Not: do we have a lot of data. Not: do we have a data strategy on paper. The question that counts is narrower and more uncomfortable: is the data belonging to a specific task complete, current, and accessible enough to hand that task over to AI, entirely or with oversight.
That is a different question per task. A task where AI summarises text from a well-structured database requires something different than a task where AI prepares a decision based on data that is recorded differently in three systems. The ambition often only states the first case out loud and assumes the second.
AI is taking over work. For some tasks that can happen fully, for some only with human oversight that approves or rejects with reason, and for some it remains human work. That division runs across every department and does not change because the year changes, but because the underlying data changes.
The difference between companies that already test this well and companies that do not yet do so rarely lies in the amount of data. It lies in whether someone can point to which data belongs to which task, who owns it, and how you would notice that data is wrong. Companies where that question is a habit tend to see fairly quickly, per task, whether AI takeover is a real conversation. Companies where that question is new only discover the data gaps at the moment AI has already produced something wrong.
The test does not ask for a data classification document. It asks for three things you can point to, not describe:
Evidence that does count: a sample where three people, independently of each other, consult the same data source and arrive at the same outcome. Evidence that does not count: a satisfaction score about "our data infrastructure" at a leadership offsite.
A task that scores low on the data dimension is not by definition a task that frees up no capacity. It means the order changes: first get the data in order, then the handover. That is a timeline, not a no.
It also does not mean something is wrong with the people currently managing the data. The test measures readiness of the work, not the quality of employees, and says nothing about roles or numbers. If an outcome touches on personnel decisions, its own legal requirements apply; you assess those separately, not on the basis of a readiness test.
What is useful: the outcome shows whether the problem lies with the data itself, with the people working with it, or with the suppliers delivering the systems. That distinction determines the next step. If it concerns how employees handle data, that touches on how you test whether your people can handle it. If it concerns third-party systems that deliver or unlock the data, how you test whether your suppliers can keep up comes into play. And if the ambition itself had not named the data need, the cause often lies more in the wording than in the data: then how you make a vague vision concrete helps.
Behind "is our data good enough" lies a bigger question: which work in this company, given the data currently available, can truly be taken over by AI, and which cannot. That question is answered per task with the work scan from FTE TO AI, not with a judgement about the organisation as a whole.
You can test the data dimension on its own, but it does not stand alone. Readiness for AI takeover consists of eight dimensions, and data is one of them alongside, for example, people, suppliers, and the sharpness of the ambition itself. An outcome on one dimension says little without the other seven.
The free readiness check provides that overview in eight short questions, one per dimension, with a picture of where you are furthest along and where you are least far along. No report, no obligation, but an honest starting point. The full ambition test, with the four layers from vision to target state and the translation to roles and decision rights, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.