Prompting Techniques Guide

Whether you’re new to AI or already using it daily, these techniques will help you work smarter.

How to Use This Guide

New to AI? Start with the Foundations section
Already prompting? Skip to Intermediate or Advanced techniques
Looking for something specific? Use the quick reference table below

I want to…	Try this technique
Get a quick answer	Basic prompting
Explore a problem I don’t fully understand	Interview
Get high-quality creative output	Rubric
Explore multiple options before deciding	Multi-draft Synthesis
Break down something complex	Decomposition
Get consistent, repeatable output	Templating
Condense lots of information	Summarisation
Just get the work done	Execution mode
Challenge my own thinking	Adversarial

Foundations

Structuring Your Prompts

Before diving into techniques, a well-structured prompt makes everything work better. Think of it as giving clear instructions to a new team member.

The core elements:

1. Role      — who should the AI be?
2. Goal      — what does success look like?
3. Task      — what specifically should it do?
4. Context   — what background is relevant?
5. Constraints — what rules or limits apply?
6. Format    — how should the output look?

Example — basic:

Write a project update email.

Example — structured:

You are a project manager writing to senior stakeholders.

Goal: Inform them of project status and get approval for timeline extension.

Task: Write a brief email (under 200 words) covering:
- Current status
- Reason for delay
- Proposed new timeline
- What you need from them

Tone: Professional but direct. No jargon.

Where this helps:

Drafting emails and documents
Preparing presentations
Writing reports
Any communication task

Core Techniques

1. Interview Technique

What it is: Instead of trying to write the perfect prompt yourself, get the AI to ask you questions until it understands what you need.

When to use it:

You have a vague idea but can’t articulate it fully
The task is complex with lots of variables
You want to uncover requirements you haven’t thought of

The prompt:

Interview me about this project until you fully understand the requirements.

- Ask one question at a time
- Ask probing, non-obvious questions
- Go deep before moving on
- Number options so I can reply with just the number
- Use y/n format for yes/no questions

Where this helps:

Scoping new projects
Writing briefs or specifications
Preparing for presentations
Planning events or initiatives
Defining job roles or processes

Tip: You can steer the interview by adding focus areas like “focus on risks and dependencies” or “prioritise budget considerations.”

2. Rubric Technique

What it is: Ask the AI to define what “excellent” looks like, then iterate until it meets that standard.

When to use it:

You want high-quality output, not just “good enough”
Creative or subjective tasks where quality matters
One-shot deliverables (presentations, proposals, designs)

The prompt:

Before you begin:
1. Think deeply about what makes an excellent [output type]
2. Create a rubric with 5-7 criteria (keep this internal)
3. Draft your response
4. Evaluate against your rubric
5. If it doesn't hit top marks on all criteria, revise and repeat

Only show me the final output.

Where this helps:

Creating presentations
Writing proposals or pitches
Designing processes
Drafting important communications
Any high-stakes deliverable

Tip: You can also provide your own rubric if you have specific quality criteria in mind.

3. Summarisation Techniques

What it is: Different ways to condense information depending on what you need.

Choose your approach:

Technique	Prompt	Best for
TL;DR	”Summarise in one sentence”	Quick overview
Bullet extraction	”Extract key points as bullets”	Scanning documents
BLUF	”Lead with the conclusion, then supporting points”	Executive summaries
Action-focused	”What decisions or actions are needed?”	Meeting notes
Audience-specific	”Summarise for [role/level]“	Tailoring communication
Progressive	”Summarise in 1 sentence, then 1 paragraph, then 1 page”	Layered understanding
Delta	”What’s new or changed compared to [X]?”	Tracking updates

Example — BLUF format:

Summarise this document using BLUF format:

## Bottom Line
[One sentence conclusion]

## Key Points
[3-5 bullets]

## Supporting Detail
[One paragraph of context]

Where this helps:

Processing long documents or reports
Preparing for meetings
Briefing colleagues or leadership
Staying on top of industry news
Reviewing contracts or policies

Tip: Combine techniques — “Summarise for a CEO using BLUF format, focus on financial implications.”

4. Templating

What it is: Define the exact structure you want, so outputs are consistent and ready to use.

