Prompting tips and techniques for better outputs

Effective prompting is built upon a foundation of four key elements: persona, task, context, and format. Building on this framework, there are two prompting techniques that are particularly effective for managing complex workflows:

1. Powerful prompt phrases: Using specific, precise language inside a single prompt to improve the quality and depth of AI's response.
   
   
2. Prompt chaining: Using a series of smaller, connected prompts to break down a larger, complicated request into individual questions or tasks.

Word choice becomes increasingly important when applying AI to complex tasks, and certain phrases can be more effective in getting you the outputs you need. Powerful prompt phrases don't just tell AI what to do. They guide how AI should get there by adding precision, setting boundaries, and forcing AI to use more complex reasoning paths.

Here are some ways to use powerful prompt phrases:

Mandating a thought process often leads to a more accurate final answer. By directing AI models to process a request sequentially, or to follow a specific process, the AI model is likely to incorporate more constraints and details before generating an output.

Example prompt phrases may include:

• “Think step-by-step.” 
• “First, [action 1]. Second, [action 2]. Finally, [action 3].” 

These phrases go beyond a persona and dictate the style of the output. This helps your message land correctly, whether you're drafting an email to a potential client, a report for leadership, or a social media post for your customers.

Example prompt phrases include:

• ”Write this for an audience of [audience type, e.g. executives].”
• ”The tone should be [tone, e.g. professional].”

This can help lower the chances of the AI hallucinating (generating factually incorrect outputs) or including irrelevant information. Use this when you need content for a specific format, like a concise product description, or based on specific information, such as a sales report.

Example prompt phrases include:

• ”Do not exceed 100 words.”
• ”Focus exclusively on [topic A] and do not mention [topic B].”
• “Use only facts, data, and  information from this file.”
• “Provide citations to verifiable sources for any factual claims.”

Telling AI what to do is good, but giving it a reference of what good looks like is even better. This is one of the fastest ways to teach the AI your specific style. Clear examples of what you want can help get outputs that match your voice, or your brand’s voice, with greater consistency. It can also help if you show it what not to do.

Example prompt phrases include:

• ”Follow this pattern: [Question: ..., Answer: ...]”
• ”Here is a good response: [Example 1]. Here is a bad response: [Example 2]. Give me a response that is similar to the good one.”

This is a powerful way to find flaws or uncover fresh perspectives. Ask the AI to act as a critic. This can be helpful for important communications and creative assets, like a high-stakes email or finalizing marketing copy.

• ”Critique this text from the perspective of a [role, e.g. potential customer].”
• ”Play devil's advocate. What is the strongest counter-argument to this?”

When you're stuck or don't like the direction, ask for options. Use this technique to break through a creative block or explore different angles for your marketing, such as for brainstorming headlines, email subject lines, or calls-to-action.

• ”Give me 3 different versions of this.”
• ”What's an alternative approach to solving this problem?”

Use this technique when the output is too shallow. For instance, you can use it to turn a list of bullet points into a detailed paragraph or to add supporting evidence to an argument.

• "Elaborate on point 2.”
• “Provide more detail and specific examples for the section about [section].”

When a task is so complex that even a well-phrased prompt isn't enough, prompt chaining – another advanced technique – may be the solution. Instead of trying to craft a perfect standalone prompt on your first try, it’s helpful to treat prompting like an ongoing conversation, which is where prompt chaining comes in. This technique is all about thinking of prompting as a way to manage an entire macro-level workflow.

Chaining prompts facilitates large tasks by breaking them into a series of smaller, connected steps within the same chat session. This process functions similarly to an assembly line: the output from one prompt serves as the input for the next, linking each stage together.

This technique involves three key steps:

1. Task analysis: Start by breaking down your complex task into a series of smaller, logical steps
   
   
2. Initial prompting: Craft a focused prompt that asks AI to complete just the first step
   
   
3. Input/output flow: Use the output from the first prompt as the context for the second prompt, and continue this iterative flow until you complete the task
   

A practical example:
Planning a vacation to Paris requires managing logistics alongside personal interests. Prompt chaining can structure this project through the following sequence.

Here’s how prompt chaining could apply to this:

PROMPT 1

I'm going to Paris for 3 days. I like art, historical sites, and parks. Suggest a few well-known places I could visit on my trip.

PROMPT 2 (SECOND PROMPT IN CHAIN)

Using those locations, create a logical day-by-day itinerary that minimizes travel time.“

PROMPT 3 (THIRD PROMPT IN CHAIN)

For each day of the itinerary, suggest a few restaurants located near the suggested locations.