Use a role when perspective changes the answer.
A role tells AI what kind of viewpoint or working perspective should guide the response. This can be useful when the same task could be approached differently by different functions.
But a role is not magic. Writing “act as an expert” does not automatically make an answer accurate. The task, context and evidence still matter.
Operations perspective
Act as an operations manager reviewing this process.
Useful when you want the response framed around efficiency, risk, coordination and execution.
Customer-service perspective
Act as a customer service supervisor.
Useful when tone, clarity and customer impact matter.
Trainer perspective
Explain this as a workplace trainer.
Useful when you want a structured explanation for beginners.
The perspective affects priorities, terminology, tone or the way the task should be evaluated.
Constraints define what the answer must and must not do.
Constraints are one of the most practical ways to improve prompt control. They help reduce unwanted assumptions, excessive detail, wrong formats and recommendations outside your actual scope.
Length
Keep the answer under 200 words.
Format
Return the answer as a five-step checklist.
Scope
Focus only on warehouse receiving operations.
Evidence
Do not invent facts that are not provided.
Language
Use simple English suitable for frontline staff.
Decision limits
Do not approve or reject the request. Only summarize the risks.
Do not give AI incompatible constraints such as “be extremely detailed” and “keep the answer under 50 words.” If two instructions compete, the result becomes less predictable.
Examples show AI what you mean instead of only describing it.
Examples are especially useful when the desired format, style or classification is difficult to explain precisely in one sentence.
Format example
Show AI what the final answer should look like when structure matters.
Style example
Provide a short sample when you want a specific communication style.
Classification example
Show one or two labelled examples when AI needs to sort information into categories.
Before-and-after example
Useful when asking AI to rewrite or improve content while preserving the original meaning.
Rewrite these warehouse instructions in a clear professional style.
Rewrite these warehouse instructions in a clear professional style. Use short direct sentences similar to this: “Verify the item code before confirming receipt. Report any mismatch before posting the transaction.”
Use only the controls that improve the real task.
Imagine you need help preparing a management summary after a stock discrepancy investigation.
Act as an inventory control supervisor.
Prepare a concise investigation summary for management.
System quantity is 840 units and verified physical quantity is 818 units. Receiving, dispatch and recent adjustment transactions have been checked.
Do not state a root cause unless it is supported by the information provided.
Use the headings: Variance, Checks Completed, Findings, Outstanding Questions, Recommended Next Action.
Act as an inventory control supervisor. Prepare a concise investigation summary for management. System quantity is 840 units and verified physical quantity is 818 units. Receiving, dispatch and recent adjustment transactions have already been checked. Use the headings: Variance, Checks Completed, Findings, Outstanding Questions and Recommended Next Action. Do not state a root cause unless it is supported by the information provided.
More instructions do not automatically create a better prompt.
Do not rely on “act as a world-class expert” instead of providing proper context and evidence.
Avoid adding rules that do not materially improve the output.
Check that tone, length, format and detail requirements can all be satisfied together.
AI may imitate the weaknesses of the example you provide, so use examples that actually represent the quality you want.
Leave room for useful variation when you are brainstorming or exploring ideas.
Before adding another instruction, ask whether it earns its place.
Would a role genuinely improve the perspective?
✓Which constraints are essential to the task?
✓Could an example remove ambiguity?
✓Am I giving instructions that conflict with each other?
✓Have I kept the prompt practical rather than overloaded?
✓Will I still review the final result myself?
✓Add a role for perspective, a constraint for control, and an example for clarity. If one of them does not improve the task, leave it out.
Use prompt controls deliberately.
Start with the task and context. Then add roles, constraints or examples only where they make the expected result clearer.
Choose the useful perspective.
Set meaningful boundaries.
Show what good looks like.
Remove anything unnecessary.
Put these controls into a structured prompt.
Use the KRIYANO Prompt Builder to combine role, context, output requirements, tone and constraints without losing sight of the main task.