As AI continues to revolutionize various sectors, many in the nonprofit domain are seeking practical ways to leverage this technology.  A prompt is more than just a question or instruction; it’s a guiding beacon for the AI, steering it towards the desired response. The accuracy and utility of the AI’s output hinge significantly on how well these prompts are crafted. A vague prompt might lead to irrelevant or inaccurate results, while a meticulously designed prompt can yield valuable insights.

 A compelling approach to structuring prompts is  Chain of Thought Prompting, which breaks complex tasks into manageable steps to achieve more accurate results. This method, developed by Jason Wei and a team at Google in 2022, is especially effective for tasks requiring detailed reasoning.

“Using Chain of Thought Prompting can transform the way nonprofits interact with AI, making complex tasks more manageable and less intimidating.”

Applying Chain of Thought in Non-Profit Operations

How can nonprofits use this approach to enhance their work? Let’s explore this with specific, easy-to-follow prompts.

Case Study 1: Meeting Notes Summarization

Initial Attempt:

My first attempt at using ChatGPT-4 to summarize meeting notes from a transcript of an audio file with AI was not effective. I uploaded a summary of a previous board meeting, to show the results I wanted, and then uploaded a written transcript of an audio file of the meeting I wanted to summarize. Unfortunately, the output from the AI was filled with errors, had significant omissions, and generally did a poor job of capturing the content of the meeting. Even as I kept adding new instructions to refine the output, the results were poor. It was frustrating. 

Revised Approach:

I then tried a more structured approach, using the sequenced-based structure of a Chain of Thought prompt. I wondered if the approach of giving the AI a more broken-down set of tasks, and going step-by-step in one chat box, would result in a better output. I uploaded the previous meeting summary, and the transcript of the audio file of meeting notes I wanted summarized.

Here’s a streamlined summary of the prompt I used (with the headers that I included in the instructions):: 

  • “Review Transcript: “Please read this entire transcript to understand the meeting’s context.
  • Identify Key Points: Highlight the major topics, decisions, and follow-up steps mentioned in the transcript.
  • Summarize Discussions: Based on the key points, create a structured summary.
  • Model Summary: Refer to these notes from a previous meeting for format and style.
  • Final Summary Creation: Using the steps above, compile a comprehensive summary of the meeting, suitable for nonprofit professionals.

This multi-step prompt significantly improved the AI’s output, making it more relevant and detailed. ChaptGPT 4’s final result was exactly what I wanted–an updated summary and thorough overview of the recent board meeting. This time there were no significant errors, and the content mirrored most of the meeting. When I asked the AI to go back and check the notes because it had not included the introduction of a new staff member, it added that without any issues.

Simplifying Prompts for Non-Technical Users

This use case showed me that it can be highly effective to break a multi-step task down into a sequence of straightforward instructions that the AI can easily follow. 

A crucial tip for crafting effective AI prompts, especially in a nonprofit context, is the value of specificity. The more precise and detailed your instructions are, the better the AI can understand and execute your request. Think of this as finding ways to be more specific in your ask, and more concrete; Here are some examples related to the use cases I’m sharing here:

  • Meeting Notes Summarization: Instead of a broad prompt like “summarize this meeting,” I saw significantly improved results with specific, guided steps. For instance, asking the AI to “identify and highlight major decisions and follow-up actions in the transcript” led to a more targeted and useful summary.
  • Crafting a Fundraising Letter (described below): Rather than simply asking the AI to “write a fundraising letter,” providing clear, specific,  instructions, such as “analyze this draft and suggest improvements based on our organization’s recent achievements,” resulted in a more impactful and relevant letter.

Remember, the key is to guide the AI through the task with the same level of detail and instruction that you would use to instruct a colleague who’s unfamiliar with the project. This approach ensures that the AI’s responses are aligned with your specific needs and objectives.

Case Study 2: Crafting a Fundraising Letter

I wanted to test the approach further, so I had ChaptGPT-4 work with me on an end-of-year fundraising letter. This time, I uploaded a file that included the organization’s mission and vision, as well as a list of accomplishments from 2023. I used the following prompt to have it summarize key information: 

“First, please look at the materials in your LLM and summarize elements that you see in effective non-profit fundraising letters for youth-centered arts organizations. Write a brief summary of the key points, less than 100 words.

Next, read and review this organization’s mission and vision and their accomplishments from 2023 and highlight 3 achievements that seem important within a fundraising and call to action context.”

The AI responded to these prompts and I read what it had posted. Then, in the same chatbox, I attached a PDF of a draft of a current end-of-year-giving letter that I had written for the organization. I then shared the following prompt:

“Please analyze this letter and what you think is effective about it and identify what could be more effective, based on the work you have done previously in this chat box and with these prompts and materials. After that, please itemize your observations and findings.”

