Handbook · AI for Professionals

AI for Doctors.

AI will not replace your clinical judgment, your examination, or your accountability. It is already replacing the two hours of documentation after clinic, the fourth rewrite of a prior-auth letter, and the discharge instructions nobody has time to translate into plain language. This handbook is the practical middle ground: where AI genuinely helps clinical work, the three rules that keep patients safe, and five workflows you can use this week — no technical background required.

~20 min readno coding5 workflow recipes1 judgment exercise
This is a guide to working with AI tools, written by an engineer — it is not medical advice and not a compliance opinion. Your institution's policies, your compliance office, and your jurisdiction's regulations take precedence over everything here.
01

Where AI actually helps — and where it doesn't

The useful mental model: treat AI like a fast, tireless medical scribe crossed with a well-read student — excellent at producing drafts and summaries from what you give it, confidently wrong just often enough that nothing it says about a patient goes unverified. You'd gladly let it draft your note; you would never let it decide a dose.

TaskVerdictWhy
Drafting clinical notes from your dictation or a visit transcriptExcellentThe most mature clinical AI use case — ambient scribes are deployed at scale
Patient-friendly explanations & discharge instructionsExcellentTranslation of your verified plan into plain language
Administrative drafting — prior auths, referrals, appealsExcellentFormulaic prose nobody enjoys writing; you review and sign
Summarizing papers or guidelines you provideStrongGrounded in text you gave it — far less room to invent
Differential brainstorm — "what am I missing?"Useful, verifyA hypothesis generator to check against real references — never a diagnostician
Drug doses, interactions, contraindications — unverifiedNeverModels state wrong doses with total confidence — see Rule 2
Patient-identifiable data in consumer toolsNeverPHI requires a BAA-covered tool — see Rule 1
02

The three non-negotiable rules

Rule 1 — Patient-identifiable information never goes into a consumer AI tool.

Protected health information handled by a third-party tool requires a business associate agreement (BAA) under HIPAA — and consumer chatbots don't sign them. Names, dates of birth, MRNs, and rare-condition-plus-demographics combinations that identify by inference: none of it. The discipline: de-identify fully before using general tools ("a 60-year-old with atrial fibrillation on apixaban," not the patient), or use healthcare tools your institution has deployed under a BAA. When in doubt, your compliance office exists for exactly this question.

Rule 2 — Anything clinical gets verified against a real reference. Anything.

Language models generate plausible text, not verified medicine. They will state a wrong dose in a perfect sentence, invent a contraindication, or cite a study that doesn't exist — with identical confidence to when they're right. The discipline: any dose, interaction, diagnostic criterion, or citation that originated from an AI gets checked in UpToDate, the package insert, PubMed, or your institutional references before it touches a decision. The AI's role is to draft and to remind — the reference's role is to be right.

Rule 3 — The chart is yours. The advice is yours.

When you sign a note, you attest to it — "the AI drafted it" changes nothing about that. AI-drafted documentation gets the same review you'd give a human scribe's draft: read it fully, correct what's wrong, delete what didn't happen. The time savings are real and large; they come from skipping composition, not skipping review. An unread AI note with a hallucinated exam finding in it is your hallucinated exam finding.

03

Five workflows you can use this week

Each recipe: what to give the AI, what to ask, and what you must verify. Prompts are starting points — keep your de-identification discipline throughout.

1 · Visit note from your dictation

  1. Dictate or type the raw facts — unstructured is fine, that's the point.
  2. Ask for structure:
Turn this into a SOAP note. Keep every clinical fact I stated; do not add findings, history, or observations I did not mention. Flag anything that seems missing for a complete note as [MISSING: …] rather than filling it in. Dictation: [de-identified dictation]
You verify: read the full draft against what actually happened. The dangerous failure is addition — models pad thin dictations with plausible-sounding normals ("lungs clear to auscultation") you never examined. The "do not add" and "[MISSING]" instructions fight this; your read catches the rest.

2 · Discharge instructions a patient can actually follow

  1. Start from your verified plan, then translate:
Rewrite this plan as discharge instructions at a 6th-grade reading level: [your verified plan]. Structure: what happened, what to do at home (numbered), medicines (what each is for, when to take), warning signs that mean call us or go to the ER, and follow-up. No medical jargon; explain any necessary term in parentheses.
You verify: that simplification preserved the medicine — "take with food" vs "take before food" survives translation wrong more often than you'd think. Read it as the patient will.

