What you will find in the article
- Agentic ad buying has moved from pilot projects toward wider Czech market adoption.
- R2B2 connects around 67 publishers and nearly 700 million monthly impressions.
- The first Czech connected-TV campaign bought through an agentic process has launched.
- Advertisers can set campaign criteria and review or adjust AI-generated media plans.
In June this year, the Czech digital advertising market saw its first purchase of advertising space by autonomous AI agents. The pilot project, involving Omnicom Media and Mafra on behalf of Skylink (now Canal+), was powered by R2B2. Rather than negotiating in the traditional way, a buyer at the agency entered the campaign requirements in natural language through a ChatGPT interface. The system then automatically identified suitable advertising opportunities and independently handled the subsequent communication and purchase. R2B2 also announced plans to expand the platform further and bring additional Czech publishers on board. The technology, known as agentic advertising, marks the next stage in the evolution of digital advertising and has now arrived on the Czech market. We spoke to František Bauma, CEO of R2B2, about what lies ahead.
AI is clearly becoming increasingly prevalent in customer services. In June, your company completed the first purchase of advertising space using AI agents. What stage has agentic ad buying reached in the Czech Republic, and how do you expect it to develop over the coming months?
We announced the first purchase made through agentic buying in June and have continued to expand our network since then. We currently have around 67 publishers on board. Among the larger names in the Czech market are Economia, Vltava Labe Media and Nova. In Slovakia, our partners include the Slovak arm of Our Media, Zoznam and Ringier. A number of smaller publishers are also involved, and we are continuing discussions with others about potential partnerships.
We have also succeeded in extending agentic buying to connected television (CTV). Together with Hybrid Company, we recently delivered the first CTV campaign bought through an agentic process in the Czech Republic, for Lázně Libverda. This demonstrated that the principles of agentic buying are not limited to websites but can also be applied to other media channels.
We are also in talks with publishers in the international markets where we operate. Agentic buying is not yet commonplace there either, just as it was not in the Czech Republic before we introduced the option. We therefore see considerable scope for further development in those markets as well, as the subject is attracting interest among publishers.
What television advertising inventory is currently available through agentic sales?
Thanks to our partnership with Hybrid Company, we have a solution that is technically ready for Titan OS, an operating system used in television sets. Specifically, we are talking about advertising placements on the TV's home screen, where viewers choose whether to watch Netflix, live television or another app, for example.
Is agentic buying already fully operational, or does it currently offer only a limited range of services while gradually moving towards a full launch with all the planned features?
The technology is ready and working. The underlying principle is quite simple and essentially the same as that used by e-commerce platforms such as Shoptet or Košík. An advertiser opens the interface of their preferred AI tool, enters a request and lets the agent handle it. For example, they might say: ‘I have a budget of CZK 100,000 and need to sell sports shoes. What should I do, and where should I invest the money?’ The agent then recommends the most suitable advertising inventory based on the brief.
Agentic buying is technically ready and operational. The underlying principle is quite simple and essentially the same as that used by e-commerce platforms.
What advertising inventory is currently fully available? You mentioned around 67 publishers in the Czech Republic and Slovakia. Does that include the biggest players? Is Seznam also involved?
Although many of the leading publishers are on board, Seznam is not among them. I expect it will eventually follow the same route, helping to establish a market standard. We currently have around 300 websites available, representing almost 700 million available impressions per month. As I mentioned, we have also recently integrated CTV on the technical side, so television advertising inventory can now be purchased as well.
How much control do advertisers have over a proposed media plan if they are unhappy with certain elements? Can they adjust it to suit their requirements?
The client gives the agent a brief, which can be very broad, specifying little more than a budget and product category. That may be enough. Equally, the advertiser can set dozens of specific criteria that the campaign must meet, including what may and may not be included. The client then receives a proposed plan, which they can adjust as needed through the chat interface.
Particularly in the early stages, I think it is important to ask why the agent has chosen a particular approach and why it has included a specific publisher or advertising placement. It is sensible to be critical and verify the recommendations, just as we would when using agents in other contexts.
Is agentic buying already the default purchasing method? When an advertiser approaches you, does everything happen automatically, or do they first have to actively choose this option?