When to use it:

You need repeatable, consistent outputs
The output needs to fit a specific format
You want to minimise editing

Approaches:

Fill-in-the-blank:

Use this template:

Subject: [Project Name] — Weekly Update [Date]

Status: [On Track / At Risk / Blocked]

Progress this week:
- [Item 1]
- [Item 2]

Next week:
- [Item 1]
- [Item 2]

Blockers: [None / Description]

Mimic:

Match this format exactly:

[Paste an example of the format you want]

Now create one for [your topic].

Schema enforcement:

Return your response as JSON:

{
  "summary": "string",
  "key_points": ["string"],
  "action_items": ["string"],
  "risks": ["string"]
}

Where this helps:

Status reports and updates
Meeting notes
Documentation
Data entry and extraction
Any recurring deliverable

Tip: Save your best templates somewhere you can reuse them.

5. Execution Mode

What it is: When you need the AI to just get work done without constant back-and-forth.

When to use it:

The task is clear and well-defined
You trust the AI to make reasonable decisions
You want output, not conversation

The prompt:

Complete this task fully without asking for clarification.

- Make reasonable assumptions and document them
- If uncertain, choose the most sensible option and proceed
- Only stop when the work is done

More autonomous version:

Act autonomously until the task is complete.

- Do not ask permission — act, then report
- Document decisions as you go
- Only hand back when finished or truly stuck

Where this helps:

Drafting first versions of documents
Research and compilation tasks
Repetitive or well-defined work
When you’re time-poor and need output fast

Tip: Review the output carefully — autonomy trades off against control.

Intermediate Techniques

6. Multi-Draft Synthesis

What it is: Generate multiple solutions, compare them, then combine the best elements into a superior final answer.

When to use it:

There’s no single “right” answer
You want to avoid settling on the first idea
Quality matters and you have time to explore options

The prompt:

Approach this problem three different ways:

1. Generate three distinct solutions
2. Compare the strengths and weaknesses of each
3. Synthesize the best elements into a final answer

Or shorter:

Give me three different approaches to this. Compare them. Then combine the best parts into a final recommendation.

Where this helps:

Strategy and planning
Creative work (messaging, naming, design)
Problem-solving with no clear answer
Proposals or recommendations
Process improvement

Tip: You can push for more diversity by adding “make sure the three approaches are genuinely different, not variations of the same idea.”

7. Chain-of-Thought

What it is: Ask the AI to show its reasoning step by step.

When to use it:

Complex problems with multiple factors
You want to understand the logic, not just the answer
Debugging or analysis tasks

The prompt:

Think through this step by step before giving your answer.

Or more structured:

Work through this problem:

1. First, identify the key factors
2. Then, analyse each factor
3. Consider the tradeoffs
4. Finally, make a recommendation

Show your thinking at each step.

Where this helps:

Problem-solving and troubleshooting
Decision-making
Analysis and evaluation
Learning new concepts

8. Decomposition

What it is: Break a big task into smaller, manageable pieces.

When to use it:

The task feels overwhelming
You need to delegate or distribute work
You want to track progress incrementally

The prompt:

Break this task into subtasks:

1. List all the components needed
2. Put them in logical order
3. Identify dependencies
4. Estimate relative effort for each

Then work through them one by one.

Where this helps:

Project planning
Writing long documents
Process design
Onboarding and training materials
Any complex deliverable

9. Few-Shot (Examples)

What it is: Show the AI examples of what you want before asking it to produce something.

When to use it:

The style or format is hard to describe
You have good examples to work from
Consistency matters

The prompt:

Here are two examples of the style I want:

Example 1:
[Paste example]

Example 2:
[Paste example]

Now write one for [your topic] in the same style.

Where this helps:

Matching brand voice or tone
Creating consistent documentation
Training others on a format
Any task where “I’ll know it when I see it”

Tip: Two or three examples usually beats one. Variety helps the AI understand what’s consistent vs. what varies.

10. Persona

What it is: Ask the AI to take on a specific role or perspective.

When to use it:

You want a specific expertise or viewpoint
The tone or approach matters
You want to simulate a conversation with a particular type of person

The prompt:

You are a [role] with [X years] of experience in [domain].

