Pentagon Model
Pillar 5: Constraints
A practical course for building clear and powerful AI prompts
The fifth and final pillar is Constraints: the boundaries that protect answer quality and stop the AI from drifting or fabricating. Constraints do not limit creativity — they steer it and make the output trustworthy and fit for professional use.
Core principle
Without constraints, the AI picks length and tone, and may "fill the gap" with info not in your data. A clear constraint keeps the reply on size and faithful to your figures.Core quality constraints (from the course template)
- Faithful to data: "do not base any conclusion on info not in the data".
- Disclose gaps: "if data is insufficient, say so explicitly".
- Decision limits: "do not issue a final decision that needs human/institutional approval".
- Style: "write in clear, direct professional language".
No constraints
Give me a final verdict on this financing request.Tight constraints
Assess it using the data only, flag any missing data, and do not issue a final decision since it needs human approval.All five pillars complete
You now hold the full template: role + context + data + task & outputs + constraints. Next you will learn to refine your prompt, then bring it all together in the Prompt Lab on a real AI.
🤖
Ask the Coach
An AI assistant that answers based on this lesson. Ask freely.
🤖 Coach answers are AI-generated; verify important details.
Check Your Understanding (2 questions)
Question 1
Which is a quality constraint that preserves answer integrity?
💡 Why: A quality constraint ties the answer to the data and prevents fabrication. Length and complexity are not quality constraints.
Question 2
"Do not state any number not present in the data" is:
💡 Why: Preventing fabrication by binding to the data is a quality constraint.
The recommended next step unlocks only after the correct answer, and your progress is saved on this device.