Skip to content

Unit B-1: Don't Give Up After One Try — Growing Output Through Dialogue

Related competencies: B1, B2 | Estimated time: 10 minutes

Why This Skill Matters

"I tried generative AI, but all I got back was a generic answer" — plenty of people say this and give up on using it. In most cases, though, the problem isn't a limitation of the tool but how it's being used. Generative AI is not a vending machine that produces a perfect answer from a single instruction; it's a sparring partner that grows its output through dialogue. Treat the first output as a rough draft, supply the information that's missing, and say what you'd like fixed. Whether or not you go through this back-and-forth makes a huge difference in the quality of what you get.

Core Concepts

  • Iterative refinement is the basic approach: Entering a prompt is not a one-and-done action. Treat the first output as a "rough draft," and give specific feedback on what's missing to direct the improvement. Just two or three rounds of this back-and-forth can transform the output.
  • Context (background information) shapes the output: The more background information and constraints you provide — audience, purpose, time, length, format — the more the output will fit your actual workplace. Conversely, if you provide nothing, the AI can only return "generic advice that could apply anywhere in the world."
  • You can add context later: You don't need to have all the context ready from the start. If you look at the output and realize, "Oh right, I only have 30 minutes," just say so on the spot — the AI will revise accordingly.

Exercise Themes and Flow

Choose the tab closest to your own profession (feel free to substitute a theme from your own work instead). The flow is the same for every profession: Step 1: Make a deliberately bare-bones request → Step 2: Add context → Step 3: Refine with follow-up instructions.

Theme: Creating a curriculum for a resident study session

Step 1 — Start by deliberately sending just this.

Please create a curriculum for a study session for residents.

Step 2 — Once you look at the output and think, "this is too generic," add context.

Please revise the plan above to match the following conditions.
- Audience: About 10 first- and second-year postgraduate residents
- Theme: Initial management of symptoms commonly seen in the emergency
  department (chest pain, abdominal pain, dizziness, etc.)
- Time: 30 minutes once a week, 8 sessions total
- Learning objective: For each symptom, participants can explain in their
  own words how to rule out dangerous conditions
- Format: Case-based, with time for participants to speak at every session

Step 3 — Refine further with follow-up instructions (examples).

For Session 1 (chest pain), please create a minute-by-minute schedule for the day.

Please add one comprehension-check question that can be used at the end of each session.

Theme: Creating a procedure handout for new nurses

Step 1 — Start by deliberately sending just this.

Please create a handout explaining venipuncture for new nurses.

Step 2 — Once you look at the output and think, "this won't work for our training program," add context.

Please revise the handout above to match the following conditions.
- Audience: Nurses in their first year on the job
- Purpose: To learn the full venipuncture procedure and the key safety-check points
- Setting: A handout to be distributed at a group training session
- Length: No more than 2 A4 pages
- Format: A two-part structure — a step-by-step checklist plus a "common
  mistakes and how to handle them" section

Step 3 — Refine further with follow-up instructions (examples).

Please add a short example dialogue to the "common mistakes and how to
handle them" section so readers can picture the situation on the ward.

Please add a brief note to each checklist item explaining why it's necessary.

Theme: Creating patient medication guidance materials

Step 1 — Start by deliberately sending just this.

Please create patient medication guidance materials for warfarin.

Step 2 — Once you look at the output and think, "this would be too difficult for this patient," add context.

Please revise the materials above to match the following conditions.
- Audience: Elderly patients and their families
- Purpose: To help them understand what to do about a missed dose, and
  which foods and over-the-counter medicines require caution
- Length: One A4 page
- Format: Bullet points intended to be printed in large text. Avoid technical jargon.

Step 3 — Refine further with follow-up instructions (examples).

Please phrase the headings as questions (for example, "What if I forget a dose?").

Please add a section explaining "why natto is a problem" in 2-3 lines,
using language a patient would understand.

Theme: Creating a home exercise handout for a patient

Step 1 — Start by deliberately sending just this.

Please create a handout explaining home exercises for a patient.

Step 2 — Once you look at the output and think, "this doesn't fit this patient," add context.

Please revise the handout above to match the following conditions.
- Audience: An elderly patient going home after surgery for a hip fracture
- Purpose: To help them safely continue strength and balance exercises
  at home after discharge
- Length: One A4 page
- Format: Bullet points for each exercise, with "target repetitions,"
  "points of caution," and "when to stop." Avoid technical jargon.

Step 3 — Refine further with follow-up instructions (examples).

Please add a one-line fall-prevention tip to each exercise (for example,
"hold on to a handrail or a chair while doing this").

Please add a section called "What if it hurts?" in 2-3 lines, using
language a patient would understand.

Steps

  1. Choose the theme for your profession from the tabs above.
  2. Send the Step 1 prompt to your generative AI exactly as written, and observe the output. Notice the feeling of "this isn't wrong, but it wouldn't work at my workplace."
  3. Send the Step 2 prompt next, and compare how much more specific the output becomes relative to Step 1.
  4. Using the Step 3 follow-up instructions (or ones you come up with yourself), refine the output at least twice more.
  5. Keep this conversation history — do not delete it. You will use this same conversation for the Area B practical exercise later on.

Example of Expected Output

  • In Step 1, you tend to get a list of generic items that could apply to any institution (which isn't wrong in itself).
  • After Step 2, the output reflects the audience's level, the time frame, and the format, becoming specific enough that it "looks ready to use as-is."
  • After Step 3, the deliverable reaches the level of detail you actually need — a minute-by-minute schedule, example dialogue, and so on.

Experiencing this shift from "generic advice" to "workplace-ready" is the goal of this unit.

Common Pitfalls and How to Handle Them

  • Making a vague request like "make it better": The AI can only guess at what to fix. State clearly, one item at a time, "what" needs fixing and "how" you'd like it fixed.
  • Packing in too many conditions at once, so the output gets long and unfocused: It's fine to add conditions gradually. Lock down the basic structure first, then refine the details with follow-up instructions.
  • As the conversation gets longer, the AI forgets the original constraints (time, length, etc.): Restate important constraints, for example, "Please keep to the 30-minute limit we discussed."
  • Every refinement request rewrites the whole thing, including parts you liked: Limit the scope by saying, for example, "Please change only XX and leave everything else as is."

Safety Note

Do not enter patient information or personal information

Do not enter information about real patients, and do not enter the real names of staff or residents either. Every theme in this unit works with entirely fictional settings. Also, do not enter confidential workplace information (such as the details of incident reports).

Self-Check Quiz

You can check your understanding of this unit. There is no pass/fail judgment, and no record is saved.

Deliverable Feedback (Practical Exercise)

The dialogue you conducted in this unit can be used directly for the Area B practical exercise. At the end of the conversation, send the feedback prompt from the Feedback AI page to receive feedback, then paste the feedback you receive into the submission field on the Area B quiz page. (It's also fine to do this using the Unit B-2 theme instead.)

Next Steps

Move on to Unit B-2: Having the AI Ask Questions and Give Critique. If you already feel confident, you can go straight to the Area B quiz.