Before AI, Agree on the Numbers
Organizations spent the past two decades learning how to manage data, but the rise of AI is revealing a deeper challenge: learning how to agree on what the data means.
A recent report from EAB, The Data Governance Imperative: How Colleges Can Address Seven Urgent Challenges Through a Shared Source of Truth, prompted me to revisit one of the most influential governance frameworks of the past two decades. At first glance, the two documents appear to address the same subject. Both focus on data governance. Both emphasize the importance of trusted information. Both seek to help organizations make better decisions.
Yet after reading the EAB report, I came away with a different impression. The two documents are not describing the same governance challenge. Rather, they represent different stages in the evolution of governance itself.
The IBM Data Governance Council Maturity Model emerged in 2007, when organizations were struggling to manage increasingly complex information environments. The EAB report arrives nearly twenty years later, in a world filled with dashboards, analytics platforms, enterprise systems, and artificial intelligence. The distance between those two moments reveals something important. The governance problem organizations face today is not quite the same problem they faced two decades ago.
The Roman Senate. Governance has always depended on more than information alone; it depends on agreement about what information means. Source: The Roman Senate’s 130-Year Afterlife, The Yeoman's Cry (Medium).
The Governance Problem We Thought We Had
When the IBM Data Governance Council introduced its maturity model in 2007, organizations faced a genuine information management crisis. Data was fragmented across systems, ownership was often unclear, and reporting practices varied significantly across departments. Regulatory requirements were becoming more demanding, and many organizations lacked the structures necessary to ensure consistency, accountability, and quality across their information assets.
The IBM framework provided a practical response. Governance became something that could be measured, assessed, and improved. Organizations could evaluate their maturity across areas such as stewardship, compliance, data quality, risk management, and value creation. Rather than treating governance as an abstract aspiration, the model established a roadmap that leaders could use to move from ad hoc practices toward more disciplined and integrated approaches.
The framework's influence was substantial. Many of the concepts that continue to appear in governance programs today can be traced back to IBM's work. Consulting firms adopted similar maturity models. Industry standards incorporated comparable terminology. Governance became a recognized management discipline rather than a niche technical concern.
The model succeeded because it addressed the challenges of its era. Organizations needed better ways to manage information. They needed clearer ownership, stronger controls, more consistent definitions, and improved accountability. The underlying assumption was straightforward: if organizations could govern information more effectively, they would be able to make better decisions.
That assumption remains largely correct. Better information generally leads to better decisions. Yet the environment surrounding information has changed dramatically since 2007.
Over the past two decades, colleges and universities have invested heavily in technology. Student information systems, customer relationship management platforms, learning management systems, advancement databases, financial systems, data warehouses, and business intelligence tools have transformed institutional operations. Information that once required days or weeks to assemble can now be accessed through dashboards in real time. Enrollment activity, retention trends, financial performance, and student outcomes can be monitored with a level of precision that previous generations of administrators could scarcely imagine.
In many respects, organizations solved the problem IBM sought to address. Information became more accessible, more comprehensive, and more manageable. Yet success in information management did not eliminate every obstacle to effective decision making. Instead, a different challenge emerged.
The Governance Problem We Actually Have
The most striking observation in the EAB report is not a recommendation or a framework. It is a diagnosis.
According to EAB, many institutions now possess "plenty of data but little confidence in it and few clear ways to understand what it says about how their colleges are operating." As a result, leaders often "debate whose data is correct instead of deciding what comes next" (EAB, p. 2).
That statement captures a challenge that feels increasingly familiar across higher education.
The problem is rarely the absence of information. Colleges and universities generate enormous quantities of data. Enrollment systems track student activity. Financial systems monitor institutional performance. Learning management systems capture academic engagement. Advancement systems record donor interactions. Dashboards and reports provide visibility into nearly every aspect of institutional operations.
Yet strategic conversations frequently stall because different parts of the institution rely upon different assumptions, definitions, and methodologies.
Consider a common question such as projected enrollment. Admissions may maintain one forecast, Finance another, and Institutional Research a third. Each forecast may be generated by capable professionals using sound methodology. Each may be technically defensible. Yet before leaders can discuss staffing levels, financial aid strategies, housing capacity, or course scheduling, they must first determine which forecast should serve as the basis for planning.
The challenge is not the absence of analysis. The challenge is establishing confidence in the analysis that already exists.
The same pattern appears throughout higher education. Retention rates vary because offices use different census dates. Program reviews draw upon different reporting sources. Public dashboards may not align perfectly with internal reports. Institutions possess data, but they do not always possess agreement.
Viewed through this lens, EAB's emphasis on creating a "single source of truth" becomes especially significant. Governance is no longer simply about managing information assets. Increasingly, it is about creating organizational alignment around information. The objective is not merely better data quality. The objective is reducing the friction that emerges when different parts of an institution operate from different assumptions about reality.
That distinction may seem subtle, but it represents an important evolution in the purpose of governance. IBM's framework focused on establishing structures that would improve information management. EAB focuses on establishing trust that enables effective decision making. The first challenge concerned the stewardship of data. The second concerns agreement about data.
Conclusion: Artificial Intelligence Makes the Difference Impossible to Ignore
Artificial intelligence has amplified the importance of this shift.
Much of the current discussion surrounding AI focuses on technological capability. Institutions are exploring predictive analytics, conversational assistants, retrieval-augmented generation systems, automated reporting, and AI-supported decision making. The assumption often seems to be that better technology will produce better answers.
The EAB report offers a useful warning. "AI can amplify weak data rather than fix it" (EAB, p. 16).
Artificial intelligence systems depend upon trusted definitions, reliable data, and authoritative sources. A chatbot connected to institutional documents can retrieve information quickly, but it cannot determine which of several competing definitions reflects institutional intent. A predictive model can identify patterns, but it cannot resolve disagreements about how key variables were defined. An AI assistant can summarize information efficiently, but it cannot establish consensus where none exists.
Many institutions are discovering that the most difficult aspect of AI adoption is not the technology itself. Building a retrieval-augmented generation system or deploying a conversational assistant has become increasingly accessible. Determining which information should be treated as authoritative remains far more difficult. Before an AI system can answer questions about enrollment, retention, graduation rates, workforce outcomes, or student success, institutions must first decide which definitions, calculations, and business rules represent institutional truth.
Seen in that light, IBM and EAB are not describing competing visions of governance. They are describing different stages in its evolution. IBM addressed the challenge of managing information in an increasingly digital world. EAB addresses the challenge of creating organizational agreement in an era defined by information abundance and emerging artificial intelligence.
The distinction matters because the promise of data has always been better decisions. Artificial intelligence promises to accelerate those decisions. Neither promise can be fully realized unless institutions first establish enough trust in their information to act upon it. Governance, once viewed primarily as a discipline of information management, may increasingly be understood as a discipline of organizational agreement.
Before institutions can teach machines what they know, they must first decide what they know themselves.
Further Reading
- IBM Data Governance Council Maturity Model (2007)
- EAB, The Data Governance Imperative: How Colleges Can Address Seven Urgent Challenges Through a Shared Source of Truth (2026)
- EDUCAUSE, 2025 AI Landscape Study: Into the Digital AI Divide (2025)