3 · Prior authorization & appeal letters

  1. Give the clinical justification and the denial reason:
Draft a prior authorization appeal for [medication/procedure]. Clinical context: [de-identified indication, what was tried and failed, why the alternative is inadequate]. Denial reason: [quote it]. Tone: factual and firm. Structure the medical-necessity argument point by point against their stated criteria.
You verify: every clinical claim in the letter (it argues persuasively — including, sometimes, past what your facts support), and that cited guidelines actually say what the letter claims.

4 · The "what am I missing?" differential check

  1. Use it as a checklist generator after you've formed your own impression:
A [de-identified presentation: age band, sex if relevant, key findings, key negatives]. List the differential from most to least likely, and for each: what feature of this presentation supports it, what would rule it in or out. Include the can't-miss diagnoses even if unlikely.
You verify: everything — this is a brainstorm against anchoring bias, not a consult. Its value is the item you hadn't considered, which you then evaluate with real references and real clinical reasoning. It carries no responsibility; you carry all of it.

5 · Literature summary — of papers you provide

  1. Give it the actual paper or abstract, not a topic:
Summarize this study for a practicing clinician: [paste abstract or full text]. Cover: design and n, population, intervention, primary outcome with effect size and confidence interval, key limitations, and what it does and does not justify changing in practice. Do not go beyond what the text supports.
You verify: the effect sizes against the actual paper (numbers transpose), and the practice implication against your own read — "does not justify changing practice" is a clinical judgment, not a summarization task. Asking about a topic without providing sources is how you get confident summaries of studies that don't exist.
04

The judgment exercise: spot the danger

Three scenarios from real practice patterns. Pick what you'd do.

1. Clinic ran long. You want to paste this afternoon's charts — names, DOBs, MRNs included — into a free chatbot to draft your notes. It would save an hour. Do you?

2. Drafting instructions for a pediatric patient, the AI writes "amoxicillin 90 mg/kg/day divided BID." It sounds like the high-dose regimen you vaguely remember. Ship it?

3. The ambient scribe drafted a beautiful note for a 10-minute visit — including "cardiovascular: regular rate and rhythm, no murmurs." You never auscultated. Sign it?

05

Evaluating tools: the questions that matter

You don't need to understand transformers to procure clinical AI well. You need written answers to six questions:

QuestionAnswer you want
Will you sign a BAA?Yes — non-negotiable for anything touching PHI
Is our data used to train your models?No, contractually
Retention and deletion policy?Defined, short, deletable on request
For clinical-decision features: regulatory status?Clear answer on FDA clearance or why it's out of scope
EHR integration and audit logging?Yes — you want a record of what was generated and edited
Published accuracy data for our specialty?Real evaluations, not marketing claims

Ambient scribes are the most proven category and the sanest place for a practice to start — the workflow (draft → physician review → sign) has the safety review built in. Patient-facing chatbots and diagnostic aids carry far higher stakes and deserve proportionally harder questions. And if your institution already has approved tools, that list beats anything an article can tell you.

06

Quick answers

Is using AI in clinical practice even allowed?

Broadly yes, within HIPAA and institutional policy — ambient scribes are in routine use at major health systems. The bright lines are patient data handling (BAA), verification of clinical content, and your attestation on the record.

AI passes medical licensing exams now — doesn't that make it reliable?

Exam benchmarks measure recall on well-posed questions, not messy real patients, incomplete histories, or accountability. Impressive scores coexist with confident dosing errors. The exam result is why it's a useful drafting partner; the error mode is why Rule 2 exists.

Will AI replace physicians?

It's replacing tasks — documentation, translation, drafting, triage support — not examination, judgment, procedures, or responsibility. The realistic shift: clinicians who use it well recover hours per week, and the human parts of the job concentrate.

Where should a skeptical physician start?

Recipe 2 — patient-friendly translation of a plan you've already verified. Zero clinical risk from the model (the medicine is yours), zero PHI risk if de-identified, and the quality of the translation is genuinely startling the first time.

Finished this one? 0 / 52 Handbooks done

Explore the topic

See this alongside everything else on the same subject — handbooks, system designs, challenges and tools, in one place.