Dominik Scholler, VP of product management and innovation at the DFL, explains why and how the Bundesliga is trialling the automated production of matches, and shares his thoughts on AI and its benefits for creative tasks and storytelling.

At SportsInnovation 2026 in Düsseldorf (4-5 March), the Deutsche Fußball Liga (DFL) showcased an AI-supported automated production setup that is currently being trialled in the Bundesliga. 

The workflow is created in conjunction with Sportcast, the host broadcaster of the Bundesliga, alongside Mark Roberts Motion Control (MRMC) and Pixelscope. 

It features 18 tracking cameras that pinpoint the movements of players and the ball on the pitch, and then five broadcast cameras on robotic heads that move based on that tracking data. Multiple layers of AI are involved, with all processing done locally on-site in real-time. Since the start of this latest trial in November 2025, there have been tests carried out on five Bundesliga 2 matches. 

During SportsInnovation 2026, this setup produced an AI-cut match feed for delegates to see on a big screen in parallel with a standard human-operated and cut feed. See the video below. 

Speaking to SVG Europe, DFL VP of product management and innovation Dominik Scholler says that this trial has been under consideration for a while. 

“We were asking for quite a long time: is there the possibility to automate the base signal production? There are now different aspects of this that are coming together [to make it possible]. 

“For the Bundesliga, we have, in general, quite a high camera standard, even for standard matches on Saturday afternoon and parallel matches, with 16 cameras. And then on Saturday evening, with 28 cameras and more. The cameras generate metadata, so you already have a very good [foundation] there. We also figured out that there are cameras that can, in theory, run autonomously and don’t need a human operator anymore. This has all come together.”

AI is a worrying prospect for some within the broadcast industry, who see it as potentially taking away their livelihoods. But the reasoning behind embarking on the trial is a lot less black and white, and potentially less existential. 

“It is a similar development to what we also see in all of Germany with the country’s demographics, and the lack of skilled workers [in other industries],” says Scholler. “On average, the typical Bundesliga camera operator is already aged 50-plus. At the same time, it’s very hard to find new talent, because it’s not a typical job that young people want to do. You have to work on the weekends. You have to work when it’s raining.

“The typical Bundesliga camera operator is already aged 50-plus. At the same time, it’s very hard to find new talent, because it’s not a typical job that young people want to do.”

“Sportcast was tackling that topic. They offered their own traineeships. We can still cover the number of camera operators that we need, but [this may change] in the future.”

Importantly, regardless of the rationale, there is a consensus that humans will still be required for football coverage. 

“For camera one, camera two, and the more complex cameras, there is no question that you need human beings behind that, because that is part of the creative process. But there are also other cameras, and that’s what we try to demonstrate, that are already very close to working autonomously. 

“It’s a mix of different things. It’s not our plan to cut jobs for camera [operators], absolutely not. We still need that. I mean, we just had Der Klassiker over the past weekend. That’s such a complex production. You need skilled people there. But, on the other hand, I think we have to prepare for the future so that in five or 10 years we can provide a state-of-the-art base signal.”

There are also benefits for the planet, he adds.

“Bundesliga is trying to be a very sustainable league. We know that a lot of the emissions that we create on match days are, besides the electricity that you need, from the trucks that have to drive to the stadium, but also all of the people travelling around Germany. Automation has the potential to reduce carbon dioxide emissions that we have.”

Just because it is technically possible to automate production, that doesn’t mean it’s automatically up to the same standard, however. Although initial findings are positive. 

“We are still in a testing phase. We just started three months ago. Basically, the system gets better with every match. I personally was positively surprised by [the quality]. It is already very good. 

“The camera standard is not yet at the level, of course, that we had at Der Klassiker. But I think, especially looking at the Bundesliga 2 production, where we also have remote production for two or three matches per round in place, it’s definitely an option. And I’m sure, if we develop it constantly, let’s say over the next two to three years, by 2029 [and the next rights cycle], we can be in a position to apply it on a productive level.

On a more general level, Scholler is also optimistic about how AI can help with football production creative tasks, especially when it comes to accessing archive footage. 

“I would say, primarily, [AI] is a supportive task. In the media product, where we provide or apply AI services or GenAI, we always have a human in the loop. But where we see a huge potential is where AI can then support creative storytelling.

“For example, if I think about our archive, and that’s the biggest football archive in the world, if [we are looking for] regular clips for one of our international shows, a normal editor would have to do research, probably even for days, to come up with a three or five-minute piece. In the future, he or she can type in a semantic search and look for a certain type or style of content and they will then get the recommendations that they need.

“We have a global product portfolio where we provide 100,000 minutes per season, which is a lot of footage. The feedback we get from our international broadcasters is that they want to know how to figure out what’s the relevant content for them. I stress this internally a lot too. I say we need good algorithms to recommend the content, because someone from Sky in the UK has completely different needs from Globo in Brazil or DAZN in Japan. So, you have to personalise so that they make the most efficient use of their time.”

Trials are progressing to make this a reality, he says. “Over the past two to three years with AWS, we have already been looking at that kind of usage. So, for example, one very important requirement is that you have good metadata. That’s what everyone says about AI, right? In the end, it’s about the quality of the metadata you have. [For] recent years, we have fantastic data, but the further you go back, the more unreliable or the more missing data you have. 

“At the same time, we have very good data when it comes to live production for the 90 minutes. But we also do a lot of support programming. We have media days. We send production teams to the clubs to do interviews. And there is not a standardised process yet on how to generate the metadata. They write scripts and put these in our portals. Of course, there is some general tagging, but it’s not done in a standardised way. This is work that needs to be done. But it’s also very, let’s say, administrative, mundane work. And a creative person doesn’t want to spend hours [doing this]. 

A ‘Meetup’ data project with AWS, and a second initiative called Smart Content, have pinpointed the two key elements that are required: semantic search and recommendations. 

“We did a POC in the past year where we ingested some footage in the system, and then we tried exactly these things. If we want to see Harry Kane – does the user find all the content? What happens if we ask for a ‘happy Harry Kane’? The first results were very promising. We want to create a content hub where we actually provide all these kinds of opportunities and serve our different stakeholders in the best possible way.”

The next step for Smart Content is to add an additional layer: localisation. 

“That’s also very interesting,” enthuses Scholler, “Learning AI at some point, if you don’t set it up effectively, it can get very expensive, especially if you basically ingest all your content and then let it create metadata. So you need a smart approach to that.” 

“In the media product, where we provide or apply AI services or GenAI, we always have a human in the loop. But where we see a huge potential is where AI can then support creative storytelling.”

To that end, broadcasters and other rights holders don’t necessarily use all of the content that is provided to them. So the DFL is thinking about adopting an approach that allows for bespoke requests that return bespoke content. 

“If a broadcaster says a clip is relevant for them in Japan and they would like to have it with Japanese language, then they would have the option to click a button, and then the AI rendering will happen in the background. By doing that, we can ensure that the content is really relevant and that we don’t have ineffective workflows or exploding costs. We want to test that with our international media partners.”

The DFL is currently trialling this with Spanish-speaking broadcasters and partners, including Relevent Sports, which has a content studio in Guadalajara. Scholler says this will be extended to other Spanish-speaking broadcasters such as ESPN in the US, Fox Mexico and all other Latin broadcasters. 

“And we will also try Spain itself, even though our presumption at that point is that probably they will then ask for a different form of Spanish from what they ask for in Latin America. That will also be interesting to see how and if the localisation tool will be able to offer that.”

SportsInnovation 2026, organised by the DFL and D.Sport, took place at the Merkur Spiel Arena in Düsseldorf, 3-4 March 2026