Lesson 7

Why Multi-Part Templates?

If you have been writing prompts for any length of time, you have likely experienced the frustration of a vague, rambling, or simply wrong output. You ask for a marketing plan and receive a generic list of buzzwords. You ask for a code review and get a summary of what the code does, not what is wrong with it. The root cause is almost never the AI model's capability. It is the structure of your prompt.

A single-sentence prompt like "Write a marketing plan for my product" leaves the model to guess everything: your industry, your audience, your budget, your timeline, your tone, and the format you expect. The model will fill those gaps with its most statistically likely guesses, which are often generic and unhelpful. This is the core problem that multi-part templates solve: they eliminate ambiguity by explicitly defining every dimension of the task before the model generates a single token.

Think of a prompt as a specification document. A good specification leaves nothing to chance. A multi-part template is a reusable specification structure that forces you to provide the essential context the model needs to produce a professional-grade result. It is not about making prompts longer for the sake of length; it is about making them complete.

The Template: [Role] + [Context] + [Task] + [Format] + [Constraints]

This lesson introduces a five-part template that you will use for the rest of this course. Each part serves a distinct purpose, and together they cover the full spectrum of what a model needs to know:

• [Role] — Who the model should be. This activates the appropriate knowledge and tone. "You are a senior marketing strategist with 10 years of B2B SaaS experience" is far more effective than "You are a helpful assistant."
• [Context] — The situation, background, and relevant data. This grounds the model in your specific reality, not a generic one.
• [Task] — The specific action you want performed. Be precise: "Create a 4-week content calendar" is better than "Help me with content."
• [Format] — The structure of the output. Specify headings, bullet points, tables, word count, or code blocks. If you do not specify, the model chooses, and its choice may not match your needs.
• [Constraints] — Boundaries, exclusions, and requirements. This includes what to avoid, what to prioritize, and any hard limits.

These five parts are not arbitrary. They map directly to the questions a professional would ask before starting any project: Who is doing this? What is the situation? What exactly needs to be done? What should the deliverable look like? What are the rules?

Worked Example: A Marketing Plan for a Real Product

Let us apply this template to a concrete scenario. Suppose you are a product manager at a company called CloudSprint, which sells a project management tool for remote teams. You need a 90-day marketing plan to launch a new feature: an AI-powered meeting summarizer that integrates with Zoom and Microsoft Teams.

Here is the full prompt using the template:

You are a senior B2B SaaS marketing strategist with 8 years of experience launching features for project management tools.

Context: CloudSprint is a project management platform used by 12,000 remote teams. We are launching "MeetingSprint," an AI feature that automatically summarizes meetings from Zoom and Microsoft Teams, extracts action items, and syncs them to CloudSprint tasks. Our target audience is operations managers and team leads at companies with 50–500 employees. Our current monthly recurring revenue is $180,000. The feature will be available to all paid plans at no extra cost. We have a 2-person content team and a $15,000 launch budget.

Task: Create a 90-day go-to-market plan for MeetingSprint. Include a week-by-week breakdown of activities, the primary channels we should use, and the key performance indicators we should track.

Format: Use a table with columns: Week, Activity, Channel, KPI. Below the table, provide a 150-word executive summary. Use professional but direct language.

Constraints: Do not include paid advertising in the first 30 days. Focus on organic channels and existing customer communication. Do not suggest webinars, as we have low attendance historically. Prioritize activities that can be executed by a 2-person team without external agencies.

Now, let us examine why each part matters in this specific example.

The Role statement ("senior B2B SaaS marketing strategist") primes the model to use industry-specific frameworks and vocabulary. It is not a magic trick, but it does shift the statistical distribution of the model's output toward more relevant patterns. The Context section provides the concrete numbers (12,000 teams, $180,000 MRR, $15,000 budget) that prevent the model from inventing unrealistic scenarios. Without these numbers, the model might suggest a $100,000 campaign or target enterprise companies with 5,000 employees.

The Task is explicit about the deliverable: a 90-day plan with a week-by-week breakdown. The Format section specifies a table with exact column names, which forces the model into a structured output that is immediately usable. Finally, the Constraints section is where you prevent common mistakes: no paid ads for 30 days, no webinars, no external agencies. These constraints reflect real business decisions and save you from having to reject the model's first draft.

Step-by-Step Application

Here is how you can apply this template to any task, step by step. Follow this sequence literally, and you will produce a well-structured prompt every time.

1. Define the Role (1 sentence). Write down who the model should be. Be specific about seniority, domain, and years of experience. Example: "You are a senior data analyst with 5 years of experience in e-commerce analytics."
2. Write the Context (3–5 sentences). Answer these questions: What is the situation? What data do you have? Who is the audience? What are the constraints of the real world? Include numbers whenever possible. Example: "Our online store has 45,000 monthly visitors and a 2.1% conversion rate. We sell handmade leather goods. Our average order value is $85."
3. State the Task (1–2 sentences). Use an action verb: create, analyze, compare, rewrite, debug, summarize. Be specific about the scope. Example: "Analyze the last 3 months of sales data and identify the top 5 underperforming product categories."
4. Specify the Format (1–2 sentences). Tell the model exactly how to structure the output. Use concrete terms: table, bullet points, numbered list, code block, JSON, 200-word summary, etc. Example: "Present the findings in a table with columns: Category, Revenue, Units Sold, Conversion Rate, Recommendation."
5. Add Constraints (1–3 sentences). List what the model must not do, and what it must prioritize. Example: "Do not include suggestions that require a larger inventory. Prioritize recommendations that can be implemented within 2 weeks."

After writing the prompt, read it aloud. If any sentence could be interpreted in two ways, rewrite it. If you have not included a number where one could exist (budget, time, quantity), add it. The goal is a prompt that a competent human contractor could execute without asking you a single clarifying question.

Common Mistakes

Expert Tip

Practice Task

Your task is to write a complex prompt using the five-part template. Do not use the marketing example above. Instead, choose one of these scenarios:

• Scenario A: You are a software developer. Ask the model to review a Python script you have written (write a short, buggy script yourself, or use a simple one you have). Request a code review that identifies bugs, performance issues, and style violations.
• Scenario B: You are a hiring manager. Ask the model to create a 30-minute interview script for a junior data analyst position. Specify the skills to test and the format of the questions.
• Scenario C: You are a content writer. Ask the model to rewrite a 200-word blog post you have written (write a rough draft yourself) to be more concise and persuasive for a technical audience.

Write the full prompt in a text editor. Then, paste it into your preferred AI tool (ChatGPT, Claude, Gemini, etc.) and run it. Evaluate the output against your constraints. Did the model follow the format? Did it respect the constraints? If not, revise your prompt and run it again.

Self-verification checklist:

• Does your prompt contain all five parts, clearly labeled or separated?
• Does your context include at least two specific numbers (e.g., team size, word count, time frame)?
• Does your format section specify the exact structure of the output?
• Does your constraints section contain at least one "do not" statement?

This task should take you under 15 minutes. The goal is not perfection on the first try; it is to internalize the habit of structuring your prompts completely before you hit enter.