What Is Prompt Engineering? Techniques That Actually Work

Why Prompting Matters More Than You Think

Two people using the same AI model get wildly different results. The difference isn't the model — it's how they communicate with it.

Prompt engineering is the practice of designing inputs that reliably produce high-quality outputs from language models. It's part writing skill, part systems thinking.

The Anatomy of a Strong Prompt

Every high-performing prompt has most of these elements:

• Role/Persona: "You are a senior financial analyst…"
• Task: "Summarise this earnings call…"
• Context: "The audience is retail investors who are non-technical…"
• Format: "Return 5 bullet points, each under 20 words."
• Constraints: "Do not include forward-looking statements."
• Examples: (if few-shot) "Here's an example of a good summary: …"

Core Techniques

1. Zero-Shot Prompting

Ask the model directly, no examples provided. Works for simple, well-understood tasks.

"Classify this customer review as Positive, Negative, or Neutral: 'The delivery was late but the product itself was great.'"

2. Few-Shot Prompting

Provide 2–5 examples of input/output before your actual query. Dramatically improves consistency and format adherence.

"Review: 'Amazing service!' → PositiveReview: 'Never again.' → NegativeReview: 'It was okay.' → NeutralReview: 'Fast delivery but wrong size.' → ?"

3. Chain-of-Thought (CoT)

Ask the model to reason step by step before answering. Dramatically improves accuracy on multi-step problems.

"Think through this step by step before answering…" or simply add "Let's think step by step."

4. Role Prompting

Assign the model a specific expert persona. Shifts the tone, depth, and perspective of responses.

"You are a senior UX researcher with 15 years of experience in e-commerce. Review this checkout flow and identify friction points."

5. Structured Output Prompting

Request JSON, tables, or specific formats. Useful for integration with downstream systems.

"Return your answer as a JSON object with keys: 'summary', 'risks', 'recommendations'. Each value should be an array of strings."

Advanced: Prompt Chaining

Break a complex task into a sequence of simpler prompts where the output of each feeds into the next.

Example — Blog Post Generation:

• Prompt 1: "Generate 10 H2 headings for a blog post about X."
• Prompt 2: "Write 200 words for each heading from this list: [output of step 1]"
• Prompt 3: "Write a compelling intro and conclusion for this draft: [output of step 2]"

This produces far better output than one massive prompt.

What to Avoid

• Vague tasks: "Write something about marketing" → too open-ended.
• Overloading one prompt: Break multi-step tasks into chains.
• Ignoring temperature: For creative tasks, higher temperature. For factual retrieval, lower temperature.
• No iteration: First draft is a starting point. Refine your prompt based on output.

Master Prompting in a Live Session

Our workshop dedicates an entire module to prompt engineering with hands-on exercises across real business scenarios.