If a client opens an AI agent without an MCP connection, it can, for example, find available advertising inventory and recommend where a campaign should run. What it cannot do is complete the automated purchase. To actually book a campaign, the advertiser needs to use our tool. We provide an MCP connector, which they link to the interface of their AI tool. That connection enables the agent to take the next step and book the campaign.
But can advertisers choose between the traditional, manual approach and automated buying, or is the latter becoming the natural, immediately available option whenever a client wants to purchase advertising space?
Before agentic buying became available, an AI agent could only recommend advertising inventory. There was no tool to actually buy the advertising. An agent might suggest a particular campaign, with a banner from one publisher and a video format from another, but that was as far as the process went. The client then had to pick up the phone or email the publisher to discuss the details. An exchange of messages would follow, with the client clarifying their requirements and the publisher explaining what they could offer.
Agentic buying takes the process a step further, all the way through to campaign execution. The agent not only recommends a campaign but also asks whether it should book it. It then communicates with the agent on the other side – the publisher's agent – and passes on the campaign request or booking. The publisher can approve or reject it. If approved, the campaign is created in the publisher's ad server.
Of course, the tool can also be used solely to recommend advertising inventory, without completing a booking. If the client is unhappy with the proposal for any reason, they can decide not to book the campaign or choose to deal with the publisher directly. They might, for instance, call the publisher and explain that the agent has recommended a particular campaign which they would like to run with them. The automated process therefore does not take away the freedom to make one's own decisions or communicate directly with publishers.
Agentic buying takes the process a step further, all the way through to campaign execution. The agent not only recommends a campaign but also asks whether it should book it.
What are the main benefits of agentic buying for advertisers?
The main benefit is speed, and that works on two levels. First, agencies and advertisers can handle many times more campaigns. But I think the second benefit is even more important: reducing waiting times and shortening the entire buying process. Today, much of the work is handled through email correspondence, which can take hours at best and several days at worst. That drags out the process of arranging a campaign. With agentic buying, however, a campaign can be processed immediately and very likely launched the same day. All that is required on the publisher's side is a single click to approve it. The publisher checks that the campaign meets its terms and specifications and that the agent has followed the brief. If everything is in order, the publisher approves the campaign and it can go live.
Have you seen an increase in campaign volumes since launching the agentic buying service in June? Does the greater speed allow you to handle more campaigns, or is that not yet reflected in the numbers?
It is still too early to assess such a trend reliably. As I said, we launched the service in June, then came the summer, and now it is October. We have not yet reached the point where large volumes of campaigns are flowing through this channel or where most demand for direct advertising purchases has shifted to it. Once we reach that stage, we will be able to measure the impact. For now, advertisers are largely trying out the system and learning how to work with it. They want to understand how it functions and are testing its capabilities.
So most campaigns are still being run without agentic buying?
Yes, most campaigns are still bought in the traditional way.
What proportion of campaigns are currently bought through agents?
Traditional purchasing methods still predominate. Given that we completed the first agentic purchase in June 2026 – the first advertising purchase of its kind in the Czech Republic – publishers are naturally still learning how to work with the new product. The existing model has been deeply ingrained in the market for years. People on both the advertiser and publisher sides know exactly what to do: send an email, make a phone call, share a spreadsheet. These habits are firmly established, and the fact is that changing this mindset takes longer than introducing the technology itself. People need time to get used to a new way of working.
Does agentic buying change the commercial terms, including pricing? Is it more expensive for clients?
In general, pricing remains in the hands of publishers. Moving to agentic buying does not in itself affect prices.
How do you expect things to develop from here? Publishers are currently testing the system, but could next year see much stronger growth in agentic buying?
I think so. We can look, for example, at a survey conducted by the European organisation IAB across 44 countries. Around 30% of respondents in senior positions said they believed agentic buying would become the main method of trading in at least some markets next year. A further 28% thought it would not become the dominant trading method but would be used daily to buy campaigns. Together, that represents 58% of respondents expecting substantial adoption. Sixteen per cent said they did not know, 8% saw no advantage in it, and the remaining 18% expected further development but were not yet prepared to predict how significant a role agentic buying would play next year.
