Healthcare organisations combine three types of work that don't mix easily: direct contact with patients or clients, physical actions that no one can perform remotely, and a paper or digital administrative burden around it that often takes just as much time as the care itself. In this type of organisation, the largest part of the non-direct hours goes into record-keeping, reporting, planning and coordination between care providers, administration towards insurers and regulators, and communication with patients and family. In addition, in most healthcare institutions multiple systems run alongside each other: an electronic patient record, a planning tool, separate software for claims, and separate communication channels per department. This fragmentation strongly determines whether a task qualifies for AI takeover, because a task that is simple in itself becomes more difficult once it has to pass through three systems.
On top of that, healthcare is one of the most heavily regulated sectors. Record-keeping obligations, retention periods, medical-substantive responsibility and consent requirements are laid down by law. This does not mean AI cannot play a role here, but it does mean the role is often supportive: preparing, summarising, flagging, rather than deciding independently. Physical care tasks — washing, nursing, injecting, physical examination — fall by definition outside what AI can take over. In this sector, it is therefore mainly about the hours around the care: the administration, the planning and the communication that surround it.
A number of tasks in healthcare are largely administrative and follow a fixed pattern, which means AI can take over most of them. Think of drafting structured reports based on fixed templates, summarising long records into a core overview for a handover, pre-sorting incoming messages in an inbox, and scheduling appointments according to fixed availability. Initial triage of questions coming in via email or form can also largely be automated: is a question urgent, is it about a repeat prescription, does something need to be forwarded to a specific department. For example, triaging a general email inbox in a healthcare organisation can pre-sort a substantial part of incoming mail before an employee sees it. How many hours this concretely frees up depends on the number of messages, the degree of standardisation and the systems that are linked to each other.
The largest group of tasks in healthcare falls in the middle: AI can deliver a first draft, a proposal or a signal, but a human assesses the result and approves or rejects it, for a reason. This applies, for example, to drafting concept reports that a care provider reviews before they go into the record, to flagging deviations in a schedule that a coordinator subsequently assesses, and to monitoring progress in long-running care pathways or quality projects. Here too, a tool that can take over monitoring progress of goals and projects can be useful as a signalling layer, as long as the final assessment remains with a human. Telephone contact with patients is another borderline case: a system that can take over answering the phone and transferring calls can handle simple routing questions, but when there is doubt about the nature of a complaint or in emotionally charged conversations, transferring to a human remains necessary. In this category, the gain lies not in removing the assessment, but in speeding up the preparation for it.
A substantial part of the work in healthcare remains, regardless of technological development, human work. This unmistakably applies to physical care actions, but also to conversations in which bad news is delivered, to complex multidisciplinary considerations in which different treatment options are weighed against each other, and to situations in which trust and presence are just as important as the content of the conversation. Final responsibility for a diagnosis or treatment plan also lies with the care provider and does not shift along with the technology. This is a structural feature of the sector, not a temporary limitation of the current generation of AI tools.
An overview of tasks that AI can take over, partly take over or cannot take over, is not personnel advice and not a basis for decisions about positions or staffing levels. Such decisions are subject to their own legal requirements, for example around works council involvement and labour law, which are not addressed here. This page describes tasks, not people or positions.
The relationships described here also do not apply only to healthcare. Anyone wishing to compare what this kind of task division looks like in another sector can look, for example, at what AI can take over in wholesale or at what AI can take over in manufacturing, where the balance between physical work, administration and regulation is different again.
Whether and how much AI can take over in your own organisation depends on which tasks occur most in your case, how your systems are connected to each other and which rules apply within your specific healthcare domain. A general sector picture such as the one above provides direction, but not an exact outcome for your situation. To get an initial indication of that, ftetoai offers a free quickscan of twelve questions, without an account, which gives an indication of what part of the hours in your profile can be taken over by AI today. The full work scan, which goes deeper into individual tasks, is still under construction; it does not yet exist at this time, and is therefore not offered here.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.