editors produce at scale, from our perspective, generative AI will be part of every feature in the future,” says Krey.