And what is your forecast for the Czech market? Is it possible to predict its development, even approximately?
It will depend above all on people. The technology is ready, and we will now see how users respond to it. We also have to recognise that widespread adoption of agentic buying could have implications for staffing, both at media agencies and among publishers. If it currently takes ten people to handle a certain number of campaigns, an agentic tool could theoretically enable two people to do the same work. Some roles may therefore change, or fewer people may be needed. Naturally, some people will be inclined to resist such a change because it may not be in their interests. Others will look at it primarily in terms of finances and costs and conclude that automation makes sense for them.
What are the main barriers to wider adoption of agentic buying? Is it primarily concern about its impact on staffing and jobs?
Much will depend on trust in artificial intelligence. We have all experienced occasions when AI works brilliantly and leaves us wondering how it managed to achieve such a result. At other times, however, it produces an answer that makes no sense or is simply wrong. Individual agencies' experiences will therefore be crucial. If an agency tests a campaign, the agent offers suitable advertising inventory and the result meets expectations, the agency will be motivated to use the tool more often. But if the first or second experience goes badly – for example, because the agent overlooks important advertising placements or suggests a more expensive solution than necessary – the agency may quickly lose confidence.
If it currently takes ten people to handle a certain number of campaigns, an agentic tool could theoretically enable two people to do the same work. Some roles may therefore change, or fewer people may be needed.
Did you develop the agent in-house at R2B2, or did you adopt an existing solution from abroad?
We started with Prebid's open-source solution, added further features and, above all, fixed a huge number of bugs. The original solution had shortcomings in areas such as access permissions. For example, one advertiser could potentially access another advertiser's campaign results. These might seem like minor issues, but they would obviously be a problem for users. We have addressed those shortcomings.
How long did development take, and how expensive was it to build the solution?
I have not calculated the cost in monetary terms, but we certainly spent several months working on it.
How much does the tool cost?
We offer various pricing models, including a technology fee, a flat fee or revenue sharing. It always depends on the particular type of advertising solution and the agreement with the partner.
Looking at the development of the Czech online advertising market, around 70% of online advertising is now sold programmatically. Is there still room for that share to grow, and could agentic buying help drive that growth?
At present, agentic sales are not programmatic in the conventional sense. To some extent, this is a philosophical question that is open to debate. In practice, the agent ultimately sets up a direct campaign in the publisher's ad server. So, if we focus purely on how the campaign is executed, this falls within the remaining share of direct sales – roughly 30% of the market.
On the other hand, if we understand programmatic advertising more broadly as the automation of ad buying, agentic buying undoubtedly fits that definition. In terms of the actual transaction and its technical execution, however, it is a direct purchase.
This creates a new commercial model somewhere between the two approaches. In my view, it combines the best of both. At its heart is a direct campaign with an automated buying process. That is the essence of this model.
Could agentic buying change the structure of advertising sales and the way online advertising is traded?
That remains to be seen. It will depend primarily on publishers and the direction they choose to take. Both existing models have advantages and disadvantages, and the same is true of agentic buying. Today, we track the balance between direct sales and programmatic advertising. We may soon also be tracking the split between direct sales, programmatic and agentic sales. We will have to see what share each model secures.
And presumably whether agentic buying expands beyond online advertising into other types of media as well...
If there is a willingness to do so, I see no reason why agents could not communicate with each other to buy any type of advertising inventory. It does not have to be limited to online advertising. Various booking and ordering systems already exist for other types of media, so in theory the same principle could be extended across media channels. When I talk to people in different industries, I can see automation advancing everywhere. Companies are trying to delegate more and more activities to agents so they can carry them out faster, more efficiently and at lower cost. It was therefore only logical that this trend would reach advertising buying too.
František Bauma, CEO, R2B2
He has been CEO of R2B2 Advertising Technology since August 2020, having previously headed product development at the company for almost five years. Since January 2025, he has also served as managing director of Caroda. Earlier in his career, he worked as a product manager and UX specialist at Webgarden.cz and was involved in web application development and training in web development.
Source: mediaguru.cz
