IAB AI Exchange: The New Script for Advertising

Two hours. Coffee. Croissants. A former movie theater in Zurich. And a topic that no longer needs a trailer: artificial intelligence in media, marketing, and advertising. On May 12, 2026, IAB Switzerland hosted the first «AI Exchange» at the Razzia. Even the venue fit the theme. Where movies once played, the focus was now on advertising’s new script: answers instead of clicks, agents instead of funnels, brands instead of manuals—and an industry that must decide whether it’s still calling the shots or merely appearing in the credits.

«From Practice, for Practice»

Martin Radelfinger, President of IAB Switzerland, set the tone right from the start: «From practice, for practice.» René Plug, Co-Chair of the IAB AI & Data Working Group, elaborated: The AI Exchange does not claim to be the industry’s exclusive AI forum. AI has long been everywhere—in programmatic advertising, retail media, agencies, publisher models, and platform strategies. But that is precisely why we need places that help make sense of it all.

And sorting through it all was necessary. Even in a compact two-hour breakfast format, it became clear just how difficult it has become to even take in all that information. Many insights are snapshots with a short half-life. What seems like a trend today may be standard tomorrow—or overtaken by a new model the day after tomorrow. This is exactly where journalistic work begins: listening again, weighing the information, connecting the dots, and turning a dense morning of information into a comprehensible industry story.

Don't sit on the sidelines

Jochen Witte, founder of fumu, opened his presentation by outlining five trends that will be relevant for media and marketing in 2026. His analogy was simple—and spot-on: The industry must decide whether it wants to watch from the sidelines or get in the game.

The first trend: The answer becomes the new stage. When users stop searching and start asking questions instead, visibility shifts to AI systems. Websites lose direct traffic, publishers lose touchpoints, and brands lose controllable entry points. Witte highlighted how bots scrape content and how differently companies respond to this. Platforms like Amazon, Reddit, and Meta largely block access to their content. Brands want maximum visibility. Publishers are caught in the middle—with very different strategies.

In Switzerland, too, there is still no uniform approach. The NZZ is strongly opposed, Tamedia protects paywall content, 20 Minuten distinguishes between long and shorter content, and Blick specifically allows Anthropic. Behind this lies a fundamental strategic question: Do you want to be present in AI systems—or do you want to protect your content?

The second trend: Advertising is moving into AI systems. As chatbots generate responses, advertising will appear there as well. «Attention follows eyeballs, money follows eyeballs,» was Witte’s straightforward logic. The big open question: To what extent will responses and advertising be intertwined? And to what extent should AI systems be allowed to use personal contexts, chat histories, and psychographic patterns for advertising?

The third trend: automation. Repetitive tasks in creation, execution, and optimization are being automated faster than many would like. But Witte also made it clear: Strategy, the creative concept, and translating that concept into impact remain core human tasks—at least in areas where the focus is not just on output, but on direction.

The fourth trend: Agentic Commerce. As AI agents begin shopping on behalf of people in the future, the traditional e-commerce store will lose its significance. The “Buy” button will move from the store to the AI dialogue. For retail media, this could mark a turning point. When machines make decisions based on price, specifications, and availability, an age-old advertising question arises anew: What does a brand mean when it’s no longer people but agents who are making comparisons?

The fifth trend: measurement. As website traffic declines, last-click attribution becomes even more absurd than it already was. Witte spoke of a renaissance of traditional methods—but «powered by AI»: media mix modeling, attention measurement, and incrementality testing. The crucial question going forward will be: Whose numbers do you believe?

Witte's message was clear: 2026 is not the year to sit back and wait. Those who don't react until 2028 will be playing by rules written by others.

The Leadership Execution Gap

Moritz Schneider, CEO of Mediaschneider, then brought the Swiss market back down to earth with a dose of self-assessment. The IAB working group conducted an AI assessment to discuss not only potential but also actual usage. Twenty-nine responses had been received by the time of the presentation—not a representative statistic, but a qualitative snapshot of sentiment.

Schneider identified the «Leadership Execution Gap» as the most important finding. About 80 percent of participants report having a dedicated AI budget. However, only 31 percent measure ROI. And only 21 percent have a detailed roadmap.

