AI applications rely on data. Not on the data you imagine you have, but on the data as it is actually stored, defined, and maintained. Data management is therefore one of the fundamental dimensions within the maturity assessment: together with organization and IT infrastructure, it determines whether there is anything to build on. An organization can score strongly on people and skills or on performance management and still find that every AI application grinds to a halt because no one can say with certainty which version of a customer file is the correct one. That order is not a choice we made; it is the order in which it works.
Data management concerns the state of the data an organization uses to work and decide. Are definitions fixed, or does "active customer" mean something different in every department. Is there an identifiable owner per dataset, or does no one know who is responsible when a field has been wrong for years. Is data captured once and drawn from everywhere, or do three exports exist that have since diverged from each other. The questions in the assessment are not about the amount of data or the type of system, but about the reliability and provenance of what exists.
The five levels run from baseline to intelligence, and for data management the difference between those levels is often more concretely noticeable than for other dimensions. At baseline, data exists mainly locally: in spreadsheets, in a colleague's head, in a system no one fully understands anymore. At foundation, the most important datasets have been identified and there is a beginning of ownership, even though definitions still diverge. What that foundation level means in practice, including for other dimensions, is described at what the foundation level means in practice, and the same applies to the starting point itself at what the baseline level means in practice. At activation, definitions are largely shared and there is a process to flag deviations. At insight, data is actively used to underpin decisions, with visibility into quality and provenance. At intelligence, data management is no longer a separate track but a fixed characteristic of how the organization works.
The plotting round reveals whether that assessment is accurate. When the CIO places data management at activation and the teams working with it daily place it at baseline, that spread says more about the actual situation than the average of both scores. It often turns out that the systems have been set up at a higher level than the behavior within them: there is a central system, but people still work alongside it in their own files because they don't trust the system.
What it costs to move from baseline to foundation depends on how fragmented the data currently is and how many departments maintain their own version of the truth. The work then lies mainly in assigning ownership and recording definitions, not in new technology. Moving from foundation to activation more often requires a technical step: designating one source, discontinuing others or making them read-only. Moving from activation to insight shifts the weight toward governance: who may change what, how quality is monitored, what happens when an error is found. How long and how heavy that process is depends on the size of the organization, the number of systems that maintain data, and the extent to which people are already accustomed to treating one source as authoritative. That cannot be captured in a fixed amount or a fixed timeline; it depends on where the spread in your plotting round concentrates.
Data management rarely stands on its own. An organization that does not trust its data will also struggle with what it costs to raise privacy and security by one level, because you cannot secure what you have not mapped. Conversely, strong data management can lay a foundation that makes what it costs to raise performance management by one level easier to answer, since performance management needs reliable data to be able to measure anything. The question of what it costs to raise ethics by one level also directly touches data management: handling data responsibly starts with knowing what you have and where it comes from. Anyone who addresses these dimensions in isolation risks having work on one dimension held back by a shortfall in another.
This assessment shows whether the organization has the data, the systems, and the people to support AI. That is a different question from which part of the work AI can actually take over. Once it is clear where the fundamental dimensions stand, including data management, it becomes meaningful to examine per task what AI can and cannot do. That is the domain of the work scan from FTE TO AI: it calculates per task which part of the work can be taken over, based on the data and systems that are actually in place at that moment.
Vraag maar wat er moet staan voordat AI in uw organisatie kan landen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.