Frans Feldberg is a familiar face on the main stage at Data Expo. As a professor of Data- and AI-Driven Business Innovation at VU University Amsterdam, director of AI training programs for professionals, entrepreneur, and driving force behind various data and AI initiatives, he bridges the gap between theory and practice in clear, accessible language. He’ll be a keynote speaker again this year: Feldberg will be in the keynote hall on Wednesday at 12:00 p.m.
A Game-Changer
What changes has Feldberg observed in his field since the last edition of Data Expo? “I think the importance of AI has become clear to many more organizations. You no longer have to convince them that it can be a game-changer for their organization. Pretty much everyone is working on it, even if not everyone knows exactly what it is. Furthermore, there’s been much more focus on the economics of AI. After all, all investments have to pay for themselves.” The computing power required for AI comes at a price, he continues: “Many solutions were initially distributed around the world virtually free of charge. People started experimenting with them and may even have structured their work around them. Now, however, none of it is free anymore, and users are being presented with the bill. They have to start making choices, from tokenmaxxing to tokenmining. This is also in light of the many pilots that yielded no results.”
More than Just Prompts
According to Feldberg, AI extends far beyond what most people think, even in 2026: “They think of language models and chatbots, always framing it as generative, but it’s so much more. AI is embedded in many of an organization’s core processes; it’s much more than just prompts. If you think that, you’re really starting from the wrong place. A language model is just a tiny piece of the puzzle.” Last year, the professor spoke at Data Expo about the different types of AI and their limitations. “At the upcoming event, I’ll be talking much more about how to successfully scale AI.”
“A language model is just a tiny piece of the puzzle”
FOBO
The latter is the exception rather than the rule, as Feldberg knows from both theory and practice. Time and again, it turns out that the vast majority of AI initiatives never make it past the experimental phase. According to Feldberg, both scientific research and market research show that the majority of data and AI projects do not deliver the desired results. This brings him to another misconception about AI: “‘You really have to experiment with this new technology!’ people keep telling each other. ‘Because it’s all happening so fast.’ Add to that the fear of becoming obsolete (FOBO for short, ed.), and you have a recipe for going completely overboard.
AI adoption also involves all kinds of new work.”
According to Feldberg, experimentation is certainly part of the process—“I shared that as a takeaway with my audience just last year”—but it must have a clear goal and be conducted under specific conditions. Experimentation, however, seems to have become an end in itself. That’s why the professor speaks of an “AI pilot syndrome,” or “AI pilotitis.” “Pilots are popping up like mushrooms, without organizations knowing what value they’re supposed to deliver. On top of that, what’s successful in the lab isn’t automatically successful in day-to-day practice. The step from a pilot to implementation and adoption within the organization is fraught with major pitfalls.”
Change Management
In his keynote, Feldberg outlines the “missing middle” between isolated proofs of concept and successful AI scaling. In doing so, he helps organizations navigate the pitfalls that stand in the way of successful data and AI projects. “What I’m saying is based on fundamental principles that aren’t earth-shattering at all and are, in part, a matter of common sense. According to a well-known rule of thumb, the problem almost never lies in the technology: that requires only 10 percent of your time and investment. Data quality and infrastructure account for about 20 percent; 70 percent of the effort and investment, on the other hand, must go toward organizational change. That’s extremely important. “Yeah, Frans, we know that by now,” people tell me, “change management is essential for every project.” But I’ve tailored this specifically to the implementation and adoption of AI.”
Toward a Shared Narrative
Feldberg’s path to successful AI scaling has a clear starting point: a shared narrative about data- and AI-driven innovation. A narrative that isn’t based on “having to do something with AI,” he explains: “Why start with the ‘how’? Start with the ‘why’ and develop a narrative in which the ‘why,’ ‘what,’ and ‘how’ are described in relation to one another.” According to the professor, it is essential that business, IT, and data professionals develop such a narrative together. “This leads people to share knowledge, collaborate, align on goals, and so on, which ultimately results in a shared understanding. We’re currently being inundated by big tech companies and vendors with stories about the unprecedented possibilities of AI. Each party has its own interest in widely sharing these stories. But is that also in our best interest, and does it align with who we are as an organization? Are we just going along with the hype, or are we making a conscious choice? To be successful, the answer to this question is essential: what is our own story about AI? What is our narrative? That sounds very 1970s, but research consistently shows that organizations with a narrative surrounding an innovation are more effective and efficient.”
The question for organizations is: what is our story? "
The What and How of AI Scaling
In his keynote, Feldberg promises to discuss “the what and how of AI scaling” as part of an organization’s own AI story. “I’m going to give attendees concrete tools. Based on scientific literature, industry reports, practical experience, and artificial intelligence, I’ve developed the components that organizations can use to create such a shared narrative. When they leave the session, they can get started that very same day by asking the right questions to create their own narrative, use that as a foundation to set up pilots, and achieve successful adoption and implementation.”
Adoption across three dimensions
Feldberg will use the TOE framework, an organization-focused theoretical adoption model developed by Louis G. Tornatzky and Mitchell Fleischer in 1990. The letters stand for technology, organization,and environment. “Across these three areas, I discuss which factors influence adoption. When it comes to technology, for example, we’re talking about data quality and infrastructure; when it comes to organization, we’re talking about alignment and hidden work,among other things. Everyone keeps talking about AI as a boost to productivity, but it also involves all kinds of new work—in the areas of data, knowledge, and value. My fellow researcher Elmira van den Broek has conducted interesting research on this: the new, often hidden work that still needs to be done. I’ll delve deeper into this as well. Finally, “environment” covers issues such as regulation, ethics, and industry dynamics—in other words, the competitive context.”
Template for your own story
“A template with questions to help you develop your own AI story” is, according to Feldberg, an important takeaway for attendees of his presentation. “Because you can’t outsource a story like that.” In 2021, he developed the Data-Driven Innovation CANvas—or DatCan for short—together with Tom Pots. In his keynote, he’ll present an AI-specific successor to this toolkit, because AI goes beyond analytics. “DatCan is a tool for data projects that has already been downloaded nearly two thousand times and is therefore widely used in both the public and private sectors. At Data Expo, I’ll be presenting DatCan AI—a first.”
“You can’t outsource a story about data- and AI-driven innovation”
Frans Feldberg, who is, among other things, a professor of Data- and AI-Driven Business Innovation at the School of Business and Economics at Vrije Universiteit Amsterdam, is a keynote speaker at Data Expo. He will speak on Wednesday, September 9, at 12:00 p.m. onthe main stage. Title of his presentation: Beyond the AI Pilot Syndrome: Why 85% Fail and How You Can Scale Up Successfully.
Order your free tickets for Data Expo, taking place on Wednesday, September 9, and Thursday, September 10, at the Jaarbeurs ( ), here.
(Photo credit: Yenlo)
