How much time goes to admin, and what can software take over?

By Joost van Houten · August 26, 2026 · 6 min read

Channels in a sandbank at low tide, seen from above

Medical specialists estimate that they spend more than a third of their working time on administration. In the largest survey of specialists, from 2017 with 2,449 participants, the average was 38%, and the specialists felt most of it took too long.¹ Software and AI can take over part of it, but the evidence is more modest than many vendors promise.

The government wants to almost halve the adminA medical specialist's working week; the dark part goes to admin

Now38%, almost 2 days2017 survey

Target20%, 1 dayAZWA, from 2030

Source: Het Roer Moet Om and VvAA (2017), see footnote 1; Supplementary Health and Welfare Agreement (AZWA, 2025).

I come from the software world. At Seeve we have looked at how outpatient work is organised, and where technology really makes a difference. I think the biggest gain lies in a care process designed around the software from the start. This article covers what we see.

The administrative burden of care

Talk to specialists and it is rarely about one big task. It is many small actions around each consultation that together fill the day:

Documentation. Recording the consultation in the record, often afterwards or at the end of the day.
Correspondence. Letters to the GP and referrers, and answering patient questions through email or the portal.
Orders and referrals. Lab tests, imaging, other specialties, medical devices.
Coding and billing. Recording diagnosis and procedures correctly so the claim is right.
Authorisations and statements. Physician statements for expensive drugs, such as CGRP inhibitors for migraine.
Quality registries. Data requested by professional bodies or insurers.
Scheduling. Booking follow-ups at the right moment and making sure results are ready by then.

Much of this exists because information is needed in several places (within one system and across different systems) and in several formats. What goes into the note must also end up in the letter, the claim and the registry. In a general EHR that has to serve every specialty at once, many of these tasks are often still done by hand. For a specialist running their own clinic, the list of administrative tasks is much longer.

What software and AI can already take over

In healthcare, AI that listens in during the consultation and drafts the note is a well-known use case. There is now also good research on its impact. In a study at UCLA Health in 2025, 238 physicians across fourteen specialties worked with or without such a tool for two months. With the best tool, time per note fell 9.5% more than in the control group, from four and a half to under four minutes. An alternative tool showed no significant difference. Physicians in both AI groups did report somewhat fewer burnout symptoms. And importantly: they reported that the notes sometimes contained clinically relevant errors, in most cases omitted information. I find that an honest picture of this use case. An AI scribe, a relatively small component, on top of an existing process saves minutes, not hours.

There is, however, a much bigger gain to be had by looking holistically at the process around it. In a clinic focused on one condition, with its own modern EHR, it would look like this:

Administrative task	What software (and in some cases AI) does	What the doctor does
Intake	Questionnaire in advance, summary of history and medication	Reads the summary, asks focused questions
Documentation	Draft note in a fixed structure per care pathway	Checks, edits, signs off
Letter to the GP	Draft from the note	Approves
Coding and billing	Derives codes from the pathway and the note	Checks exceptions
Scheduling	Books follow-ups by protocol and orders tests in advance	Overrides where needed
Patient questions	Answers practical questions, queues medical ones	Answers medical questions
Quality registry	Fills in automatically from structured data	Nothing

The difference with a general EHR is that the care pathway for a single condition is largely fixed, and in some cases can be adjusted to make it more efficient. The modern, specific EHR software then knows what belongs in a note, which code fits and when the next check-up is due. Any task you can standardise this way, you can largely automate, without errors.

What this means for a clinic is easiest to see if you follow the money. Below is a simplified breakdown of €100 in revenue at a hospital outpatient clinic, next to what it looks like at a focus clinic where software does most of the admin. It is our own estimate, meant to show the order of magnitude.

At a hospital outpatient clinic, an estimated €15 of every €100 in revenue goes to adminIndicative breakdown of €100 in revenue

Hospital outpatient clinic

Focus clinic with its own software

Source: Seeve estimate (2026). Indicative; the parts differ by hospital, clinic and specialty.

The €15 of admin and the hospital overhead do not disappear by themselves. In a focus clinic that money goes to software, to a doctor who can see more patients, and to what the clinic keeps to invest.

What stays with the doctor

The doctor makes the diagnosis, chooses the treatment, talks with the patient and signs off. Software drafts and handles routine work; responsibility does not move. I think it should stay that way, for three reasons.

Errors. Even in the UCLA trial, notes sometimes contained clinically relevant mistakes. A doctor who approves without reading makes the system less safe.
Privacy. Recording and processing a consultation requires patient consent, agreements on where the data is stored, and strong security. That has to be in place before you start.
Regulation. Software that supports clinical decisions can fall under the EU Medical Device Regulation (MDR). A draft note is different from a treatment recommendation, and that line has to be drawn deliberately.

For a specialist thinking about starting a clinic, this may be the most important point. Administration does not disappear when you leave the hospital. In your own clinic you take on extra tasks, such as billing, staff and compliance. Doing that well takes systems built around the care pathway from the start. That is what we are building at Seeve. I am happy to hear questions or ideas at joost@seeve.ai.

Footnotes

¹ A caveat on that 38% figure. It comes from a survey by Het Roer Moet Om and VvAA, and the sample was not shown to be representative. A later measurement (the (Ont)RegelMonitor from 2021) only looked at pointless rules and came out lower. So there are no precise figures. The direction is clear, though, and in the Supplementary Health and Welfare Agreement (AZWA) the government agreed that from 2030 administration may take up no more than 20% of working time.