02 — Prompt Engineering: Prompts

Practice writing better prompts using the GCSE framework (Goal, Context, Source, Expectations).


Part 1 — Building a GCSE prompt

Work through these in order. Each step adds one more element to the same prompt and you will see the output improve at each stage.

Step 1 — Goal only. Submit this and note the response:

Write a summary.

Step 2 — Add Context. Submit and compare:

Write a summary for a team of non-technical staff.

Step 3 — Add Source. Submit and compare:

Write a summary for a team of non-technical staff based on 
the AI Usage Guideline we have been working on in this workshop.

Step 4 — Add Expectations. Submit and compare:

Write a summary of our AI Usage Guideline for a team of non-technical staff. 
Cover the 3 most important rules they need to follow. 
Use plain language, bullet points, and keep it under 150 words.

After all four rounds, ask:

Compare the four summaries you just produced. Which one is most useful 
and why? What specifically made the difference?

Part 2 — Rewriting weak prompts

Read each weak prompt, identify what is missing from GCSE, then rewrite it before submitting.

Weak:

Write me something about AI.

Weak:

Summarise this.

Weak:

Help me with my email.

Weak:

Give me ideas.

Weak:

Make this better.

After rewriting each one, submit both versions — the weak and your improved version — and compare the outputs side by side.


Part 3 — GCSE framework practice

These are fully built prompts using the GCSE framework. Submit them as-is first, then try modifying one element at a time to see how the output changes.

I am an HR manager at a 200-person Malaysian manufacturing company. 
Write a short internal announcement (under 120 words) informing staff 
that the company has adopted Microsoft 365 Copilot and will be rolling 
out training next month. Tone: reassuring and positive.
Act as a sceptical finance director reviewing a new AI Usage Guideline 
for a Malaysian company. List 5 specific questions or concerns you would 
raise before approving it. Focus on financial risk, data security, and 
employee liability. Be practical, not theoretical.
I need to explain to a non-technical colleague what "prompt engineering" 
means. Write a 3-sentence explanation using an analogy from everyday 
life in Malaysia. Avoid technical jargon entirely.
I am preparing a 30-minute lunch-and-learn session on Microsoft 365 
Copilot for a team of 15 people in a logistics company. They have 
never used AI tools before. Write a session outline with 4 segments, 
including a hands-on activity. Keep the tone light and practical.

Part 4 — Iteration practice

This exercise runs as a single conversation. Do not start a new chat between rounds.

Round 1 — Get a first draft:

Write a short policy statement on the responsible use of AI at work. 
Audience: all staff. Length: 3 short paragraphs.

Round 2 — Localise it:

Make it more specific to Malaysian workplace culture. 
Add a reference to the PDPA (Personal Data Protection Act 2010).

Round 3 — Shorten it:

Shorten it to under 80 words and make it suitable for a 
company intranet homepage. Keep the PDPA reference.

Round 4 — Add a perspective:

Now rewrite the same statement from the perspective of an employee 
reading it for the first time. What might feel unclear or concerning? 
Suggest 2 improvements based on that perspective.

Round 5 — Final check:

Review the final version of the policy statement. What is missing 
that a complete AI usage policy should cover? List up to 3 gaps.

Part 5 — Format control

These exercises focus specifically on controlling output format. Submit each prompt and observe how specifying format changes what you get.

List 5 benefits of using Copilot in Outlook. Give the output in 
3 different formats:
1. Bullet points only
2. A numbered list with one-line explanations
3. A table with "Benefit" and "Estimated Time Saved Per Week" columns
Explain how Microsoft 365 Copilot uses RAG (Retrieval-Augmented 
Generation) to access company data. Give me 3 versions:
1. A 2-sentence plain language summary for a non-technical employee
2. A 5-bullet technical explanation for an IT team
3. A one-line version suitable for a slide title
Describe the difference between Copilot Chat and Copilot inside 
Word. Format the response as a comparison table with 4 rows covering: 
where you access it, what context it has, best use cases, and 
one limitation.

Part 6 — Role and perspective

Using “Act as” shifts how Copilot frames its response. These exercises show how the same request produces different outputs depending on the perspective you give.

Submit this base prompt first:

Review this sentence from an AI Usage Guideline: "Staff may use any 
AI tool of their choice for work tasks as long as it improves 
productivity." What are the problems with this statement?

Then submit it again with each of these roles prepended, one at a time:

Act as a cautious IT security officer. Review this sentence from an 
AI Usage Guideline and flag any security risks: "Staff may use any 
AI tool of their choice for work tasks as long as it improves 
productivity."
Act as a legal advisor specialising in Malaysian data protection law. 
Review this sentence from an AI Usage Guideline and flag any 
compliance risks: "Staff may use any AI tool of their choice for 
work tasks as long as it improves productivity."
Act as a pragmatic operations manager who wants to roll out AI tools 
quickly. Review this sentence and explain why it is actually fine as 
a starting point, with one suggested improvement.

Compare the three responses. Notice how the same sentence gets framed very differently depending on who Copilot is “acting as.”


Part 7 — Prompting inside M365 apps

These prompts are designed for use inside specific M365 apps, not Copilot Chat. The context is shorter because the app already provides the document or data.

In Copilot in Word (with a document open):

Summarise this document in 5 bullet points for a senior manager 
who has 2 minutes to read it.
Rewrite the introduction to be more engaging. The reader should 
want to keep reading after the first paragraph.
What sections of this document are unclear or could cause 
confusion for a non-technical reader? List them with a 
suggested fix for each.

In Copilot in Outlook (with an email thread open):

Summarise this thread. What has been agreed, what is still 
unresolved, and what action is needed from me?
Draft a reply acknowledging the feedback and confirming I will 
make the requested changes by end of this week.

In Copilot in Excel (with a data table selected):

What are the top 3 insights from this data? 
Highlight anything that stands out as unusual.
Create a bar chart showing the distribution of responses 
in the Productivity Rating column.

Part 8 — Stress testing your prompts

Once you have a prompt that works, push it further. These follow-up questions help you get more out of any response.

After any Copilot response, try:

What did you leave out that might be relevant?
What assumptions did you make when writing this?
Give me an alternative version with a completely different structure.
What are the weakest parts of this response and how would you improve them?
If a critical reader reviewed this, what objections might they raise?