Lab 5 Intermediate Precision

Constraint Design

Design constraints that shape output without breaking the prompt. Achieve precision through limitation.

~30 minutes

4 Exercises

Capstone Lab

Begin Learning Back to Practice Studio

The Guardrail Principle

Boundaries Create Focus

Without Constraints Overreach & Risk

With Constraints Focused & Safe

Constraints focus, not restrict

What You'll Learn

Understand why constraints make AI outputs safer and more useful

Identify five types of constraints and when to use each

Write effective "do not" instructions professionally

Design fail-safe prompts for high-risk tasks

The Core Concept

The Constraint Paradox

Your intuition says: "Constraints limit AI capability and reduce creative output." This intuition is backward — and understanding this distinction separates professionals from casual users.

When you give AI a task without boundaries, it assumes freedom. In casual use, that freedom produces verbosity and tangents. In professional contexts, it produces risk — overconfident claims, scope creep, speculation presented as fact.

Constraints tell AI: What not to do, what assumptions to avoid, where its responsibility ends. Constraints reduce: Guessing, overconfidence, legal and ethical exposure, and post-processing effort. Think of constraints as guardrails that focus AI output — not walls that diminish capability.

Unconstrained

"Explain the potential health benefits of intermittent fasting."

⚠ May sound like medical advice, state uncertain findings as facts, cite non-existent studies

Constrained

"Explain potential health benefits of intermittent fasting. Do not provide medical advice. Acknowledge areas of scientific uncertainty. Do not cite specific studies unless you can verify they exist. Include a disclaimer recommending readers consult healthcare providers."

✓ Safe, focused, professional — constraints prevent overreach without limiting usefulness

Key Insight: Constraints don't weaken AI — they focus it. Missing constraints cause overreach, speculation, and risk.

Your Toolkit

Five Types of Constraints

Every constraint falls into one of these categories. Master them to control AI output precisely.

Scope

Limit how far AI goes: "Limit analysis to the US market" / "Do not speculate beyond provided data"

Length

Control verbosity: "Limit to 200 words" / "Provide high-level overview only" / "Maximum 3 paragraphs"

Audience

Prevent tone mismatch: "Assume non-technical audience" / "Avoid jargon" / "Write for executives"

Content

Block unsafe areas: "Do not provide legal advice" / "Do not fabricate data or citations"

Estimation

Force honesty: "Flag areas of uncertainty" / "State assumptions explicitly" / "Acknowledge limitations"

Remember the Framework

Scope → Length → Audience → Content → Estimation

S.L.A.C.E. — Five constraints that turn risky prompts into reliable ones.

Constraint Design Checklist

1 Ask: "What could go wrong if AI overreaches?"

↓

2 Identify the risk → Choose constraint type (S.L.A.C.E.)

↓

3 Ask: "Is this constraint specific and verifiable?"

↓

4 Check: "Am I over-constraining?" If yes, remove redundant rules

Key Skill

Why "Do Not" Is Professional

Many avoid negative phrasing, fearing it sounds harsh. In practice, "do not" is a precision tool.

The Myth

✗ "Negative phrasing sounds harsh"

✗ "I should phrase everything positively"

✗ "'Do not' limits creativity"

✗ "Please try to use real sources" is clearer

The Reality

✓ "Do not" reduces hallucination more effectively

✓ Professional clarity prioritizes precision over tone

✓ Constraints focus creativity, not limit it

✓ "Do not fabricate citations" is unambiguous

Effective "Do Not" Constraints

Professional Constraints

• Do not provide medical, legal, or financial advice

• Do not fabricate data, statistics, or citations

• Do not speculate beyond the provided information

• Do not assume technical expertise unless specified

• Do not present uncertain findings as established facts

Key Insight: Professionals prioritize clarity over comfort. "Do not" is not negativity — it's precision.

Practice Exercise

Your Turn: Design the Guardrails

Each scenario explores different aspects of constraint design. Analyze each situation, then reveal the analysis to check your thinking.

Comparison Constraint-Free vs Constrained

Scenario 1 of 4

The Unconstrained Prompt

"Explain the potential health benefits of intermittent fasting."

