Stijn Christiaens

Stijn Christiaens returns again and again to a single correction: people say "data governance" but mean something more precise, and getting that precision right is the difference between organizations that scale AI and the ones stuck at pilot stage. He is blunt about the semantics: "The term 'data governance' is frequently misunderstood. It does not govern data itself. Instead, it governs the management of data," and he pushes the point further, noting "we can only govern the people that do something with data" . That reframing, governance as a discipline about people and processes rather than about data as an inert object, runs underneath his commentary on the changing role of the Chief Data Officer, whom he sees shifting into a curator of context: "CDOs who can build and manage this infrastructure of context transform their businesses" . He extends this into a broader argument about meaning itself being something organizations must own rather than outsource, flagging work on "Why You Cannot Rent the Meaning Layer" as strategically important thinking .

A second, closely related theme is that governance can no longer stop at storage. Citing McKinsey's finding that "only 7% of companies have fully scaled AI across their organizations," he argues the real bottleneck is data, and that "AI governance can no longer stop at the storage layer. It has to extend to retrieval, assembly and generation, to every point where data influences an AI output" . For him this means unstructured content, documents, contracts, transcripts, emails, has to be treated as a governed data product, with lineage, metadata, quality controls and semantic consistency, not an afterthought bolted on once a model is already in production . This is not a new insight he arrived at recently; he frames it as a long-standing conviction, "Collibra's north star since day one," with only the urgency having changed as AI amplifies and spreads data quality problems faster than governance models can catch . In interviews he sharpens this into a definition of "AI ready data" that spans both the classic quality requirements for training models and the much larger, historically neglected task of preparing the 80-90% of organizational data that is unstructured, using semantic layers to turn documents into usable knowledge products ¹⁷. He is equally direct about the misconceptions leaders bring to this work, chiefly the belief that AI requires no data investment and the assumption that AI systems are already reliable rather than still emerging ¹⁷.

Christiaens is skeptical of speed as a virtue in itself. "The mistake is to confuse speed with judgment" is a line he elevates approvingly , and it connects to a recurring worry about what happens to human capability when AI does more of the thinking. Recounting a question from an audience at his alma mater about skill atrophy, he draws the analogy to aviation, where policy forces pilots to manually perform a share of tasks so core skills do not disappear, and endorses a similar prescription for knowledge work: protect time for unstructured thinking before consulting AI, build in checkpoints where engineers, product managers and architects verify AI output in their own domain, and require teams to show their work, prompts, edits and citations checked, before anything ships . This is paired with a harder-nosed view of where AI value actually gets captured. Quoting the argument that "the scarce input in most of the 'messy work' required to implement autonomous AI is not IQ, but the workflow redesign, organizational authority, proprietary data, and the political ability to overcome resistance" , he signals that the bottleneck to AI payoff is organizational and political, not technical.

He is attentive to the physical and geopolitical infrastructure underneath all of this. He highlights research showing AI agents can consume "up to 136.5 times more energy per query than conventional generative AI" as an explanation for exploding token costs , and he flags the gap between stated intent and actual investment in sovereignty, noting that "sixty percent of respondents said that rising geopolitical risk makes them more likely to pursue sovereign technology solutions, yet only 15% have made AI sovereignty a CEO or board-level priority, and fewer than 13% see it as a growth driver rather than a cost" . On enterprise architecture he endorses the view that heterogeneity is not a transitional headache but a permanent fact: "No real enterprise runs on one model, one cloud, one architecture... 50 Shades of Hybrid... is the permanent condition, not a phase" . He extends the same systems-level thinking to Europe's competitive position, pointing to fragmentation, speed and depth of capital as the real constraint rather than any deficit of ambition .

On monetization, his takeaways are practical and procedural rather than aspirational: manage data assets with a product mindset, treat data monetization as a team sport, and realize the value that is created rather than simply generate it . This mirrors his long-held conviction, voiced in interviews, that organizations should treat data as a valuable asset rather than exhaust, and that trustworthy data products are the precondition for everything downstream, including AI that he expects to become more powerful and commoditized even as it demands more data, not less, to function well ¹⁶. Across his commentary on organizational failure, he keeps returning to the same diagnosis: the recurring mistakes with new technology, from on-premise systems to cloud data warehouses, are not technology problems but people problems, and cross-functional collaboration is the real barrier ¹⁶.

Underneath the governance commentary is a founder's voice shaped by pivots and category creation, and he is candid about how untidy that process actually is. He resists the romantic startup narrative directly: "The romantic notion of the entrepreneur is a little bit of an illusion, but I do recommend start, right? So people start" ¹⁹, and warns that