Think back to the last time you saw a doctor while you were sick.
In most cases, the diagnosis took a minute or two, the prescription took maybe another minute and you’d be out in 5 mins max.
If you reflect back, you’d realize, half the conversation would just be about reassurance. (I often tell my husband that this is what we pay the doctor for)
Lately, people show up already holding the diagnosis.
Mostly, along with some idea of the treatment as well.
A mother typed her son's 3 years of medical history into ChatGPT, and it came back with the diagnosis 17 doctors had missed.
Another woman was told she had rheumatoid arthritis.
She asked ChatGPT instead, it suggested Hashimoto's, and an ultrasound 2 months later found thyroid cancer.
Nobody who had access to doctors talked to them through a screen before March 2020.
Then, within 3 months, 50 million Indians did, 8 in 10 of them for the first time, and Practo saw a 500% increase in online consultations.
COVID broke the assumption that you had to be in the room.
Each step felt unimaginable until it happened, but each removed a reason to be in the room.
For most of medical history, the doctor walked in holding three things you didn't have.
He had the readings from your body, the training to make sense of them, and the confidence to say “this is fine, come back if it gets worse.”
The first to slip away from him is the readings.
When Stanford ran the Apple Heart Study across 419,000 people, they found that a watch flagging an irregular pulse was right about 84% of the time.
That is a stroke risk caught by a $400 device.
The same watch also sends plenty of healthy people into a panic over nothing, pretty much every day, which is the cost nobody prints on the box.
The training is still theirs, though I'm less sure for how much longer.
The same profession that warned patients off self-diagnosis has let AI into the room itself, albeit mostly for the paperwork so far.
Though most of that is documentation, billing notes and discharge summaries, rather than diagnosis.
The profession that spent a decade telling patients to stay off Google is no longer arguing about whether AI belongs near a diagnosis, only about how much of it should.
And it's not hard to see why.
When a Nature study in April put Google's medical AI against 20 doctors across 159 simulated cases, it scored higher on 30 of the 32 measures, including empathy.
It doesn't care, which is exactly why it scored higher.
It has unlimited time, no patient waiting outside, and no irritation at the obvious question asked for the 8th time that day.
The third thing, the confidence to say this is fine, is where I’m stuck.
For most routine visits, that confidence is the whole reason the patient came in. And these days, you've had it since the night before anyway.
The mother, with the 17 doctors sat at her kitchen table and worked through 3 years of records at her own pace, despite the diagnosis having been in that file the whole time.
No appointment was ever built to sit with one patient that long.
And the usual objection I see is that you can't trust AI with your health, but trust was never the thing holding this back.
People follow through when they feel they have some say in what's happening to them.
A waiting room and a 5-minute slot were never going to give them that.
A phone at 2 am does.
For the routine visit, a lot of what the doctor was giving you alongside the diagnosis was permission to stop worrying, and that half is the part a phone can now hand you at 2am.
The first generation of medical AI was mostly fraud.
Babylon Health, once valued at $4.2 billion, went bankrupt in 2023.
A BMJ study this year still found about half of medical chatbot answers problematic, 1 in 5 seriously so.
The medicine that needs a body in a room hasn't moved at all, like the physical exam, the emergency, or the chronic case managed across years.
What moved was the permission.
And none of this is exclusive to healthcare.
Each is applying expertise that took years to build, and each is also handing over permission to stop worrying.
The client has been paying for both as if they were one thing.
But that bundle is coming apart.
So if you run a service business, look back at your last 20 client interactions.
How many needed your expertise to solve something the client couldn't have solved alone?
And how many were a smart, anxious person who needed someone qualified to say yes, that's right, you can stop now?
If that second number is more than half, you've got about 18 months.
And if you’re on the paying side, you’ve probably run this experiment already without naming it.
The next time you walk into an appointment already holding an answer you found the night before, sit with the gap between what you arrived with and what the doctor adds on top.
That gap is what you’ve been paying for.
If that gap is small, the trade you've been making is about to come undone, in healthcare first and everywhere else right behind it.
If you've read this far, reply and tell me what permission you've stopped paying for. I want to see what this looks like in other industries.
Nandini