AI Cannot Govern What No One Has Defined (EACOE)
A September 2026 CIO.com article, “How CIOs use AI to better manage data lifecycles,” captures a question that many executives are now asking: if AI needs high-quality data, can AI itself be used to produce that data? The article cites Gartner research indicating that 63 percent of organizations either do not have, or are unsure whether they have, the right data management practices for AI, and that organizations will abandon 60 percent of AI projects unsupported by AI-ready data. It then presents a set of senior practitioners from Volvo Cars, Harvey Nash, Sanofi, AOP Health, the DP World Tour, and Freshworks describing where AI is, and is not, helping them manage data.
The cited article identifies the right problem. It does not, however, resolve it. This research paper from EACOE makes three arguments.
First, data and information are not the same thing. Data is data: recorded values, symbols, signals, documents, and transactions. Information is processed data: data that has been organized, defined, related to other data, and placed in a business context so that it means something for a purpose. The cited article uses the two terms interchangeably, and that slippage hides the real work. Organizations do not fail at AI because they lack data. They fail because they have not converted their data into governed, defined, trustworthy information.
Second, the cited article’s central recommendation is circular. Its logic can be reduced to two sentences: AI needs good data, so use AI to get good data. A probabilistic system cannot establish the authoritative meaning, ownership, and quality thresholds against which its own outputs must be judged. Without a defined reference point of truth, AI applied to unmanaged data does not produce AI-ready data. It produces faster, larger, and more convincing versions of the original problem. The practitioners quoted in the article confirm this: every one of them who is succeeding with AI in data management is doing so on top of an existing, human-owned discipline of curation, definition, permissioning, or governance.
Third, the missing element is a method. The AI-Ready Data book, AI-Ready Data: What Leaders Must Get Right Before AI Works, and the AI-Ready Data Master Class supply that method. They teach organizations to start with meaning and then move to mechanics: define the things of interest to the business in an ontology, forward engineer that meaning into technology-neutral data models and governed data assets, organize delivery through a Data Distribution Center™, and only then apply AI to accelerate bounded data-management tasks under accountable human review.
The conclusion for CIOs and business leaders is direct. AI can be a powerful accelerator of data and information management. It is not a substitute for it. The sequence matters: business purpose and governance first, defined and managed information second, controlled AI assistance third, and human validation throughout.
