Sonia starts her morning with a queue of appointment requests, each one carrying a referral order a patient uploaded, a photo, sometimes handwritten.
She doesn’t read them cold. The AI has already read every order overnight: is the date valid, is the doctor’s licence current, does the name on the order match the patient’s account. Handwriting included.
For each request, the AI proposes an action, accept or reject, with the reason attached. Sonia reviews the proposals, not the raw images.
She rejects two. One order is expired; one name doesn’t match. She approves the rest.
The rejected patients don’t hit a dead end. They get a notification with the next step already attached, a link to re-upload a corrected order, or the right service with its next free slots.
The approved patients get a reminder the day before their visit. They answer YES or NO, confirming or freeing the slot for someone else on the waitlist.
Nothing here replaces Sonia’s judgment; she still makes the final call on every request. What changed is what she spends her morning on: not deciphering handwriting and checking licence numbers, but deciding.
The waitlist means no freed slot sits empty, and every rejection tells the patient exactly what to do next.