In healthcare, much of the time goes into work that directly involves the patient or client, and into work that surrounds it: record-keeping, documentation, handovers, planning, accountability to insurers and regulators. That second category often takes up a large part of office and care hours, without anyone experiencing it as "administration" until it is measured. In addition, there are regulations on record-keeping obligations, medical responsibility, and data protection that are non-negotiable. That makes the outcome of an AI ambition in this sector different from that in a sector without those regulations: not slower by definition, but dependent on where responsibility for a decision ultimately rests.
AI is already taking over parts of the work today, not as a future vision but as something already running in some teams and not yet in others. Three categories run through almost every care process. AI can perform some tasks independently: summarizing a long record, generating a first draft letter, flagging an abnormal value in a series of measurements. Other tasks require human oversight that approves or rejects with reasoning: a draft diagnosis, a proposed treatment plan, a risk assessment for a client. And a third category remains human work: the conversation with the patient, the judgment call in a borderline case, the responsibility that ultimately lies with a practitioner and not with a system.
The difference between organizations that already make use of this and organizations not yet ready for it rarely lies in the technology. It lies in whether there is a process that can receive an AI proposal, assess it, and reject it with reasoning, and whether someone is authorized to make that decision. Where that process does not exist, an AI ambition remains an intention on paper, even if the software has already been purchased.
A commonly heard ambition in healthcare is that AI will reduce the administrative burden so that more time is left for care. That is a clear statement, but it presupposes capabilities that are not automatically present: a record structure that AI can read, a quality process that can check a draft report before it becomes final, and practitioners who know when to reject an AI suggestion rather than adopt it. If any one of those three is missing, the ambition will not free up hours, no matter how much budget goes into the project.
The same pattern occurs with ambitions around triage or planning: AI can make an initial assessment of urgency or capacity, but who checks that assessment, and on what information may that person deviate from it? Without an answer to that question, an appearance of acceleration arises while responsibility remains unmarked.
In a management team, one person says "AI-supported" and means that a system makes a suggestion that a nurse always reviews. Another means that the system already handles most of the input itself. Both pictures can be correct, but not at the same time for the same ambition. That difference becomes visible as soon as a vision is translated into a concrete situation on a concrete ward; how that works is described on a page about making a vague vision concrete. Once that translation has been made, the next question often arises: who will be authorized to decide that an AI proposal is adopted without separate review, and who must remain involved? That question about decision rights is addressed on the page about decision rights in an organizational change.
Similar points of tension also occur outside healthcare, though with a somewhat different nature of regulation: in professional services, for example, it more often concerns professional/technical responsibility than patient safety, as described on the page about AI ambitions in professional services. The underlying question is always the same: which work in this specific organization can truly be taken over by AI, and which work only in theory. That question is answered by the FTE TO AI work scan per task, not at sector level but at the level of the organization's own process.
An ambition such as "AI-supported care" only becomes manageable once it has been broken down into what the system may do independently, what it may propose under oversight, and what remains with the practitioner. As long as that distinction has not been made, a management team is using the same words to talk about different plans. A first step in this is the free readiness check: eight short questions, one per dimension, that show where the organization is furthest along and where it is not. The full ambition assessment, with the four layers from vision to target state and the link to roles and decision rights, is currently under development.
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.