At Rise of AI 2026 in Berlin, a panel with Kenza Ait Si Abbou (founder and CEO of scailers, former CTO and board member of FIEGE), Claudia Pohlink (Director of Data and AI at FIEGE) and Dr. Ulli Waltinger (Siemens), who also moderates, discusses what it takes to move enterprise AI from pilots to competitive advantage. Pohlink argues vibe coding is excellent for experimentation but a liability for ownership in regulated, industrial settings, and describes chatting with 15 years of her own downloaded LinkedIn data as the moment she understood AI's practical potential. Ait Si Abbou defines scaling as embedding AI in corporate strategy so it changes how a company makes money, not accumulating long lists of unimplemented use-case ideas, and stresses that architecture, data management, governance and culture must be in place first. Waltinger cites Siemens' internal "everyday AI" rollout, SiemensGPT, with more than 190,000 users. The panel debates Europe's AI sovereignty gap, weak marketing of European AI achievements such as Aleph Alpha and Black Forest Labs, agent governance, the shrinking role of the Chief Data Officer, and closes with bold 2030 predictions on digital twins, automated governance and voice interaction.
Key takeaways
- Kenza Ait Si Abbou defines scaling AI as embedding it in corporate strategy so it changes how a company makes money, not accumulating unimplemented use-case lists.
- Claudia Pohlink says vibe coding is excellent for experimentation but a liability for ownership and explainability in regulated, industrial settings.
- Siemens' internal SiemensGPT tool has more than 190,000 users, cited by Ulli Waltinger as an example of Siemens' 'everyday AI' rollout.
- Both panelists say many companies have fired or renamed Chief Data Officers even though the orchestration function they performed is still needed.
- Kenza Ait Si Abbou argues European enterprises largely still avoid using available European AI solutions, undermining calls for tech sovereignty.
Detailed notes
Vibe coding: brilliant for experimentation, risky for ownership
Asked whether widespread "vibe coding" (a term the moderator traces to Andrej Karpathy's framing of English as the new programming language) marks a golden age or just generates messy "spaghetti data," Kenza Ait Si Abbou says the answer is both: it is brilliant for experimentation and helps people get results faster, but in regulated or industrial environments, where nobody can explain a system's behavior after the fact, for example during an unexplained warehouse blackout, it is a liability without ownership and explainability. She says her own top productivity habit since starting scailers alone is using Claude for back-office and delegated tasks while keeping strategic client work for herself.
What "scaling AI" actually means
Ait Si Abbou explains that her company's name reflects a common failure: companies proudly report hundreds of AI use-case ideas, but only a handful ever get implemented, let alone scaled. She defines scaling as adopting AI as part of corporate strategy in a way that changes how a company actually makes money, which requires a modular, scalable architecture, proper data management and governance, a defined operating model, and, most importantly, a culture of leadership, change readiness and psychological safety around failure. Once that maturity exists, she argues, the specific use case chosen matters far less than the organization's underlying readiness.
Building the data foundation
Claudia Pohlink describes her downloaded 15 years of LinkedIn data (a one-day export) as the personal "aha moment" that convinced her of AI's tangible value, and situates Siemens' internal SiemensGPT tool, cited by Waltinger as having more than 190,000 users, in the same category of hands-on exposure. For building a data organization, she describes a model learned from Deutsche Bahn of three pillars, data foundation and governance, AI use cases, and "3C" (communication, community, change), alongside FIEGE's own approach of separating foundational data platform work from solution-specific work and pushing data products such as dashboards out to the business units that use and fund them, who can also decide to terminate a product that is no longer needed.
Europe's AI sovereignty gap
Pohlink recalls Ait Si Abbou once telling her Germany has the talent, knowledge and infrastructure for AI but lacks marketing, citing Aleph Alpha, Black Forest Labs and SAP as underappreciated European achievements. Ait Si Abbou counters that European enterprises still will not use available European solutions, recalling how hard it was to sell Deutsche Telekom's sovereign Open Telekom Cloud years earlier, and attributes this partly to a missing "failure culture" in Germany and to public institutions that require prior references before working with startups, making it hard for new German solutions to gain traction. On the market hype around agentic AI and Anthropic's roughly 300 billion dollar valuation, Ait Si Abbou argues stock prices have become disconnected from company performance, calling the capital-markets reaction the real site of AI hype, while enterprise adoption itself is "still doing the homework."
Agent governance and the fading Chief Data Officer role
Pohlink argues the durable foundation for agents, multi-agent systems, or any future "AI as an operating system" is digitizing core master data and processes now, for example translating them into markdown, rather than chasing each new buzzword. Ait Si Abbou warns that as anyone can now build agents with tools like Copilot or Power Automate, organizations need the same kind of asset and access-rights oversight they once applied to IT service management, with distinct governance for human and technical (agent) users, or the result is chaos. Both agree many companies have fired or renamed their Chief Data Officers even though the underlying orchestration function, data stewardship, governance, quality, is still needed somewhere in the organization, since boards often do not see its value until years after the investment.
Q&A and 2030 predictions
Asked for bold, unhedged 2030 predictions: Pohlink expects personal digital twins/avatars of busy executives to become a commodity, and predicts data and AI governance will be fully automated and built in by design; Ait Si Abbou disagrees on both counts, calling personal digital twins an ethical risk and arguing governance remains inherently a human activity, and instead predicts human-machine interfaces may largely disappear as agents act directly, with voice interaction becoming far more common. Responding to an audience member's fear, drawn from the Sully Hudson River landing story, that professionals such as diagnosing doctors might one day be punished for not following AI recommendations, Ait Si Abbou says mistakes are inevitable for both humans and AI and must be learned from rather than avoided. Closing, Ait Si Abbou says clarity follows a decision rather than preceding it, and Pohlink urges dropping an overly cautious German mindset in favor of more American-style boldness in data and AI.
Quotes
Vibe coding is really brilliant for experimentation, but when it comes to ownership, it's a disaster.
Kenza Ait Si Abbou
In my experience clarity follows the decision, it doesn't precede it.
Kenza Ait Si Abbou
Let's throw away all that German mindset and let's be more brave and do something.
Claudia Pohlink
Questions answered in this talk
What does it actually mean to scale AI in an enterprise?
According to Kenza Ait Si Abbou at Rise of AI 2026, scaling means adopting AI as part of corporate strategy in a way that changes how a company makes money, which requires a scalable architecture, solid data management and governance, and a culture ready for change, not just a long list of pilot use cases.
Is vibe coding safe to use in enterprise and industrial settings?
Claudia Pohlink says at Rise of AI 2026 that vibe coding is brilliant for experimentation but risky for ownership in regulated or industrial environments, where nobody may be able to explain what a vibe-coded system is doing when something goes wrong.
Why is Europe struggling to build AI sovereignty despite having AI talent?
Kenza Ait Si Abbou argues at Rise of AI 2026 that Europe has the talent, knowledge and infrastructure but weak marketing and a limited 'failure culture,' and that even existing European solutions such as Aleph Alpha are often not adopted by European enterprises themselves.