In the world of AI, the quality of your prompts determines the quality of the output. A well-crafted prompt is like a clear set of instructions for a brilliant but very literal assistant. To get the best results, it’s helpful to structure your prompts using a framework. A popular and effective one is the Persona, Task, Context, Format (PTCF) model.

Persona: Act as Who?

The persona sets the stage for the AI. By telling the AI who to act as, you guide its tone, style, and expertise. Are you looking for a formal explanation from a university professor, or a simple, encouraging guide from a fitness coach?

Example:

<persona>
Act as a seasoned software engineer...
</persona>

<persona>
You are a helpful and witty marketing expert...
</persona>

<persona>
Imagine you are a travel blogger...
</persona>

Task: What Needs to Be Done?

This is the core of your prompt. Be specific and unambiguous about what you want the AI to do. Break down complex tasks into smaller, manageable steps if necessary.

Example:

<task>
...write a Python script that...
</task>

<task>
...create a three-day itinerary for a trip to Paris...
</task>

<task>
...explain the concept of blockchain in simple terms...
</task>

Context: What Info Do We Have?

Providing context gives the AI the necessary background information to complete the task accurately. This can include data, examples, constraints, or any other relevant details.

Example:

<context>
...for a web application that uses React for the frontend and Django for the backend.
</context>

<context>
My flight arrives at 9 AM and I'm interested in art museums and historical sites.
</context>

<context>
The target audience is non-technical, so avoid jargon.
</context>

Format: Expected Format of Output

Specify the format you want the output in. This could be a list, a JSON object, a block of code, a table, or a specific writing style. This saves you time and effort in post-processing the AI’s response.

Example:

<format>
Provide the output as a JSON object with the keys 'name', 'capital', and 'population'.
</format>

<format>
Write the answer in a series of bullet points.
</format>

<format>
The code should be well-commented and follow PEP 8 standards.
</format>

Putting It All Together

Here’s an example of a prompt that uses all four elements:

Prompt:

<persona>
Act as a senior data analyst. 
</persona>

<task>
Your task is to write a SQL query to find the top 5 most active users from the `users` and `activity` tables.
</task>

<context> 
The `users` table has columns `user_id` and `name`, and the `activity` table has columns `user_id` and `login_date`.
</context>

<format>
Provide the output as a single SQL query, with a brief explanation of how it works.
</format>

Prompt PTCF as draft:

<persona>
Act as a senior  
</persona>

<task>
Your task is to write 
</task>

<context> 

</context>

<format>
Provide the output as a
</format>

By using the PTCF framework, you can create prompts that are clear, concise, and effective, leading to better and more reliable results from your AI assistant.

🎉🎉🎉