Last Wednesday, I forgot one crucial question in the hospital corridor
I was accompanying my mom to a follow-up appointment. My paper bag held test reports, lab results, and records from several hospitals. When the nurse called my name, my first instinct was not to panic but to pull out my phone and check the questions I’d written for the doctor—only to realize I had still missed one. That evening, I stared blankly at a dozen sheets of paper. I had the records, but nobody had connected them. I’ve been there too: I kept assuming that saving every file was the same as organizing the care.
Then I tried using ChatGPT to create a “care brief.” I started with the patient’s situation, confirmed diagnoses, and most recent follow-up date, then pasted in the medical-record text and had it organize everything into five sections: patient information, questions for the next visit, confirmed facts, plain-language explanations of medical terms, and a dated care log. It cannot diagnose or recommend medication on its own, but it can turn scattered materials into a checklist I can bring into the exam room.
Not medical answers, but a meeting brief I can open anytime
A friend I know, Lin Yan, is a freelance brand consultant in Hangzhou. While her mom received cancer treatment, she had to coordinate with nurses, collect mailed records, and track medication and appointments every week. She photographed the paper documents on her phone, masked identifying details, then gave them to AI to organize. She dictated her notes with Wispr Flow because speaking in the hospital corridor was easier than typing. Her words: “I used to spend more time trying to explain everything clearly than comforting my mom.” Now she brings a printed one-page summary to every appointment, with three things that must be confirmed and a short record of recent changes.
If I wanted to replicate this, the key wouldn’t be finding a more sophisticated prompt. It would be preparing the input first. Before the first step, I remove private data such as full ID numbers and home addresses. Then I open ChatGPT or Claude, start a new conversation, and click “Add file” or “Upload file.” If I don’t have medical-record files, I can manually enter five items: basic information, current treatment, known facts, the next doctor, and questions for this visit. Medical-record image recognition can make mistakes, so I check every date, medication name, and number against the original.
My replication cost: ¥0, about one hour, starting with one checklist
Money: I use the standard ChatGPT or Claude chat at no extra charge. Wispr Flow’s promotional prices and plans change, so I’m counting the transcription membership as ¥0 here; the actual price should come from the purchase page that day. Time: the first setup takes about 60 minutes—20 minutes transcribing the material, 20 minutes asking AI to summarize it, and 20 minutes checking every item. Each later update takes about 15 minutes. Technical barrier: taking photos, copying text, and uploading files are enough. No coding or technical knowledge is required.
The mistake I made first was uploading an entire medical record at once, only for the important points to disappear in the noise. Now I keep it to one page and force AI to separate “facts from the source” from “questions to confirm with the doctor.” For unfamiliar medical terms, I only ask for the original meaning, never a treatment plan. When stopping or adding medication—or dealing with urgent symptoms—I contact the doctor or hospital directly. Not everyone needs this method. If my family has little follow-up care and the records are already minimal, sticking with the hospital’s existing process makes sense. There’s no urgency to try it now.
I pick the lightest step that matches what I already have
If I’m just starting and only accompany my family occasionally, I create a blank one-page checklist with only “three questions for this visit” and “next appointment date.” There’s no need to upload a full set of records yet. If I already have one or two regular clients or am caring for a family member continuously, I append changes by date, generate a one-page follow-up summary, and keep a separate “To Confirm” section. If I’m scaling up and team members take turns handling follow-ups, I make each shift handover include the date, owner, and next step while keeping human sign-off. AI only organizes; it does not make medical decisions. That saves communication time without handing responsibility to a tool that may occasionally misread the material.