Complete each knowledge check to save your progress on this device and browser only. No account, patient information or continuing education credit is involved.
Lesson 1 Foundations 5 min
Know what AI can—and cannot—do
Choose a useful first task and recognize the limits of a convincing answer.
Open lesson
A language model generates text from patterns in its training and the context you provide. A fluent answer does not establish that its facts, references or reasoning are correct. Treat its output as a draft to inspect.
Start with a bounded administrative task: turning a synthetic meeting agenda into a checklist, for example. Keep diagnosis, treatment decisions and patient-specific advice outside these exercises.
Before using a tool at work, check your organization’s approved-use policy, access controls and data-handling requirements. For these lessons, use invented examples or public material only. Removing a name alone does not make real patient information appropriate to upload.
Try it yourself
Write down one repetitive task you do. Identify the input, the draft you want and the person who must review it before use.
Lesson 2 Prompting 6 min
Give your prompt a clear brief
Use context, a specific task, constraints and a reviewable format.
Open lesson
A useful prompt explains the context, requests one clear task, sets boundaries and specifies the output format. Include only the approved information needed to do the task. More text is not always better.
Example: ‘Using this fictional agenda, draft a 30-minute staff meeting plan. Keep all four topics, reserve five minutes for questions and return a table with duration, topic and owner. Mark missing owners as unassigned.’
Ask the model to flag missing information and avoid inventing details. If the first draft misses the brief, name the exact problem: ‘The durations total 40 minutes. Revise them to total 30.’ You still need to check the revision yourself.
Try it yourself
Rewrite ‘make this better’ as a brief that names the audience, desired output, length and facts the model must preserve.
Lesson 3 Review 7 min
Review the output, not the confidence
Check facts, omissions, source fidelity and suitability for the audience.
Open lesson
Compare the draft with the supplied input. Check names, dates, numbers and assignments; look for omitted constraints and new claims. Verify calculations independently rather than asking the same model to certify itself.
When a draft cites a source, open the original. Confirm that it exists, supports the specific statement and is appropriate for the intended use. A real URL can still be attached to a claim it does not support.
Read for audience and consequences. Is the wording understandable? Could it imply an unapproved policy or promise? Have a designated human reviewer approve the final material. For clinical educational material, follow your organization’s qualified review process before distribution.
Try it yourself
Take a synthetic draft and label each sentence: supported by the input, needs verification, or should be removed. Correct it before sharing.
Lesson 4 Practice 6 min
Run a small, measurable pilot
Define success, set a stopping rule and review responsibility before you begin.
Open lesson
Choose one approved, low-stakes workflow and name an owner. Write down what the AI will draft, what information it may use, who reviews every output and what the process does when the tool is unavailable.
Compare a small set of synthetic or approved administrative examples with your existing process. Record total time including review and corrections, factual errors, missed requirements and whether the result is usable. Do not assume faster generation means less total work.
Set a stopping rule before the pilot: for example, pause if output introduces unsupported policy statements or approved-data boundaries are unclear. Review findings with the relevant organizational stakeholders before expanding access or task scope.
Try it yourself
Create a one-page pilot brief with: task, permitted inputs, reviewer, three success measures, stopping rule and review date.
Put the idea into practice.
Try a ready-to-adapt workflow with a synthetic example.