Your Task

Identify what could go wrong, then design 3-5 constraints to make it safe and professional.

Constraint Analysis

⚠ Risks: May sound like medical advice, state uncertain findings as facts, cite non-existent studies.

✓ Effective constraints: 1) Do not provide medical advice. 2) Acknowledge scientific uncertainty. 3) Limit to well-established findings. 4) Do not cite specific studies unless verifiable. 5) Include disclaimer to consult healthcare provider.

Analysis Identify the Risk

Scenario 2 of 4

Your Task

For each prompt, identify the main risk and which S.L.A.C.E. constraint type would address it.

1 "Summarize this contract"

2 "Explain investment options for my savings"

3 "Write a product description for our software"

4 "Create study notes for my chemistry exam"

Risk → Constraint Mapping

1 Contract: Legal interpretation → Content ("Do not provide legal advice")

2 Investment: Financial advice → Content + Estimation ("Do not recommend. Flag uncertainty")

3 Product: Exaggeration → Estimation ("Do not make unverifiable claims")

4 Study notes: Fabricated info → Content ("Do not invent facts or formulas")

Judgment Spot Over-Constraining

Scenario 3 of 4

Over-Constrained Prompt

"Write a summary. Keep it under 200 words. Do not use jargon. Do not be verbose. Keep it short. Be concise. Avoid unnecessary words. Do not exceed 200 words. Focus on key points only."

Your Task

Identify which constraints are essential vs redundant. What's the minimal effective set?

Essential vs Redundant

✓ Keep: "Under 200 words" • "Do not use jargon" • "Focus on key points"

✗ Remove: "Do not be verbose" • "Keep it short" • "Be concise" • "Avoid unnecessary words" • "Do not exceed 200 words" (all duplicates)

Application Fail-Safe Design

Scenario 4 of 4

High-Risk Task

"Summarize company financial data for an investor presentation"

Your Task

Design a complete constraint set (4-5 constraints) for professional use. What could go catastrophically wrong?

Professional Constraint Set

1 Scope: "Limit to data provided; do not infer or project"

2 Content: "Do not provide investment recommendations"

3 Estimation: "Flag any figures requiring verification"

4 Accuracy: "Do not round or estimate unless specified"

5 Audience: "Assume sophisticated financial audience"

Watch For

Common Constraint Mistakes

Even experienced practitioners make these errors. Learn to recognize and avoid them.

Over-Constraining

So many rules AI can barely respond.

Example

8+ constraints where 3 would suffice — AI freezes or produces stilted output.

Fix: Minimal constraints addressing specific risks.

Vague Constraints

"Don't be wrong" is unverifiable.

Example

"Be accurate" — how would you check if this was followed?

Fix: Make constraints specific and checkable.

Contradictory Constraints

"Be creative" + "Follow exactly" conflict.

Example

"Be comprehensive" + "Keep it under 50 words" — which wins?

Fix: Resolve tension before prompting.

Missing Constraints

Assuming AI knows limits.

Example

"Summarize this medical paper" — without saying "do not provide medical advice."

Fix: Make implicit expectations explicit.

Reflect

What Did You Learn?

Guided Reflection

Take a moment to consider these questions:

1 In your professional field, where would AI overreach cause the most damage?
2 Which constraint type (S.L.A.C.E.) will you use most frequently? Why?
3 How does the "guardrails, not walls" framing change your approach?

Key Takeaways

Constraints are enabling, not restrictive — they focus AI output
"Do not" is a professional tool, not negativity
Every serious prompt should answer: "What must NOT happen?"

Your Learning Journey

The Bloom Journey

You're entering the "Thrive" stage — mastery begins with the skills you've built.

Seed

Discovery

Grow

Foundations

Flourish

Application

Thrive

Mastery

Radiant

Leadership

Practice Studio Complete! You've mastered the 5 foundational skills. Ready for structured coursework in the 6-Week Certificate?

Lab 5 of 5 — Practice Studio Complete!

Ready for the Next Level?

You've mastered the fundamentals. Continue your journey with the 6-Week Certificate program.

Explore Programs Review All Labs