The AmplifierReal cases · Healthcare · Physician
Dr Ismail, 41: a small-town clinic, guidelines that change monthly, and forty patients a day
A doctor who reads the journals at midnight, answers the same WhatsApp questions all day, and will not let a tool near a prescription.
Planned for release phase 2. The page is complete as a draft; interactive parts run in the browser and do not yet save to an account.
In brief
The situation. Dr Ismail runs a general-practice clinic in a town in Karnataka. Forty patients a day, a receptionist, his wife managing the accounts. He tries to keep up with changing guidelines and mostly fails; he answers patient WhatsApp messages between consultations; he has heard colleagues talk about AI and has a settled position — nothing that touches a diagnosis or a prescription — that he is not going to move. He is right about that, and wrong that it means there is nothing for him.
Step 1 — Define the End State
"Every patient leaves with printed instructions they can follow, in Kannada, that I have approved. I read the one guideline change a month that matters to my patients, not the twelve that don't. The same forty WhatsApp questions are answered by the receptionist from a sheet I wrote. Nothing a tool produces is between me and a clinical decision." Audiences: the patient, the receptionist, and himself as the clinician.
Step 2 — Map the Flow
His list: patient instruction sheets for the twenty most common conditions, a guideline digest, the receptionist's answer sheet. AI review adds: a review date on every sheet; a way for the receptionist to log which questions are not on the sheet, so the bank grows from reality; and translation checked by a native reader, not trusted. It suggests a symptom-checker for patients. He declines, and the framework agrees with him: cost of wrong is high, output not evaluable by the patient.
Step 3 — Assign Tasks
Which twenty conditions, and what each sheet must say: his. Drafting the sheets: AI, from his dictated notes. Translating: AI, then verified by his receptionist's mother, who reads Kannada better than either of them. The guideline digest: AI summarises what changed; he reads the source for the one change that matters. WhatsApp: receptionist from the sheet; anything not on the sheet, to him. Every diagnosis, every prescription: him, and only him.
Step 4 — Select the Tool
A voice tool to dictate his notes between patients, a conversational tool with long context for the sheets and digests (source documents pasted in, never recalled from memory), a document tool to hold the approved bank. No patient-facing AI.
Step 5 — Prepare and Direct
Criteria: every sheet has his signature line and a review date; every digest item links to the source; nothing goes to a patient that he has not read in full; the receptionist's sheet says "ask the doctor" as the answer to anything about symptoms.
Steps 6 and 7 — Execute and Evaluate
The sheets are good; the translations are not — two contain phrasing a native reader finds alarming. Caught by the criterion. The digest tries to summarise too much; he tightens the brief to "changes affecting a general practice in a small town". The receptionist's log fills with questions not on the sheet; the bank grows by a page a week.
Step 8 — Integrate and Stress-Test
Assembled after a month. Small fix: sheets printed too small for older patients. Structural: none, this time — because his end state was specific about what the tools would never touch, and the flow never drifted. The Avoider's instinct, held as a criterion rather than a wall, turned out to be exactly the right shape.
What changed
Patients leave with instructions they follow. He reads one change a month and reads it properly. The clinic runs on a sheet he wrote and a rule he kept. His position on diagnoses did not move, and did not have to.
A principle held as a criterion ("nothing between me and a clinical decision") is an Amplifier's; the same principle held as a refusal to engage is an Avoider's. Translation is a verifiable output — verify it with a human.
A composite case, not a client engagement; the person is invented, the situation is not.