AI Platform for Journalism and Communication: BeatSquares

Branislava Lovre: Welcome to AImpactful. Today we are talking about local journalism and how it can find its way to new audiences. Because the stories are there. The reporting is there. But sometimes it needs a little help reaching the right people. And maybe AI can be the part of that journey. Our guest today is René Bosch, who decided to focus exactly on this challenge: how to help newsrooms bring their local reporting to the people who actually need it, with a little help from technology. Welcome, René. It’s great to have you with us.

René Bosch: Thanks for having me.

Branislava Lovre: As with every episode, I like to start from the very beginning. Can you share what inspired you to launch Beat Squares and what did the transition look like from working in large media organizations to building your own project?

René Bosch: I would say what really inspired me was, firstly, the mass of information we have about what is happening regionally and locally. Like these are the parts where we started our whole business and how hard it is for them to actually reach the audiences that are interested in it. So, it’s mainly in a print newspaper. If you look at the German media market, still circulation of print is pretty high compared to other markets. But you have the issue that let’s face it: old people are reading print newspapers. So, you have all of this information that is… that 80% of the population are practically left out, they cannot access it. So that was really the starting point where we started to think and say, “Hey, there’s this little thing called AI,” and maybe it doesn’t have to destroy journalism. Maybe it can help make journalism better and actually distribute journalism to more people. And that was really what kicked us off and where we started going.

Branislava Lovre: You mentioned this gap between the reporting that exists and the audiences that could benefit from it. When you started thinking about how to solve this, what were the main problems you wanted to address, both from the newsroom side and from the audience side?

René Bosch: Like, the first thing we wanted to solve is to make it accessible. So that’s where we, for example, went into setting up WhatsApp channels with Süddeutsche Zeitung and really distributing the news that were already there from articles in a format that’s much better readable in WhatsApp for audiences. So, it’s really accessibility that we wanted to make better.

And from the newsroom point of view. When you think about working in the newsroom your main work is firstly doing research, getting the facts right, getting everything into an article, making a good story about it. But then if you look at the last 15 years what happened is that there’s this massive amount of work that comes afterwards. You should do a post on Instagram, you should do a story on Instagram, you should do a YouTube reel, you should make a video out of it, it should become part of your newsletter. All of this work is really far away from what you want to do as a journalist, which is research, but has to be done to reach these diverse audiences.

So that would, from a newsroom perspective that’s what drove us to say how can we get as much automation as possible by also keeping as much control in the hands of journalists that we have. So, when we talk about AI or if we talk about what we’re doing, it’s really about controllability. Every step that we’re doing is actually editable by a journalist and also changeable in a great sense, so he always gets the original sources that are used, what the AI did, and he can very easily through our fact viewer understand and say, okay, are all the names correct? Are the dates correct? Did he get the facts right?” Things that really matter in journalism where you say, okay, in ChatGPT can be a mistake. But when we talk about journalistic projects or products, there can’t be any mistakes. So that’s where we put really a lot of work into.

Branislava Lovre: So, you’re essentially helping newsrooms do more with what they already have.

René Bosch: We’re coming from a time where newspapers, 20 years ago, of course, everybody knew about the daily newspaper in their region, but that’s long gone. And right now, if you especially want to target young audiences, you have to go there and expose yourself and tell them that you’re there. And the step that we’re doing differently is when we’re helping companies doing that, for example, by generating vertical videos out of news articles. We’re not using that to build up big followership on TikTok or big followership on Instagram, for example.

But we’re trying to offer the people something and tell them… With every vertical video we’re uploading, we’re telling them: hey, do you want to come in our WhatsApp community, for example, or do you want to come in to our newsletter, which tells you every week exactly what you want to know from your city. And by doing so, we can get to the second step of this funnel, which is the loyalty part. And we really think that is crucial for publishers in the coming years to build up followerships, and also not just one big community, but really target people better to really give them the content that they’re interested in.

Because let’s face it, if you’re living in a city and you have young kids, your interests are very different from somebody in their 20s who’s just thinking about going out and seeing what’s in the cinema or whatever. And so, we’re trying to really get the stories to the people without having them stuck in the walled gardens.

Branislava Lovre: For people who are hearing about Beat Squares for the first time, what would they actually get? How does it all work in practice?

René Bosch: So, what Beat Squares does is: we plug into big content archives and that doesn’t have to be regional journalism. That could be a book publisher who has a lot of factual books in his archives, or that might be a data broker who has a lot of data in his archives. And what we’re doing is we’re actually analyzing all of the items, might it be articles, might it be books, inside of these content archives. And we’re then checking how much information do you have for which target group and how can it fit. And if we find a fit as I already said as an example, for example, young families with kids, or commuters who use a car to go to work.