The bottom line: Money is being invested. But it's often unclear whether it's making a difference.

The picture varied by group. Agencies have strong strategic approaches but lag behind in governance and measurement systems. Media companies often have a clear vision but are held back operationally by processes. AdTech companies are the furthest along but also need clearer frameworks. Advertisers show a strong commitment but often struggle with outdated infrastructure and data silos.

This may be the most honest assessment of the morning: The industry wants AI. But in many places, its organizational structure hasn't yet caught up with its ambitions.

A card for the Swiss market

Henning Gronki then presented the work being done on a Swiss Martech Map. His central thesis: While global Luma-Scapes display an impressive number of logos, they are often of little practical use in everyday life. After all, no one picks up a globe when driving from Zurich to the Bernese Oberland.

The Swiss Market Map is therefore not intended to highlight complexity, but rather to provide guidance. By the market, for the market. Featuring providers, experiences, and applications that are relevant to the Swiss context: multilingual, fragmented, small-scale, yet demanding.

Gronki highlighted three areas where guidance is particularly important. First, data processing: measurement, attribution, unified data layer, and marketing mix modeling. Second, creative and personalization: rapid creation of many ad variations—not as a replacement for creative agencies, but as operational support. Third, publishing and sales tools: for example, protection and monetization models for publishers, AI-powered yield optimization, or discoverable inventory for agency systems.

The Market Map thus becomes more than just an overview of providers. It is an attempt to make the Swiss advertising industry navigable before the market is permanently reshaped by global platform logic.

Publishers: Between Protection and Visibility

Lukas Görög, founder of the Academy for Artificial Intelligence and Digitalization, spoke about the publisher’s perspective. And he began with a refreshing dose of perspective: Although he advises AI organizations, he wouldn’t call himself an AI expert. Anyone who does so should go to sleep, wake up—and by then, Anthropic, Mistral, or GPT will have already moved on.

His topic: What happens to content monetization when users stop searching and start asking questions?

Görög described the shift toward zero-click. Google is no longer the search engine it once was. In international markets, media companies have lost as much as 50 to 60 percent of their traffic. Markets without strong paywall models have been particularly hard hit. The traditional media funnel is shrinking: less search traffic, less platform engagement, and fewer conversions.

If everyone can build more, the idea becomes all the more important— Lukas Görög

Generative Engine Optimization, or GEO for short, is certainly being heavily hyped, but it must be viewed with a level head. Many media companies still generate barely more than five percent of their traffic from generative search engines. Nevertheless, GEO is important. Those who do good SEO work have already done part of the job. Key factors include strong author pages, topic clusters, engagement, backlinks, and FAQ structures.

The issue of licensing becomes even more fundamental. Should media companies block AI crawlers or make content available through deals? Görög cited international examples such as *Le Monde*, which has entered into a business partnership with OpenAI. At the same time, there are publishers that consistently block access. Hybrid models are also conceivable: content is made available to AI systems after a certain period has elapsed since its publication.

The bottom line: Not making a decision is also a decision. Publishers must strategically determine whether their content should primarily be protected, licensed, or made visible.

First-party data remains a requirement

Beyond all the questions about AI, Görög reminded us of an almost classic truth: first-party data is becoming more important, not less. Those who have direct contact with their readership—whether through email, registration, login, or a community—remain more capable of taking action. In a world where platforms and AI systems are redistributing access to audiences, maintaining your own direct contact becomes a strategic lifeline.

His concept of «liquid content» was also fascinating. AI lowers the technical barrier to creating new journalistic formats. Infographics, quiz formats, scrollytelling elements, and interactive presentations can be created more quickly. Not because journalists are suddenly becoming developers, but because programming language is, in part, becoming part of everyday language. This opens up opportunities, but it also raises the bar. When everyone can create more, the idea itself becomes all the more important.

The Hybrid Media Organization

Görög’s most concrete example was that of the hybrid media organization. AI is no longer just a tool, but is increasingly being viewed as a colleague. A bot with an email address, access to data, stylistic expertise, and a clear assignment. His example was named «Luigi»—an AI employee who prepares LinkedIn posts and communicates with a personal writing twin.

