Lesson 8 of 8 · Prompt Writing That Works
Project: Build a Prompt Pack for One Weekly Task
Put the whole course together: pick one real weekly task, build a small prompt pack for it, test it and measure the time you save.
This final lesson is a small project. You will take one task you do every week and build a prompt pack for it: a short set of prompts that takes you from raw material to finished result. By the end you will have something you can use every week, and real numbers on how much time it saves.
Step 1: Choose the right task
Good candidates are tasks that are:
- Repeated: weekly or more often
- Text-heavy: reading, writing, summarising, organising
- Low-risk to draft: you will still check the result before it goes out
Examples: weekly sales summary, client follow-up emails, meeting minutes, social media posts, stock or payment reminders, a team newsletter.
Write down how long the task takes you today without AI. This is your baseline.
Step 2: Map the task as a chain
Using lesson 5, break the task into 3 to 5 steps. For a weekly client follow-up, that might be:
- Summarise last week's notes per client
- Decide the next action per client
- Draft one email per client
- Review the emails as the client would read them
Step 3: Write one prompt per step
For each step, write a prompt using everything from this course:
| Lesson | What to include |
|---|---|
| Consistency (1) | Fixed length, audience and format |
| Context (2) | Goal, situation, constraints, source material |
| Examples (3) | One or two examples of a good output |
| Structure (4) | Tables or CSV where it helps |
| Chaining (5) | One prompt per step, run in one chat |
| Review (6) | A review prompt for the final step |
| Templates (7) | Blanks in [BRACKETS] and a "when to use" note |
A prompt pack for client follow-ups might start like this:
PROMPT 1 OF 4: Summarise notes
From my notes below, make a table with columns: Client, What happened
last week, Open issue, Promised date. One row per client. Use only what
is in the notes; write "None" if something is missing.
"""
[PASTE THIS WEEK'S NOTES]
"""
Step 4: Test it three times
Run the full pack on three real weeks (or three real cases). After each run note:
- What you had to fix by hand
- Any mistake the AI made
- Total time, including checking
Each time you fix the same thing by hand, improve the prompt so you do not have to.
Step 5: Measure and decide
Compare your average time with the baseline from step 1.
- Saved real time and quality is good? Keep the pack and share it with your team.
- Saved little time? The task may be too short or too specialised. Try the pack on a different task.
- Quality issues? Add examples (lesson 3) or a review step (lesson 6).
Step 6: Save and share
Save the final pack as one document: a title, when to use it, the prompts in order, and one filled-in example run. Add the date and a version number. Share it with one colleague and ask for feedback.
Final practice task
Complete steps 1 to 6 for your chosen task. When you have measured your time saved, mark this lesson complete: you have finished the course!
Where to go next
- Apply these skills in your field: see all free courses.
- Find the right tool for each job in the AI tools directory.
Key takeaways
- A prompt pack is a short chain of tested prompts for one real, repeated task.
- Map the task into steps, write one prompt per step, and use everything from this course.
- Test on real cases, fix the prompts instead of the output, and measure time saved.
- Save the pack with a version and share it so your team benefits too.
Finished this lesson?Mark it complete to track your progress on this device.