PORC - A Foundational Structure

At its core, your prompt should consist PORC. This stands for four key things:

1. Purpose - explain what you are trying to achieve. Aim to be as simple and as specific as you can.
2. Outcome - tell the machine what kind of format the finished product needs to be. Is it an image, a spreadsheet, a web page, an essay, something else?
3. Requirements - expand on the parameters that you will need to follow to achieve your finished product. Use examples like "a finished product will include the following."
4. Context - provide as much background about the topic as possible. The more information you provide, the better your output will be.

After providing this information in your initial message to the bot, you can expect a simple pattern to follow.

1. You provide information (the prompt): What you want to achieve; any specific requirements; important context or background
2. The AI responds with its best attempt: Based on your instructions; drawing from its training; following any format you specified
3. You review and refine: Check if the output matches your needs; identify what's missing or incorrect; provide additional guidance or clarification

Each time this pattern repeats, you should be looking to expand on the four key areas mentioned above. Your responses to the machine will provide additional context about your ask, while also helping it to better understand your intended outcome.

A Simple Example

Let's say that you are going to be hosting some colleagues for a dinner event, and want to take the guesswork out of planning.

Notice how each exchange adds more context and helps narrow down the specific help you need. For simple requests, this approach tends to be more effective than trying to include everything in one long prompt.

A (Slightly) More Complicated Example

Now, let's say that you had a fantastic dinner experience with your colleagues (thanks AI) and your team is ready to start writing a proposal for the work that you have been discussing.

As the conversation expands, focus on the patterns that begin to emerge:

• Start Broad, Then Narrow
  • Begin with your basic need
  • Let the AI help you identify what other information is needed
  • Add details progressively
• Use the AI's Questions
  • Pay attention to what the AI asks for
  • These questions often highlight important details you might have missed
  • Use them to improve your future prompts
  • If the AI asks you multiple questions, it is completely fine to provide multiple answers. Just be sure to ask for confirmation before you proceed
• Build on Previous Responses
  • Each exchange adds to the context
  • You can reference earlier parts of the conversation
  • The AI maintains understanding of the ongoing discussion

Common Beginner Mistakes

The "Kitchen Sink" Approach

Would you ask an intern to tackle this project on their first day and ask them for the same thing without providing any information first?

Probably not. And, if you did, it wouldn't go the way you hoped. The same logic applies to AI.

Why it's problematic:

• Too much information at once
• No room for clarification
• Difficult to address all points effectively

The "Too-Vague" Approach

If you were to ask this question to any marketer, they'd probably chuckle (internally, of course).

The term "marketing" is so broad and, since you've provided no information about where you currently are, where you want to go, and what resources you have at your disposal, you're not going anywhere quickly.

Why it's problematic:

• Lacks specific goals
• Requires too many follow-up questions
• Wastes time with basic clarifications

The Assumption of Knowledge

"Ok, who the heck is Jenkins, what are these configurations, and what am I doing where?" (The Bot)

Historical information is the foundation of every conversation. If you are starting from scratch with a conversation, don't dive right into phase 4/6. The bot isn't going to have a clue where to begin.

Why it's problematic:

• Assumes AI knows about your specific setup
• Doesn't provide necessary context
• Can't lead to useful, specific advice

Better Approaches

The Iterative Start

Why it works: You are providing the bot exactly what it needs in the structure that it prescribes. AI works best when you keep your prompts simple.

The Structured Start

Why it works: You clearly identify your Purpose, Context, Requirements, and Format in a succinct prompt. Yes, there is room to expand on it, but the foundational information and need has been established.