A notebook or document, a small idea, and access to an approved AI tool if you choose to use one. Use fictional or permitted material. Follow the agenda at your pace; the times are planning estimates.
Choose a task you can judge
Pick a small task you have done enough times to recognize good work. Define who uses the output and what they must be able to do with it. Use permitted or fictional inputs.
Practice example
A consultant turns an approved discovery summary into a workshop outline. The output must state one learner outcome and include practice, not just slides.
Write the input, intended user, useful output, and one boundary.
Make your judgment visible
Write three acceptance checks and two failure examples before prompting. Explain what you notice that a beginner might miss. Keep decisions requiring your expertise with you.
Practice example
Checks: the learner outcome is observable; the exercise practices that outcome; the agenda fits the time. Failure: inventing participant needs not supported by the brief.
Create a review rubric with pass, revise, and stop conditions.
Build and inspect one attempt
Give your approved AI tool the task, permitted source, rubric, output format, and missing-information rule. Save the first draft and your corrections. Do not upload confidential client material without permission.
Practice example
Prompt: Draft a workshop outline using only this fictional brief. State one learner outcome, an exercise, a time budget, and unanswered questions. Do not invent learner research. Check against the attached rubric.
Keep your prompt, original output, corrections, and the reason for each correction.
Try a second case
Use a different input with the same instruction and rubric. Include a missing or ambiguous detail. Count preparation, generation, review, and correction time. One successful attempt does not establish reliability.
Practice example
The second brief leaves the audience experience level unspecified. A useful draft flags that gap instead of silently assuming beginner learners.
Record what passed, what failed, full task time, and where human judgment was needed.
Write the handoff recipe
Document when to use the workflow, permitted inputs, instructions, review checks, stop conditions, and owner. Decide the next test yourself. Teaching the recipe is a separate step from proving buyer demand.
Practice example
Handoff: Use for an initial outline, never final learner promises. Stop if the brief lacks a learner goal or data permission. The facilitator reviews every agenda before sharing.
Leave with a one-page workflow recipe and a dated next-test plan.
Want a second pair of eyes?
Bring one task and your review rubric to an existing 90-minute AI Working Session. We can work through the inputs, output, and human checks together.
Explore the AI Working SessionConsidering a paid offer built on this workflow? Use the buyer-evidence worksheet on Outcome Driven, another site by Ashley. Your workbook is not sent there.
Finish with a useful reflection
What did you make? What can you explain? What did you verify, and what remains uncertain? Keep your work and try the process again on a new example.