That may sound playful, but it’s a serious organizational model. AI isn’t simply tacked onto existing processes. It changes roles, routines, and responsibilities. Görög also demonstrated how his small organization works with AI personas and was able to increase organic traffic as a result.

His closing message was simple: Experiment. In many areas, he said, it’s better to make one more mistake than to try to do everything perfectly. That, too, was a recurring theme of the morning: Speed isn’t everything. But without speed, strategy quickly becomes a thing of the past.

From Doers to Orchestrators

Michelle Scholz from Jung von Matt TECH wrapped up the session. Her presentation offered the agency’s perspective—and one of the most powerful statements of the morning: «Decision-making is the new bottleneck.»

Scholz made it clear: The AI hype is over. The industry is now in a phase of productivity. At the same time, FOMO is rampant. LinkedIn, OMR, tool lists, agent frameworks—everywhere, people are promising that one trick will solve everything. In reality, however, organizations vary in maturity, complexity, and resilience.

The promised efficiency gains do not automatically lead to less work. Rather, productivity gains are often reinvested immediately. Employees are not any less exhausted in the evening; rather, they are exhausted in a different way. This is because different decisions, different skills, and different responsibilities are required.

The big shift: away from simply doing things, toward orchestrating them. Away from output, toward responsibility.

The brand must become "Top of Machine"

Scholz’s central idea was the new role of the brand. In a world where production has been democratized, brands are not becoming obsolete—they are becoming more important. When customers ask LLMs in the future which car, product, or solution they should choose, it will no longer be enough to be ’top of mind.« Brands must also become »top of machine.«.

This means that a brand can no longer exist solely as a PDF, a CI manual, or a Frontify repository. It must become machine-readable, interoperable, and technically interpretable. Scholz spoke of a brand OS—with strategy, narratives, a design system, coding guidelines, and building blocks that AI systems can understand.

Her example from a museum in Winterthur showed how an interface can be created using such brand building blocks. Not merely as a demo gimmick, but as an indication of where marketing is headed: the brand as a system that provides guidance to both people and machines.

That is a crucial point. If AI is trained on the average, the brand must become more precise, not more vague. Attitude, perspective, and judgment become differentiating factors.

Agencies aren't going away—but they are changing

The subsequent Q&A session focused on the agency of the future. Scholz made it clear: We don’t necessarily need new AI-based job titles for everything. Rather, everyone needs to become more technically savvy. Roles are becoming more fluid. Designers are building front-end demos. Developers are focusing more on shipping, integration, and technical quality. Teams are working more agilely along the value chain, rather than along old disciplinary boundaries.

The discussion also turned to the issue of pricing. When clients simply demand productivity gains in the form of budget cuts, it becomes difficult for agencies. According to Scholz, those who will succeed are the ones who focus on what should not be delegated: attitude, perspective, judgment, the role of challenger, and an outside perspective.

This may be the most important answer to the age-old question of whether AI will replace agencies. AI replaces output. But it does not automatically replace direction.

A wake-up call without scaremongering

In closing, René Plug summed up what had defined the morning: The presentations were technology-focused, but not overly obsessed with technology. They weren't about gadgets, but about business models. About changes in the market. About commercial realities.

That was precisely the strength of the first AI Exchange. It didn’t just show what AI is capable of. It highlighted where the industry is already taking action, where gaps still exist, and where the next lines of conflict lie: visibility or protection? Automation or dependence? Efficiency or effectiveness? A brand as a PDF or a brand as a machine-readable system? Measurement via platforms or our own truth?

The answer to this morning’s question is therefore nuanced. What really works? AI works when it is embedded in processes, data, roles, and responsibilities—when it is understood not as a magic wand, but as infrastructure—and when organizations experiment, measure, learn, and make clear decisions.

What remains a promise? Everything that AI markets as a shortcut. Everything that confuses efficiency with impact. Everything that seeks to delegate branding, strategy, and judgment to machines without first defining what the organization actually stands for.

The advertising industry doesn't need to panic. But it shouldn't stand on the sidelines either. The rules of the game are being written right now. If you want a say in the matter, you have to play along.