Use AI in Medicine / Course 3

Clinical AI Literacy: From Prompt to Governed Workflow

Learn to specify, inspect, compare, test, and govern the systems that place large language models inside clinical workflows.

intermediate 65 minutes 5 lessons
Who this is for
Physicians who already use ChatGPT or another general-purpose AI tool.
Course outcome
Design an AI-assisted workflow that protects clinical judgment, data quality, and accountability.

The model generates. The prompt specifies. The agent coordinates. The harness governs. The physician remains accountable.

This course moves from individual instructions to system-level supervision. It treats Claude Code, Google Antigravity, OpenAI Codex, and DeepSeek Harness as dated architectural examples rather than clinical products, then applies their visible controls to a bounded clinical workflow.

Course Outcomes

What you will learn

  • Write a clinical prompt as a bounded specification with explicit evidence and stopping conditions.
  • Inspect the context, tools, memory, permissions, and logs surrounding an AI model.
  • Compare agent platforms through durable architecture rather than temporary model rankings.
  • Analyze clinical AI failures across unsupported additions, factual alterations, and omissions.
  • Design a minimum viable clinical harness with meaningful physician supervision.

Before You Begin

Prerequisites

  • Completion of Clinical AI Workflows or the ability to map an AI task from source data to human checkpoint.

Ordered Syllabus

Complete the lessons in sequence.

Each article remains at its original URL. Its position here supplies the learning context.

  1. 01 Lesson 1 / intermediate / 12 min The Prompt Is a Clinical Specification Define the source, context, objective, prohibitions, and evaluation that govern a clinical AI task. Apply the SCOPE framework to a clinical prompt. Read lesson
  2. 02 Lesson 2 / intermediate / 11 min Learn the Harness, Not Just the Model Inspect the context, tools, memory, permissions, logs, and human checkpoints that turn a model into a working system. Distinguish the model from the harness surrounding it. Read lesson
  3. 03 Lesson 3 / intermediate / 13 min Same Model, Different System Compare agent platforms through context, tools, memory, permissions, observability, execution boundaries, and failure behavior. Complete a Harness Card for an AI system. Read lesson
  4. 04 Lesson 4 / intermediate / 14 min The Anatomy of a Medical AI Hallucination Trace unsupported additions, factual alterations, and omissions from source input through clinical consequence. Apply the three-error check to an AI-generated clinical artifact. Read lesson
  5. 05 Lesson 5 / intermediate / 15 min Build the Minimum Viable Clinical Harness Design a bounded clinical AI workflow that can retrieve, generate, verify, challenge, stop, and preserve a useful record. Apply the eight-stage clinical harness sequence. Read lesson