Let’s talk about why the most advanced AI in your hospital might be the one keeping the doctor from actually seeing you.
Picture a physician in a public hospital in Pretoria. She has forty patients scheduled. The system gives her twelve minutes per consultation. But eight of those minutes go to typing notes, filling templates, selecting codes for billing, and ticking boxes that the hospital’s compliance system demands.
The remaining four minutes are what the patient gets.
Now the hospital installs an AI scribe. It listens to the conversation, generates the notes automatically, and promises to cut documentation time in half. The vendor calls it a productivity breakthrough. The administration approves the purchase. The doctor gets four minutes back.
That sounds like progress.
But here is what actually happens. The AI scribe does not reduce the documentation burden. It makes the documentation more thorough, and therefore more demanding. The billing codes get more precise, which means more claims are accepted. The compliance notes are more complete, which means fewer audit flags. The hospital sees better metrics. The vendors celebrate their ROI.
And the doctor still spends most of her time managing the record of the encounter instead of having the encounter.
The machine did not remove the burden. It optimised the burden. It made the system that steals the doctor’s time more efficient at stealing it.
This is the pattern that repeats across every industry that adopts AI for productivity: the machine speeds up the broken workflow instead of questioning whether the workflow should exist at all.
In healthcare, the cost is not just inefficiency. It is the slow disappearance of the human from the room. A generation of doctors is being trained to trust the machine’s notes more than their own memory. To read the AI summary before looking at the patient. To let the algorithm flag what matters instead of developing their own clinical instinct.
The system that was supposed to free them is quietly deskilling them.
There is a better path, but it is not the one the vendors are selling. The machine should not write better notes. It should make notes unnecessary for the parts that do not matter. It should handle the insurance codes, the referral letters, the prescription renewals, and the follow-up scheduling so that the doctor enters the room with nothing but attention and exits with nothing but a decision.
That requires a design philosophy that almost no healthcare AI vendor has adopted: make the machine invisible, not central. Make it the assistant that cleans up after the human interaction, not the interface through which the interaction happens.
The principle that must survive is simple. The doctor treats the human. The machine handles the burden around the room. Anything that reverses those roles is not progress. It is just faster extraction dressed as innovation.
You have got this.
⚠️ File-mutation verifier: 1 file(s) were NOT modified this turn despite any wording above that may suggest otherwise. Run git status or read_file to confirm.
• /dev/null — [write_file] Failed to write file: /bin/bash: line 2: /dev/.hermes-tmp.63179: Operation not permitted