If we found that connection, we’re deciding and saying what would be a format that these people would like to consume, or in which format would they like to get the information from you that are already in the archives, and then we go to work and produce the format, and obviously, if you’re a car commuter, probably a newsletter is not the best way to consume something because you’re driving a car, but it might be a podcast. So, we’re producing out of your articles, inside of your content archive, a new format.

And the whole business sense behind that is: we’re saying, we want to heighten the return on research. So once the research is done, how can we get it to as many people in as many formats as possible.

Branislava Lovre: What surprised you the most about how audiences responded?

René Bosch: The first feedbacks are actually really good. So, we started a project with Süddeutsche Zeitung. It’s one of the big quality newspapers in Germany, and what we said, we want to reach every municipality around Munich and build channels just for them. In this case we said, WhatsApp, maybe in an international context I have to explain a little bit, WhatsApp is amazingly big in Germany. 81% of the people in Germany use WhatsApp. So, we said, how about we create a channel for each of these municipalities.

What we did is, we took all of the articles from Süddeutsche Zeitung and analyzed them, said, are they fitting for the respective municipality. But we also take stories from official sources there. So, what is the city hall saying, for example. And it really goes down to stuff: when is the trash going to be picked up. Because that’s what people care about in a local context. Right.

And then we started this project and we were all really curious because it was the first one, we did in that way. We thought, okay, how will this work, will somebody subscribe, or… Will it not interest anyone? Or people are not interested in local news, especially. And what happened is the absolute opposite. So in a little bit more than two weeks, we got more than 7,000 subscribers into the channels and we are doing constantly, we’re working together with Süddeutsche Zeitung on it. Obviously, there’s editors in the process reviewing what we’re doing, what is actually published, and we’re doing surveys in the channels and we get amazing feedback, amazing engagement rates. And in the end 7,000 subscribers in two and a half weeks, which is much, much more than we actually expected.

Branislava Lovre: A lot of newsrooms tried using AI and then gave up because it didn’t work as expected. From your experience, what do they usually get wrong?

René Bosch: So, what we see is that it’s really hard, and I think we’re in this, I call it the second phase of AI. You know we all started 2023, 2024… We were all… you had the feeling like you can give to the models whatever you want and they will just magically make everything you want. Exactly and perfect. And everything is very easy. And now we’re a little bit in the second phase. When we talk to people, they tell us, “You know, actually, we thought about this and we thought about this and we tried, and we figured out it’s super hard, or it works like in a project for a while and then it stops working, or it’s super hard to implement it and to improve it.”

So, from a business perspective, obviously, I would always say, if you want to try it, talk to Beat Squares, we will help you do it because that’s our job, and we’re just trying to get better every day and make a better toolset for publishers. But if I would have one point where I would say, “Okay, where should you start?” It’s, or maybe let’s make it two points. It’s firstly the structured output large language models. I will tell you more about that in a second. And the second is really fact verification, because I think trust is the most important value that we have as journalists or also as content, factual content producers in general. And the first step, structured output, is one means to that. So, it’s really writing, not only prompts, and just hoping for the best that the things come out to you that you want.

But structured output means that I really tell the model very closely what I want to have in what position. And it gets a little bit technical. I’m trying to spare you with the details, but it’s very important to think of an article, for example, as kind of a dataset, right? It has a lot of data and it has dates and it has facts in it. It has maybe locations in it. So, you can all save that in a much more structured way and then use that really to send it into what we call the strategy, which is a pattern of both large language model steps, but also deterministic use of the content that you have. And then you can really use it to make out of an article or several articles, a very structured way of working with it.

And then what we always do is we call AI the most stupid intern you can have. And we’re not doing so because we don’t like it. We do so because we think you have to explain it very exactly what you want it to do. If you do so, you have a good chance to get a good outcome. But if you don’t and if you just throw it in and hope for the best the outcome sometimes will be very good, but most of the times it won’t. So having structured output and then using a good verification engine, for example, the one from Beat Squares, that is really the things you should look out for.

Branislava Lovre: Earlier you mentioned the idea of high return on research. Can you explain what that means in practice for a newsroom?

René Bosch: Yeah, absolutely. What we saw when I worked for Axel Springer, is that an article in general, from a business perspective, has one immense problem, and this is the time it has to earn money for itself. Every article costs a lot of money because there is a journalist involved, there’s a whole structure around it involved. So, an article mostly is on the platform, on the digital platform of the news media, and it’s in the print newspaper. And these are the two places where it can earn money, right. What we saw is that mostly 99% of the traffic that it needs to earn money, is done in the first 48 hours. So, after 48 hours you could say from a business perspective, and I’m only talking about the business perspective, obviously, the article becomes worthless. It’s lying in the archive, but you cannot earn money with it anymore. So that makes it very hard, especially if you have very complicated topics, or very niche topics, or very intensive topics, to really build a business out of it.

