An AI project that gets stuck at IT is usually not a technical problem. It is a project that never got an owner outside IT. IT builds a model, a pipeline, an integration — and waits for someone to say which work will now run differently. If that person is not there, or does not have the mandate, the project stays where it was built.
In parts of every company, work is already shifting. Not as an announcement, but as a practice: part of the work can be done independently by a system, part can be done by a system as long as a human approves or rejects the outcome with reason, and part remains human work because judgment, responsibility, or contact with a customer cannot be separated from it. This threefold division runs through virtually every process — not per department, but per task.
The difference between a company where this works and a company where the project gets stuck at IT is rarely found in the technology. It lies in who has determined which tasks fall into which category, and who has the authority to say: this work shifts, and this stays as it is. Without that determination, IT builds something that no one dares to put into production, because no one has made the decision that the decision requires.
An AI project that gets stuck at IT is usually missing one of three things:
An ambition that goes beyond a pilot. If the question "can we build this" has been answered but the question "which work changes structurally as a result" has not, there is nothing to scale up. Which ambition is better postponed for a year addresses the distinction between an ambition that is achievable now and an ambition that first requires something else.
An answer to what a system may and may not decide without oversight. That answer does not lie with IT, but with the department that performs the work and with compliance. What do you do when the ambition is feasible but compliance is not describes what happens when the technology is ready before the frameworks are.
Decision rights outside IT. Someone in the line — not in the architecture — must be able to say that a task shifts and that the freed-up capacity is deployed elsewhere. As long as that decision rests with IT, the project will remain there too, because IT does not have that mandate and should not have it either.
An ambition assessment establishes what the ambition is at four levels — the vision, the desired state, the concrete target, and the state that target requires — and tests per dimension whether the organization is ready for it. This is not a measurement of what a system can technically do. It is an estimate of what the organization still needs to arrange before work actually shifts: decision rights, oversight, data, responsibility.
That estimate is never definitive. It is only as good as the information on which it is based, and that information is often incomplete at the moment a management team starts with it. A dimension that appears ready because no one has asked the right question is not ready — at that moment, that outcome mainly says something about what has not yet been investigated. Conversely, a dimension may appear poor while the problem is small as soon as someone is given the authority to solve it. The assessment orders where the organization stands; it does not replace investigation into the tasks themselves.
The underlying question — which work in this company can genuinely be taken over by AI, partly or fully — is answered by the work scan of FTE TO AI per task, not per department or role. That distinction is necessary because a project that is assessed at the role level always gets stuck at IT: roles do not change, tasks do.
A single AI project that gets stuck is usually a symptom. Often there is no ongoing picture of how the eight readiness dimensions develop over time, and the same dimension remains the bottleneck year after year without anyone naming it — see what do you do with a dimension that has been stuck for years. An ambition that looks small on paper often requires capabilities that only become visible once someone maps them against the eight dimensions; that translation is central to which capabilities does my organization need for this ambition. And a project does not get stuck at IT if management is already at odds about what the ambition means — that is what how do I get my management team aligned is about.
Regarding personnel decisions that follow from such a process, the organization's own legal requirements always apply; an assessment of ambition and readiness is not a basis for such decisions and is not intended as such.
The free readiness check consists of eight short questions, one per dimension, and provides an initial picture of where the organization is furthest along and where it is not. It is not a judgment and not an endpoint, but a starting point for having the conversation outside IT before another project is built that no one is waiting for. The full ambition assessment, with the four layers and the five confidence gates, 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.