Six parts
Role, facts, task, output, your numbers, guardrails.
In that order, every time. A prompt with all six produces a draft a team can check in ten minutes. A prompt missing any of them produces something that sounds right and is not yours.
01
Give it a role, in one sentence
Not "you are a helpful assistant". An operations advisor to the executive team of a multi-location practice. A compliance officer preparing a board note. The role sets the altitude and the vocabulary.
02
Give it the facts, with the sources
The figures exactly as published, the organization named, the outlet or journal named. A model that is handed facts reasons from them; a model that is handed a topic invents.
03
State the task the move implies
Bounded and finishable: set a scorecard, draft a workflow with owners, list the questions for the vendor, write the memo. "Tell me about ambient AI" is not a task.
04
Name the output precisely
A one-page memo with these four sections. A table with these columns. A workflow with a step, an owner and a check per row. The output you name is the output you get.
05
Leave slots for your own numbers
[your organization], [your EHR], [your current after-hours documentation hours], [the number of clinicians in the pilot]. The slots are what make the output yours rather than generic.
06
Set the guardrails inside the prompt
Use only the facts given and the slots filled. Cite the sources by name. Flag what is uncertain or must be verified. No patient data. This is a draft for the team to check. Nothing is medical advice.
The master prompt
Paste it, fill the brackets.
This is the frame every item's prompt follows. Use it on any article, any memo, any decision that lands on your desk.
Run this
The prompt to act on this item, for your own model.
Paste it into Claude, ChatGPT or Copilot and fill the bracketed slots with your own numbers. The output is a draft for your team to check, built only from the sources on this page. Never paste patient data. How these prompts are written.
You are an operations advisor to the executive team of [your organization: type, size and locations]. I am the [your role]. I will give you facts from a published source and my own numbers. Work only from those. The facts, from [the source, named]: [paste the item's facts here, with every figure exactly as published]. My numbers: [your current figure for the thing the item measures]; [the size of the team or pilot involved]; [the system or vendor you run today]. The task: [the move from the item, stated as something to finish this week]. The output: a one-page memo for the executive team with four sections, in this order: what happened and why it matters to us; what we would measure and the baseline we have today; the first three steps with an owner and a date each; what we do not yet know and how we would find out. Rules: use only the facts I gave you and the numbers I filled in; cite the source by name wherever you use its facts; flag anything uncertain or that needs verifying before we act; include no patient data and ask for none; say at the top that this is a draft for the team to check and that nothing in it is medical advice. Plain prose, short paragraphs, no bullet symbols, no emojis.
Why it works: a model given a role, facts, a bounded task and a named output stays inside what it was given. The slots make the output yours. The guardrails keep it honest. That is the same discipline Your Chief AI Officer Daily Brief applies to itself: every item is drafted from named sources and checked by rule before it publishes.