In response, the AI then clearly outlined what it thought worked in the letter and why, and what did not and why. The work was excellent.

Next, I asked the AI to suggest a new draft of my letter that would combine the best elements from all of its work, and minimize the identified weaknesses. My prompt was:

“Combine the strong elements from the analysis with the summarized key information to draft a more effective new fundraising letter.”

These step-by-step prompts helped generate a strong fundraising letter, tailored to my needs, and in the organization’s voice. I took the AI’s version and added some of the suggested elements to my original, then completed the draft myself before sharing with the team.

Understanding Incremental Learning in AI

A key aspect to remember while working with AI is the concept of incremental learning. While AI tools like ChatGPT don’t retain personal data across sessions, they excel at building upon the information provided within a current interaction. This means you can progressively refine your prompts based on the AI’s responses in real-time. 

For example, in our fundraising letter case, I started with a basic prompt and then iteratively refined it, incorporating the AI’s feedback to fine-tune our final output. This approach encourages a dynamic interaction with the AI, treating it more as a collaborative tool that adapts and responds to your evolving needs within a single session. This iterative process is particularly useful in complex tasks where initial prompts may require adjustments to achieve the desired outcome.

Concluding Thoughts

Learning the art and skill of effective prompt writing and being prepared to iterate, edit and revise, are two key factors to being successful in using AI to help streamline your work.  The AI is not going to do all the work for you and deliver miraculous, perfect documents, but it can significantly save time in crafting drafts, building outlines and suggesting revisions and new approaches.

Also, as you work on a specific task or problem, remember to notice what the title is of the chatbox you are working in, so you can return to it next time you need to do this task.

The art of using AI effectively involves iterative refinement — a process where initial instructions are fine-tuned, gradually honing the AI’s output to precisely match your specific needs. This skill, combined with careful prompting, can transform your AI into a very useful tool, indispensable for core  tasks in our non-profit work..

While the specific examples and prompts in this post can serve as a starting point, I encourage you to adapt and refine these prompts to fit your own scenarios. Remember, AI is a tool to aid your creativity and efficiency, not replace it.

How are you using AI in your work? Let me know!

I’d love to hear how you’ve used AI in your non-profit work. Your experiences, questions, and insights are invaluable and can help others in our community embrace this technology.

 

Many thanks to my colleague Sharyl McGrew, for feedback on an earlier draft.

This post is part of a series called PROMPTS THAT WORK: Getting Started With AI. 

Other posts include:

As AI continues to revolutionize various sectors, many in the nonprofit domain are seeking practical ways to leverage this technology.  A prompt is more than just a question or instruction; it’s a guiding beacon for the AI, steering it towards the desired response. The accuracy and utility of the AI’s output hinge significantly on how well these prompts are crafted. A vague prompt might lead to irrelevant or inaccurate results, while a meticulously designed prompt can yield valuable insights.

 A compelling approach to structuring prompts is  Chain of Thought Prompting, which breaks complex tasks into manageable steps to achieve more accurate results. This method, developed by Jason Wei and a team at Google in 2022, is especially effective for tasks requiring detailed reasoning.

“Using Chain of Thought Prompting can transform the way nonprofits interact with AI, making complex tasks more manageable and less intimidating.”

Applying Chain of Thought in Non-Profit Operations

How can nonprofits use this approach to enhance their work? Let’s explore this with specific, easy-to-follow prompts.

Case Study 1: Meeting Notes Summarization

Initial Attempt:

My first attempt at using ChatGPT-4 to summarize meeting notes from a transcript of an audio file with AI was not effective. I uploaded a summary of a previous board meeting, to show the results I wanted, and then uploaded a written transcript of an audio file of the meeting I wanted to summarize. Unfortunately, the output from the AI was filled with errors, had significant omissions, and generally did a poor job of capturing the content of the meeting. Even as I kept adding new instructions to refine the output, the results were poor. It was frustrating. 

Revised Approach:

I then tried a more structured approach, using the sequenced-based structure of a Chain of Thought prompt. I wondered if the approach of giving the AI a more broken-down set of tasks, and going step-by-step in one chat box, would result in a better output. I uploaded the previous meeting summary, and the transcript of the audio file of meeting notes I wanted summarized.

Here’s a streamlined summary of the prompt I used (with the headers that I included in the instructions):: 

  • “Review Transcript: “Please read this entire transcript to understand the meeting’s context.
  • Identify Key Points: Highlight the major topics, decisions, and follow-up steps mentioned in the transcript.
  • Summarize Discussions: Based on the key points, create a structured summary.
  • Model Summary: Refer to these notes from a previous meeting for format and style.
  • Final Summary Creation: Using the steps above, compile a comprehensive summary of the meeting, suitable for nonprofit professionals.

