WEAK PROMPT
"Can you take a look at my personal website and let me know what you think of it?"
AI DAY 4 / PROMPTING PLAYBOOK
Better prompts.
Better results.
A practical guide to communicating more effectively with generative AI through five principles of prompting.
01 / GIVE DIRECTION
Give AI a
point of view.
Tell the AI what role it should take, what context matters, who the audience is, and what kind of response you want. Clear direction gives the model a better understanding of the goal behind your request.
IMPROVED PROMPT
"Can you act as a professional in the IT industry and give me actionable feedback on what to improve for my personal website in terms of visual appeal or material to add?"
WHAT CHANGED?
Adding a defined role, relevant context, and a clear goal gives the AI more information about what the user actually expects from the response.
WEAK OUTPUT
Generic & Broad
The response lacks enough direction to understand the specific perspective, audience, or goal of the request.
IMPROVED OUTPUT
Focused & Actionable
With a defined role and clearer context, the response can provide more relevant and specific recommendations.
01
WHY IT WORKS
Generative AI responds to the information and direction contained in a prompt. Giving it a role, context, audience, and goal reduces ambiguity and makes it easier for the AI to produce a response that matches what you actually need.
02 / SPECIFY FORMAT
Shape the
response.
Don't leave the structure of the response up to chance. Tell the AI how you want information presented so the result is easier to read, compare, or follow.
EXAMPLE PROMPT
“Can you provide feedback on my personal website in a bullet-point list?”
THE KEY
A few extra words can completely change how the information is delivered.
CHOOSE YOUR FORMAT
Same task. Different presentation.
01
Bullet List
“in a bullet-point list.”
Best for quick, scannable information.
02
Table
AREA
FEEDBACK
Design
Content
Projects
“in a table.”
Best for organizing and comparing information.
03
Step-by-Step
01
02
03
“as a step-by-step list.”
Best for processes and instructions.
02
WHY IT WORKS
Specifying a format tells the AI how the information should be organized, not just what information to provide. A clear format can make the same response easier to scan, compare, understand, or put into action.
03 / PROVIDE EXAMPLES
Show, don't
just tell.
Examples help show the AI the type of response you expect. Depending on the task, you can provide no examples, one example, or multiple examples to give the AI more guidance.
HOW MUCH GUIDANCE?
More examples can give the AI a clearer pattern to follow.
ZERO-SHOT
0
0 Examples
PROMPT
“Write a short positive review for a coffee shop.”
→
AI relies only on your instructions.
ONE-SHOT
1
1 Example
EXAMPLE
“Great coffee, friendly staff, and a relaxing atmosphere.”
↓
PROMPT
“Write a short positive review for a coffee shop. Use this example as a guide.”
→
AI has one example to follow.
FEW-SHOT
2+
2+ Examples
EXAMPLE 01
“Great coffee, friendly staff, and a relaxing atmosphere.”
EXAMPLE 02“The drinks were delicious, the service was quick, and the shop felt welcoming.”
↓
PROMPT
“Write a short positive review for a coffee shop. Use these examples as a guide.”
→
AI can recognize patterns across multiple examples.
MORE EXAMPLES
→
MORE GUIDANCE
03
WHY IT WORKS
Examples give the AI a reference for the kind of response you want. Zero-shot prompting relies only on instructions, one-shot prompting adds a single example, and few-shot prompting provides multiple examples that help establish a clearer pattern.
04 / EVALUATE QUALITY
Don't prompt
blindly.
AI-generated responses should be reviewed before they are trusted or used. A confident answer is not automatically a good answer.
DON'T JUST ACCEPT THE FIRST ANSWER
Generate. Review. Evaluate.
GENERATE
↓
AI RESPONSE
GENERATED
“Working from home can reduce commuting time, provide greater flexibility, and help employees become more productive.”
↓
REVIEW
01
Relevant?
Does the response actually answer what you asked?
02
Accurate?
Are the facts and claims in the response reliable?
03
Complete?
Did the response address everything you asked for?
04
WHY IT WORKS
AI can produce responses that sound confident even when information is incomplete, irrelevant, or incorrect. Evaluating the response helps identify weaknesses before relying on the information.
05 / DIVIDE LABOR
Break complexity
into steps.
Large prompts can ask the AI to handle too many things at once. Breaking a complex task into smaller stages allows you to focus on one part of the problem at a time.
BREAK DOWN THE TASK
One goal. Multiple focused steps.
LARGE TASK
“Help me plan a weekend trip.”
↓
Destination
Find a destination that fits the type of trip you want.
Budget
Estimate the major costs and create a realistic budget.
Itinerary
Organize activities into a simple schedule for each day.
Final Plan
Combine the completed pieces into one organized trip plan.
ONE LARGE TASK
→
SMALLER FOCUSED TASKS
05
WHY IT WORKS
Dividing complex work into smaller tasks gives the AI a clearer objective at each stage. It also gives you more opportunities to review the results and make adjustments before moving on to the next part of the task.
BEYOND THE FIVE
Prompting is a
process.
Strong prompting is more than writing one perfect instruction. These additional practices can make AI interactions more useful, reliable, and intentional.
Add Context
Give the AI the background information it needs to understand the situation. Relevant context helps reduce assumptions and produces responses that better match your specific needs.
Verify
Check important facts, sources, calculations, and claims instead of automatically trusting the output. AI can produce convincing information that is incomplete or incorrect.
Refine
Use follow-up prompts to correct weaknesses, clarify your expectations, and guide the AI toward a stronger final result. Effective prompting is often a conversation rather than a single request.
QUICK REFERENCE
Build a better prompt.
A useful prompt often combines several pieces of information. Think about what the AI needs to know before you press send.
ROLE
+
CONTEXT
+
TASK
+
FORMAT
+
CONSTRAINTS
ROLE
Who should the AI act as?
CONTEXT
What background does it need?
TASK
What exactly should it do?
FORMAT
How should the answer be presented?
CONSTRAINTS
What limits or requirements should it follow?