Anatomy of an effective prompt

Section 10.2.1 : Anatomy of an effective prompt

🎯 Learning objective

Understanding the fundamental components of an effective prompt and mastering the principles that transform a vague instruction into a precise request producing actionable results. This is THE foundational skill of the AI era.

What is a prompt?

A prompt is the instruction you give to an AI model. It's the interface between your intention and the model's capability. Output quality depends directly and massively on prompt quality — it's the #1 skill of the AI era.

Think of it as the difference between telling an assistant "Make me something good" and "Write a 3-page report on X, with an executive summary, quantitative data, and actionable recommendations for Tuesday's board meeting". Same assistant, same skills — but a radically different result.

The 6 components of an effective prompt

Every prompt can be broken down into 6 fundamental building blocks. Not all are always necessary, but knowing them allows you to construct prompts suited to any situation — from the simplest to the most complex.

1. The role (who is the AI) — Defining the persona and expected expertise

The role is the most underestimated and most powerful component. By defining who you're talking to, you steer the vocabulary, depth, and angle of the response.

Examples by level of detail:

Basic: "You are a marketing expert."
Intermediate: "You are a senior marketing director with 15 years 
of experience in B2B SaaS."
Expert: "You are a CMO of a B2B SaaS scale-up (50-200 employees) 
who has already led 3 product launches in the French market. 
You are pragmatic, data-driven, and you hate BS marketing 
with no measurable ROI."

2. The context (the situation) — Providing necessary background information

Context is everything the AI cannot guess. The more relevant context you provide, the less the AI fills in gaps with assumptions (often wrong ones).

Our CRM SaaS startup sells to French SMBs (10-50 employees).
We raised €2M in Series A 6 months ago.  
Our current revenue is €300K ARR with €25K MRR.
We have a marketing budget of €5,000/month.
Our main competitor is HubSpot (but we target 
a segment they ignore: micro-businesses of 5-15 people).
Our average sales cycle is 14 days.

The context rule: include everything a new colleague would need to know to understand your situation and produce quality work.

3. The task (what to do) — A clear and specific instruction

The task is the central element of the prompt. It must answer two questions: what to do and what level of detail.

4. The format (how to present) — The expected response structure

This is the component that transforms a "correct" response into a "directly usable" one. Without a specified format, the AI defaults to linear text — rarely optimal.

Most useful formats:

Present as a table: channel | budget | target KPI | 
concrete actions. Add a 3-line executive summary.

5. The constraints (the limits) — What to avoid or respect

Constraints prevent the AI from drifting toward responses that are too long, too technical, or off-topic.

Constraints:
- Maximum 500 words
- No Anglo-Saxon marketing jargon (use French terms)
- Only channels realistic for a €5K/month budget
- Don't suggest TV or radio campaigns (too expensive)
- Include at least 2 zero-cost channels (community, SEO)

6. The example (the model) — An example of what you expect

The example is the most powerful component for getting a precise format. A single example is worth a thousand words of instruction.

Expected format example for a channel:
| Channel: SEO Content |
| Budget: €1,500/month |
| KPI: 50 qualified leads/month |
| Actions: 8 blog articles/month, targeting "SMB CRM France", 
  pillar pages on 3 topics |

The precision spectrum — From beginner to expert

Prompts are distributed across a spectrum of precision. The good news: moving from one level to the next takes only 30 extra seconds of thought.

The investment: 2 minutes of prompt writing saves you 30 minutes of unusable results you'd have to redo. It's the best ROI of your day.

The 7 most common prompting mistakes

Practical exercise: before/after (the most important of the chapter)

Duration: 15 minutes

Transform these 3 vague prompts into expert prompts using the 6 components:

Prompt 1 — Before: "Write an email for my client"

Prompt 1 — After:

You are an experienced B2B account manager, warm 
but professional tone.

Context: Our client (e-commerce SMB, 30 employees) 
has been using our CRM for 6 months. Their NPS is 8/10 but 
they only use 40% of the features.

Task: Write a check-in email proposing a discovery session 
for advanced features.

Format: Subject (< 50 chars) + body (< 120 words) + clear CTA.
Constraint: No sales tone, focus on value for them.

Prompt 2: Transform "Give me content ideas" into an expert prompt

Prompt 3: Transform "Analyze this report" into an expert prompt

For each transformation, identify which components (Role, Context, Task, Format, Constraints, Example) you added and estimate the expected quality improvement.