This multi-step prompt significantly improved the AI’s output, making it more relevant and detailed. ChaptGPT 4’s final result was exactly what I wanted–an updated summary and thorough overview of the recent board meeting. This time there were no significant errors, and the content mirrored most of the meeting. When I asked the AI to go back and check the notes because it had not included the introduction of a new staff member, it added that without any issues.

Simplifying Prompts for Non-Technical Users

This use case showed me that it can be highly effective to break a multi-step task down into a sequence of straightforward instructions that the AI can easily follow. 

A crucial tip for crafting effective AI prompts, especially in a nonprofit context, is the value of specificity. The more precise and detailed your instructions are, the better the AI can understand and execute your request. Think of this as finding ways to be more specific in your ask, and more concrete; Here are some examples related to the use cases I’m sharing here:

  • Meeting Notes Summarization: Instead of a broad prompt like “summarize this meeting,” I saw significantly improved results with specific, guided steps. For instance, asking the AI to “identify and highlight major decisions and follow-up actions in the transcript” led to a more targeted and useful summary.
  • Crafting a Fundraising Letter (described below): Rather than simply asking the AI to “write a fundraising letter,” providing clear, specific,  instructions, such as “analyze this draft and suggest improvements based on our organization’s recent achievements,” resulted in a more impactful and relevant letter.

Remember, the key is to guide the AI through the task with the same level of detail and instruction that you would use to instruct a colleague who’s unfamiliar with the project. This approach ensures that the AI’s responses are aligned with your specific needs and objectives.

Case Study 2: Crafting a Fundraising Letter

I wanted to test the approach further, so I had ChaptGPT-4 work with me on an end-of-year fundraising letter. This time, I uploaded a file that included the organization’s mission and vision, as well as a list of accomplishments from 2023. I used the following prompt to have it summarize key information: 

“First, please look at the materials in your LLM and summarize elements that you see in effective non-profit fundraising letters for youth-centered arts organizations. Write a brief summary of the key points, less than 100 words.

Next, read and review this organization’s mission and vision and their accomplishments from 2023 and highlight 3 achievements that seem important within a fundraising and call to action context.”

The AI responded to these prompts and I read what it had posted. Then, in the same chatbox, I attached a PDF of a draft of a current end-of-year-giving letter that I had written for the organization. I then shared the following prompt:

“Please analyze this letter and what you think is effective about it and identify what could be more effective, based on the work you have done previously in this chat box and with these prompts and materials. After that, please itemize your observations and findings.”

In response, the AI then clearly outlined what it thought worked in the letter and why, and what did not and why. The work was excellent.

Next, I asked the AI to suggest a new draft of my letter that would combine the best elements from all of its work, and minimize the identified weaknesses. My prompt was:

“Combine the strong elements from the analysis with the summarized key information to draft a more effective new fundraising letter.”

These step-by-step prompts helped generate a strong fundraising letter, tailored to my needs, and in the organization’s voice. I took the AI’s version and added some of the suggested elements to my original, then completed the draft myself before sharing with the team.

Understanding Incremental Learning in AI

A key aspect to remember while working with AI is the concept of incremental learning. While AI tools like ChatGPT don’t retain personal data across sessions, they excel at building upon the information provided within a current interaction. This means you can progressively refine your prompts based on the AI’s responses in real-time. 

For example, in our fundraising letter case, I started with a basic prompt and then iteratively refined it, incorporating the AI’s feedback to fine-tune our final output. This approach encourages a dynamic interaction with the AI, treating it more as a collaborative tool that adapts and responds to your evolving needs within a single session. This iterative process is particularly useful in complex tasks where initial prompts may require adjustments to achieve the desired outcome.

Concluding Thoughts

Learning the art and skill of effective prompt writing and being prepared to iterate, edit and revise, are two key factors to being successful in using AI to help streamline your work.  The AI is not going to do all the work for you and deliver miraculous, perfect documents, but it can significantly save time in crafting drafts, building outlines and suggesting revisions and new approaches.

Also, as you work on a specific task or problem, remember to notice what the title is of the chatbox you are working in, so you can return to it next time you need to do this task.

The art of using AI effectively involves iterative refinement — a process where initial instructions are fine-tuned, gradually honing the AI’s output to precisely match your specific needs. This skill, combined with careful prompting, can transform your AI into a very useful tool, indispensable for core  tasks in our non-profit work..

While the specific examples and prompts in this post can serve as a starting point, I encourage you to adapt and refine these prompts to fit your own scenarios. Remember, AI is a tool to aid your creativity and efficiency, not replace it.

How are you using AI in your work? Let me know!

I’d love to hear how you’ve used AI in your non-profit work. Your experiences, questions, and insights are invaluable and can help others in our community embrace this technology.

 

Many thanks to my colleague Sharyl McGrew, for feedback on an earlier draft.

This post is part of a series called PROMPTS THAT WORK: Getting Started With AI. 

Other posts include: