← LEARN

Anatomy of aGood Prompt.

Better prompts are not about using complicated wording. They are about giving AI enough direction to understand the task, the situation and the result you actually need.

01 / THE CORE IDEA

A good prompt reduces uncertainty.

When AI receives a short instruction with little context, it has to make more assumptions about what you want. Some assumptions may be reasonable. Others may produce the wrong tone, structure, level of detail or even the wrong interpretation of the task.

A stronger prompt gives the model enough direction to produce an answer that is closer to your actual need on the first attempt.

GOOD PROMPTING IS NOT

Writing long or complicated instructions for every task. The goal is to include the information that meaningfully changes the quality of the result.

02 / PROMPT COMPONENTS

Eight building blocks you can use when they matter.

Not every prompt needs all eight. Think of them as controls you can add when you need more precision.

01

Task

State clearly what you want AI to do. A vague task usually produces a vague result.

EXAMPLE

Draft a supplier follow-up email.

02

Role

Give AI a useful perspective when specialist framing would improve the result.

EXAMPLE

Act as an operations manager.

03

Context

Provide the background information AI needs to understand the situation correctly.

EXAMPLE

The shipment is three days late and the customer is waiting.

04

Audience

Tell AI who will read or use the output so it can adjust language, detail and tone.

EXAMPLE

Write for a senior manager.

05

Output format

Specify how the answer should be structured instead of leaving the format to chance.

EXAMPLE

Use a short email with subject line and three paragraphs.

06

Tone

Define how the response should sound when communication style matters.

EXAMPLE

Professional, polite and firm.

07

Detail level

Tell AI whether you need a quick answer, a practical explanation or a detailed analysis.

EXAMPLE

Keep it under 150 words.

08

Constraints

Set boundaries, requirements or things the model should avoid.

EXAMPLE

Do not promise a delivery date that has not been confirmed.

03 / WEAK VS STRONGER

Small changes can significantly improve the output.

WEAK

Write an email about a late delivery.

STRONGER

Act as an operations supervisor. Draft a professional email to a supplier about a shipment that is three days late. Ask for the confirmed dispatch status and revised ETA. Keep the email under 150 words, remain polite but firm, and do not make assumptions about the cause of the delay.

The stronger version tells AI what role is useful, what happened, what the email must achieve, how it should sound and what it should avoid.

ROLE

Operations supervisor

TASK

Draft supplier follow-up email

CONTEXT

Shipment is three days late

GOAL

Get status and revised ETA

TONE

Professional, polite and firm

CONSTRAINT

No unsupported assumptions

04 / REAL-WORK EXAMPLE

Build the prompt from the situation you actually have.

Imagine you need AI to help prepare an inventory discrepancy summary for management after an investigation.

01
Task

Prepare a concise inventory discrepancy summary.

02
Context

System stock is 500 units and verified physical stock is 476 units. Receiving, outbound and adjustment transactions have already been reviewed.

03
Audience

The summary will be sent to the inventory manager.

04
Output

Use headings for variance, checks completed, findings and recommended next action.

05
Constraint

Do not invent a root cause. State clearly when the cause has not yet been confirmed.

ASSEMBLED PROMPT

Act as an inventory control supervisor. Prepare a concise inventory discrepancy summary for the inventory manager. System stock is 500 units and verified physical stock is 476 units. Receiving, outbound and adjustment transactions have already been reviewed. Structure the summary with headings for variance, checks completed, findings and recommended next action. Do not invent a root cause, and clearly state if the cause remains unconfirmed.

05 / QUICK CHECK

Before sending your prompt, check these six things.

01

Is the task specific?

✓
02

Have I provided the necessary context?

✓
03

Does AI know who the output is for?

✓
04

Have I specified the format I need?

✓
05

Are important constraints included?

✓
06

Can the result be reviewed before use?

✓
DON'T OVERBUILD

If a simple prompt already gives you the result you need, stop there. Add more structure only when it improves accuracy, consistency or usefulness.

06 / PRACTICAL TAKEAWAY

Prompting is mostly about clarity.

Start with the task. Add the context that changes the answer. Then specify the output requirements that matter.

01Define

What should AI do?

→
02Contextualise

What does AI need to know?

→
03Specify

What should the output look like?

→
04Review

Did the result meet the need?

BUILD IT WITH KRIYANO

Use the Prompt Builder to structure your next prompt.

Add your task, role, audience, context, output format, tone, detail level and constraints in one guided workspace.

Open Prompt Builder
NEXT LESSON

How to Give AI Better Context

Learn which background details improve AI output and how to provide useful context without making prompts unnecessarily long.

CONTINUE →