Sidecar Blog

The Ultimate Guide to Prompt Engineering: Improve Your Input for Better Output  – Sidecar

Written by Mallory Mejias | Jul 5, 2023 12:29:01 PM

Picture this: Your team just hired a new marketing person. Let’s call her Jill. On her first day, Jill is asked to create an email campaign encouraging your members to attend the annual conference. She tries her best but doesn’t quite grasp your organization’s brand voice and style in her email copy.  

Is Jill to blame? Of course not! She’s new, and she doesn’t have a great grasp on your organization just yet. Plus, it’s possible that she didn’t have access to the right information to successfully capture your organization’s voice in the first place.  

Now replace the word Jill with Generative AI. Sure, sometimes ChatGPT produces lackluster (and even comical) results, but we should approach this problem the same way we would with a new employee. Have I provided enough context? Are my instructions clear enough? 

Enter prompt engineering. Think of prompts like a recipe or a map for Artificial Intelligence (AI) to follow. If you're not getting the output you want, you might want to take a look at your input. When in doubt, experiment with these six types of prompt engineering so that AI is working with you—not against you.

When in doubt, experiment with these six types of prompt engineering so that AI is working with you—not against you.  

Related: Free Event! Introduction to Prompt Engineering for Association Professionals Learn More >

1. Get Started with AI by Using Instruction 

This foundational prompt focuses on clear and concise directions for the AI model to follow. It’s essential for generating coherent responses and serves as the basis for all other types of prompt engineering. 

Example Instruction Prompt: 

Write an introduction for a bootcamp on generative AI for associations. 

Related: Get Usable Content For Your Association Using These 3 ChatGPT Prompts Learn More >

2. Get Better Results from AI by Assigning Roles

By assigning a specific identity or perspective for the AI model to adopt, you encourage a better-aligned response. This creative approach primes the model to deliver content that matches the desired output. 

Example Role Prompt: 

Act as an event planner for Sidecar, specializing in engaging young professionals. Write an introduction for a bootcamp on generative AI for associations. 

3. Work Through Logic Problems by Using a Chain of Thought Prompt

This technique structures the prompt sequentially, allowing the AI model to reason through complex problems or concepts. It is best suited for straightforward problems that don’t require a ton of creativity. 

Example Chain of Thought Prompt: 

If John has 5 pears, eats 2, buys 5 more, and gives 3 to a friend, how many pears does he have?  

Let's think step by step. -> Without this line, the AI model gets the problem wrong! 

4. Use a Few-Shot Prompt When You Have Existing Content to Use as a Model

Using limited examples to guide the AI model, this type of prompting is useful when data is scarce. By introducing just a handful of examples, you can steer the AI to generate meaningful responses.  

Example Few-Shot Prompt: 

Write a personalized email inviting recipients to join a bootcamp on generative AI for associations. 

##Start Examples##  

First Name: Aiden  

Age: 26  

Title: Associate Manager of Membership Retention  

Recent Activity: Watched “Making Virtual Mentorship Work”  

Email: [insert previous email example] 

##End Examples## 

First Name: Mia  

Age: 57  

Title: CEO  

Recent Activity: Watched “Overcoming Hotel Contracting Challenges in a "Seller's" Market”  

Email:  

5. Use a Think - Reason - Plan - Act Prompt to Work Through Complex Challenges

This prompt technique guides the AI model through the steps of “thinking” and reasoning to execute solutions for complex, multi-step challenges. It's effective for tackling intricate problems and presenting a well-thought-out response. 

Example Think-Reason-Plan-Act Prompt: 

Write a renewal notice to first-year members of the Association for Really Awesome Accountants. Provide your thought process and plan before executing the task. 

6. Use Prompt Chaining to Assess, Plan and Execute Tasks

Combining multiple prompts creates a fluid cognitive process, allowing the AI model to navigate complex tasks with ease. By using a sequence of prompts, you can thoroughly guide the AI through each aspect of the task.  

Example Prompt Chaining Sequence: 

Step 1: What are 3 high-level goals for a 4-part bootcamp on generative AI for associations? 

Step 2: Break the first goal into tasks. 

Step 3: Start accomplishing one of those tasks. 

Experimentation is Key to Making AI Work for Your Specific Situation

To make AI work for you and not against you, experimenting with prompt engineering is key. With a little practice, you'll be a master at crafting prompts that generate valuable content catered to your association.  

We should also mention you're only scratching the tip of the AI iceberg! To dive deeper into prompting, download Betty Bot's Guide to Prompt Engineering and unlock a world of AI potential for your organization. Happy prompting!