Approach this as you would in your professional context.

Examples:

“You are a sceptical CFO reviewing this proposal”
“You are a customer who has never used our product”
“You are a junior team member asking clarifying questions”

Where this helps:

Testing how different audiences might react
Getting specialised perspectives
Role-playing difficult conversations
Preparing for presentations or pitches

Advanced Techniques

11. Reflection

What it is: Ask the AI to review and improve its own output.

When to use it:

Quality matters more than speed
You want to catch errors or gaps
The first output was okay but not great

The prompt:

Review your response:

1. What's missing or weak?
2. What assumptions did you make?
3. How could this be clearer or more useful?

Then provide an improved version.

Where this helps:

Important documents or communications
Analysis where accuracy matters
Any output that will be widely shared

12. Adversarial / Red Team

What it is: Ask the AI to find problems, holes, or weaknesses.

When to use it:

You want to stress-test an idea
Preparing for tough questions
Risk assessment

The prompt:

Act as a critical reviewer. Find the weaknesses in this:

- What could go wrong?
- What am I missing?
- What would a sceptic say?
- Where are the gaps in logic?

Be direct and specific.

Where this helps:

Preparing proposals or pitches
Risk planning
Pre-mortems
Strategy review
Preparing for Q&A

13. Debate / Steelman

What it is: Explore multiple sides of an issue before concluding.

When to use it:

Complex decisions with valid arguments on both sides
You want to avoid confirmation bias
Stakeholders have different views

The prompt:

Before giving a recommendation:

1. Present the strongest case FOR this approach
2. Present the strongest case AGAINST this approach
3. Identify what would need to be true for each to be right
4. Then give your balanced assessment

Where this helps:

Strategic decisions
Policy changes
Investment or resource allocation
Any decision with significant tradeoffs

14. Socratic

What it is: Instead of giving answers, the AI asks questions to help you think through something.

When to use it:

You want to develop your own thinking
Coaching or mentoring situations
You’re not sure what you actually need

The prompt:

Don't give me answers. Instead, ask me questions that will help me think through this problem myself.

Ask one question at a time. Go deeper based on my responses.

Where this helps:

Personal development
Working through difficult decisions
Coaching conversations
Clarifying your own thinking

Combining Techniques

The real power comes from combining techniques. Here are some effective combinations:

Discovery + Execution:

1. Interview me about this project (Interview)
2. Then complete it without further input (Execution)

Quality-focused delivery:

1. Break this into components (Decomposition)
2. For each, define what excellent looks like (Rubric)
3. Execute and self-evaluate (Reflection)

Decision support:

1. Summarise the situation (BLUF)
2. Present both sides (Debate)
3. Stress-test the options (Adversarial)
4. Make a recommendation (Chain-of-thought)

Quick Reference: Job Function Examples

Function	Useful techniques	Example use
Project Management	Decomposition, Templating, Summarisation	Break down project plans, consistent status updates, meeting summaries
Sales	Persona, Adversarial, Few-shot	Anticipate objections, match successful email styles
Marketing	Rubric, Interview, Templating	High-quality content, campaign briefs, consistent brand voice
Finance	Chain-of-thought, Summarisation, Adversarial	Analysis with clear logic, digest long reports, stress-test assumptions
HR	Interview, Templating, Persona	Job descriptions, consistent documentation, simulate candidate perspective
Engineering	Decomposition, Reflection, Execution	Break down tasks, review code/designs, generate documentation
Leadership	BLUF, Debate, Socratic	Executive summaries, balanced decision-making, coaching
Operations	Templating, Execution, Decomposition	Standard procedures, process documentation, workflow design

Tips for Better Results

Be specific — vague input = vague output
Iterate — your first prompt rarely needs to be your last
Show, don’t just tell — examples beat descriptions
Constrain the output — length, format, what to include/exclude
State the goal — what will you do with this output?
Save what works — build a library of prompts that deliver

Getting Started

Pick one technique that matches a task you do regularly. Try it this week. Refine it. Then add another.

The goal isn’t to memorise all of these — it’s to have options when the default approach isn’t working.

Resources

OpenAI Prompt Packs — ready-to-use prompt templates for common tasks