Participation in “AI and Digital Marketing: From Promise to Reality”, organized by IAB Spain.

A surprising data point opened the debate: 85% of companies in the European digital ecosystem already use AI in marketing. But Jesús Serrano put it in context with brutal honesty: only 7% of those projects generate real revenue impact. That gap is what framed the entire conversation.

In advertising creativity, the examples say it all. Coca-Cola openly embraced 100% generative content for Christmas and set a trend, even if the technical result was still improvable. McDonald’s did the opposite: it used AI without disclosing it, received the backlash, and pulled the ad. Serrano’s takeaway: an amateur with AI does not outperform a creative professional with AI. The specialist will be the one for whom AI provides superpowers over the generalist.

In the next panel, Cristina Lera García and César Alonso focused on proprietary data. The competitive advantage is no longer about which model you use, everyone has access to the same ones, but about how you manage your first-party data and how you feed it into AI systems. The conversation moved from “human in the loop” to something more mature: the human as the architect of context, not the executor within the process.

Borja Lizarraga Rodríguez and Sylvain Weill closed with insights from more than 200 projects across Europe: pilots that fail to scale almost always fail for the same reasons—scattered or unstructured data, processes designed for a world without AI, and teams lacking the necessary skills. Their framework: reinvent workflows, upskill teams, and build strong data foundations before talking about agents.

The final session, where I participated, was moderated by Gonzalo Martínez Rojo in a panel on regulation that I shared with Natalia Martos Díaz and Francisco José Montalvo Abiol. The EU AI Act is already in force in its first phase. But the real problem, illustrated with the case of an Ibex 35 bank, is not only complying with the AI Act: AI systems often bypass existing legislation by default, intellectual property, biometric data and image rights, simply because the marginal cost of generating content is almost zero and there is no visible friction. Under European law, what a machine generates without sufficient human intervention has no author and no rights. The practical recommendation: document the creative process, record the prompt, describe the human contribution, and verify the licenses of every tool used. Compliance by design as a competitive advantage, not as a barrier.