And what we want to do, and that the term we called with “return on research,” is trying to break that theme and to say, “What forms could we use this article in, so it would have more chances to actually earn money for the journalist who wrote it?”

That is really the thing where we say, we want to do this inside of a podcast. If we want to do this inside of a vertical video, if we want to do this inside of a messenger channel, we can use this article several times. And by doing so, we actually have the possibility that it doesn’t have to be stuck in those two days. But maybe it’s still included in a weekly overview, and it’s contributing to a podcast a month later. Or even if you say going to the end of the year like you say, “Okay, there’s this yearly overview,” and we just automatically include it in there. So this is what it’s all about to say, how can you actually get the content to the audience that is interested in it. Not even at the day where it’s produced or the day after, but through a much, much longer time.

Branislava Lovre: That makes a lot of sense, extending the life of good journalism. But there’s also a lot of fear right now about zero-click search and AI systems answering questions directly. Is this something you are worried about, or do you see it differently?

René Bosch: We’re now too much concerned about search engines and not about find engines. Right. And I think when we talk, especially about journalistic formats people don’t tend to search for it. They much rather find it. So, you can still see, we have very big news platforms that people tend to go to, or I’m always half-jokingly tell people, do you really think in five years, you will switch on your television and then you will only see a text field and you will write exactly what movie you want to see and what the plot should be? I really don’t think so. I think it’s really about concentrating on what publishers do best. And that’s reporting, and that’s having stories and actually producing them in formats that people like to consume them. And then there will always be a place where people come and where people enjoy. And then we don’t have to worry so much about a zero-click future.

Branislava Lovre: You seem optimistic about the future, so where do you see the real opportunity? How far can a small team actually go today if they use AI the right way?

René Bosch: The beauty of what we have right now, if we think about all of this new technology as a tool and not as something that will replace everything magically, is that one to three persons can already get very far, in for example, covering what is happening in a city, because they are really freed of a lot of work that they normally would do, and they can now do it automatically. Just think about it like the way we are thinking about it. You can be three people in a city and you can have a podcast, you can have five, six, seven newsletters catering to different audiences. You can have vertical video on social media, but you can focus yourself on researching, talking to people, doing interviews, asking what people really care about, because these are all things AI cannot do.

I don’t think we will go to a future where robots will go out and interview people. I just think that’s not something that will be happening, at least in the nearest future. So that is the positive aspect. I think that you can reach much more diverse audiences with less work, transforming your content or your research into different formats.

Branislava Lovre: We talked a lot about results today, but I imagine the beginning wasn’t easy. What helped you most when you were just starting out?

René Bosch: Like, what was definitely really helpful, was support by other people. So, when we started the company in August ’24, the first thing we did is really reaching out to everyone that we knew in the media sphere and talking to our idea and how we want to shape it. And it really made our idea so much more crisp and so much more catered to what newsrooms especially needed, that we really had a much clearer view after that.

And then also what was really helpful to us, we had the Media Innovation Center here in Babelsberg, which is close to Berlin. They supported us both financially, but also consulted us on where to go, gave us some contacts. That was really, really helpful for us to have this also first proof point to say that, look, there’s somebody who believes in us and somebody who believes we’re going in the right way.

And then going further, really just people daring to see a positive future. Because we at the beginning, we were fighting a lot with people who were I would say near depressive, and saying, “Okay, local journalism is going down and we cannot save it.” And it’s like going out there and telling people that, no, there is a positive vision and that it is possible. And I think now we already proven that it is, was very, very helpful and helped us a great thing in the start.

Branislava Lovre: Starting a company is never easy. For anyone listening who might be thinking about taking that step, building something in this space, what advice would you give them?

René Bosch: When you have an idea and you come to the point where you would say to yourself, “I see a possibility that it might work,” that is enough to start. It’s really, you don’t need the perfect idea, and you don’t need the super polished slides and everything. And you don’t need to be an expert in everything. If you have an idea, and maybe you are not the only one who thinks that it could work. But there are some other people also telling you, “Yeah, it doesn’t sound bad,” then it’s good to start. And don’t take too much advice from people that tell you that it won’t, because that will mostly hold you back. But that’s my thing. Don’t wait for the perfect moment and everything to be polished, but just like start.

Branislava Lovre: Great, thank you so much for your time and for being our guest today.

René Bosch: Thank you so much for having me. And all the best for you, Brana, and happy to talk to you again.

Branislava Lovre: Thank you for watching. Another episode of AImpactful. See you next week.