Make the decision
Decide what AI must know before it can produce a useful first draft—and what you need to keep out of the prompt.
Pathway 1 · Lesson 0360 min
Prompt Engineering Foundations
The COSTAR framework and structured communication
Learning Objective
By the end of this lesson, you will be able to: Write structured prompts using the COSTAR framework, apply at least three advanced prompting techniques, and demonstrate the difference between vague and precise prompts.
2.6×
improvement in first-pass quality with structured prompts
Engage
"What if the quality of every AI output you receive is determined entirely by what you put in — before you hit send? That's not a metaphor. It's literally true."
Explore
A comprehensive introduction to prompt engineering techniques — covers the fundamentals you need before applying the COSTAR framework.
Explain
Why Most Prompts Fail
The average person types a prompt the same way they'd type a Google search: short, vague, context-free. 'Summarise this report.' 'Write an email.' 'Analyse this data.'
AI has no idea:
The result is a generic output that requires significant rework. This is the rework paradox: 37% of AI time savings are consumed by fixing poor outputs.
📖
The COSTAR Framework — by Sheila Teo
COSTAR is a universal structure for AI prompts: • C — Context: background, situation, relevant constraints • O — Objective: the specific task or goal • S — Style: writing approach (formal, analytical, persuasive) • T — Tone: emotional register (confident, empathetic, urgent) • A — Audience: who will read the output and their expertise • R — Response: desired format (table, bullets, JSON, word count)
Elaborate · Worked Example
COSTAR in Action: Before and After
Context
Task: Write a message to a team about a new AI tool rollout.
The Situation
❌ WEAK PROMPT: 'Write an email about our new AI tool.' ✅ COSTAR PROMPT: Context: I'm a department head at a 200-person financial services firm. We're rolling out Microsoft Copilot to all staff next Monday. Some team members are anxious about job security. Objective: Write an internal announcement email that generates excitement while addressing concerns. Style: Professional but warm — not corporate-speak. Tone: Reassuring and energising, not dismissive of concerns. Audience: Mixed team of analysts and admin staff, varying AI experience. Response: 250 words maximum, with a clear subject line and 3 bullet points on what changes next week.
The Outcome
The COSTAR prompt produces a targeted, contextually appropriate email on the first pass. The weak prompt produces a generic template that needs complete rewriting.
Explain
Five Advanced Prompting Techniques
Beyond COSTAR, five techniques dramatically improve output quality:
1. Role Assignment
'Act as a senior McKinsey consultant with expertise in digital transformation...'
Gives the model a persona that shapes its vocabulary, depth, and perspective.
2. Chain-of-Thought
'Think step by step before providing your final answer.'
Forces the model to reason through the problem before concluding. Dramatically improves accuracy on complex tasks.
3. Few-Shot Learning
Provide 1–3 examples of the output you want. The model learns your format, style, and quality bar from examples — not just instructions.
4. Constraints
'Do not include... Limit to 200 words... Focus only on... Avoid jargon...'
Boundaries prevent the model from going off-track.
5. Meta-Prompting
'Before answering, list what additional information you need from me.'
Lets the AI tell you what's missing from your brief — then you fill the gaps.
✅
Combine Techniques for Maximum Impact
The real power comes from combining multiple techniques in a single prompt. Role + COSTAR + Chain-of-Thought + Constraints together produce outputs that are 2.6× better on the first pass than unstructured prompts. The investment is 2 minutes of prompt writing. The saving is 30 minutes of rework.
Explore
Practical demonstrations of chain-of-thought, few-shot learning, and role assignment in action.
Explain
Context Engineering: Beyond the Single Prompt
Context engineering is the emerging discipline of designing the entire information environment for AI — not just the single message you send.
Four principles:
Think of it as briefing a new team member on their first day. What do they need to know to succeed? Give them that — and nothing more.
Try It · Reflect & Apply
Practice: Take a task you've previously given AI with a weak prompt. Rewrite it using the full COSTAR framework. What changed? What would the output difference be?
Your notes are saved in your browser for this session.
Evaluate · Knowledge Check
1 / 3In the COSTAR framework, what does the 'A' stand for and why does it matter?
Fluency Practice Studio
Prompt Craft Studio
Practise at work
Rewrite a real weak prompt using COSTAR. Then add one example output, one constraint, and the instruction: ‘List any information you still need before answering.’
Quality check
Score your prompt from 0–6: one point each for Context, Objective, Style, Tone, Audience, and Response format. A score below 5 is usually a signal to improve the brief, not blame the model.
Spaced revisit
Reuse the prompt on a new task in seven days. Record what context you added or removed and how the first-pass output changed.
External video references
Keep Learning on YouTube
Open in a new tab
Prompt Engineering Full CourseEmbedded lesson videoA comprehensive introduction to prompt engineering techniques — covers the fundamentals you need before applying the COSTAR framework.Watch · Watch on YouTube Advanced Prompt Engineering TechniquesEmbedded lesson videoPractical demonstrations of chain-of-thought, few-shot learning, and role assignment in action.Watch · Watch on YouTube
Ready to move on?
Mark this lesson complete and continue to "The 4D Framework".