For Impact

Blog

Five ways we use AI at For Impact

WOW Email | | Nick Fellers

We’re well into the hype cycle of AI. However, when we survey clients and training attendees less than 20% say they’re actively and deliberately using AI in their day-to-day workflow so we think it makes sense to highlight some of the ways we use AI to assist our fundraising and workflow.

Primer note: This is not an AI literacy article. I’m assuming some level of AI knowledge, but if all of this is new to you, here are a few terms to get you started:

  • GenAI. The field of AI has been around for a while. When you hear it conversationally used in 2024, most people are talking about Generative AI, the broad set of technologies that are capable of generating new content. At the time of this writing, these aren’t actually intelligent in the way humans are intelligent – instead, they’re predicting patterns, such as what a good sentence might look like, or what a picture of a bear shouting, “Be For Impact!” might look like.
  • More on Gen AI tools. You can access many tools for free and/or you can pay a monthly fee to access a premium version that is better and has more features (such as the ability to search the web in real time). We use the big three plus one: ChatGPT (from OpenAI), Claude.ai, and Google Gemini (and perplexity.ai, which is sort of a search engine that uses ChatGPT and Claude.ai). We also use a number of design tools (often built on top of OpenAI).
  • Trial and error. All of our prompting takes some amount of trial and error to refine the right prompt. It’s rare that we get magic on the first try, but the overall process is still a huge productivity boost. It’s also important to take into consideration hallucinations and bias.

Here is a video walkthrough of this article to demonstrate some of these examples.

Five applications of AI at For Impact

  1. Generative.
    Technically, all the use cases I’m sharing are ‘generative’ but I want to focus on using AI to generate ideas quickly, or advance your starting point.

    Critics of AI are usually quick to point out imperfections. Often they will say something like, “It’s not ready for prime time.” Or, “I just don’t see how this replaces what a human can do.”

    We ask them to think about these tools in a different way – not as replacements but as brainstorming accelerators. They can be used to generate ideas very quickly. Your role then becomes that of a curator, selecting the very best ideas to incorporate or enhance.

    Examples we use every day to advance starting points: Assemble job descriptions, write predisposition letters, identify possible funding rationales

    You can ‘prime’ the generation by asking AI by feeding it some examples or parameters. Examples:

    • Upload four predisposition letters you’ve used in the past. Then point AI to the website of a specific foundation and ask it to write a predisposition letter for the foundation.
    • Ask AI to read For Impact’s take on a funding rationale and then read your website and create three possible funding rationales.
    • Feed AI your existing job descriptions and ask it to generate a new job description for a new role based on the style of your previous job descriptions.

      Our Innovation and Design team uses generative AI design tools to develop 8+ initial concepts for engagement tools. AI has doubled our design productivity!

  2. Prospect Research.
    We use AI tools extensively in our prospect research – specifically to help identify prospects and to synthesize areas of alignment. This research can be hit or miss depending on the prompt, depending on the tool, and depending on the ever-changing limits around the tool – we really have to play around with various prompts.

    Some specific prompt examples:

    • “Identify 10 peer (nonprofit) organizations of XYZ organization.”
      This gives us a great list to start to do prospect research. We can then dive through the websites, reports, etc. to see who is funding in this space. To be sure, the names are just leads. They all need some level of research and strategy to determine quality.
    • “Who are top funders to XYZ cause?”
    • Or, you can get more specific.
      “Make a list of the top 100 philanthropists in the US. Next, tell me the top funders to XYZ cause that are NOT on the top 100 list.”
    • “Help me identify 20 potential funders for xyz organization.”
    • “Help me identify 10 funders already supporting xyz organization.”
  3. Editorial.
    This is the use case familiar to most readers.

    Some simple prompts (paste your copy into body of prompt or attach file):

    • “Read this email. I don’t want to change anything, just correct for really obvious spelling or grammatical errors.”
    • “My grant narrative is 1000 words. I need to get this down to 500. Can you take a shot at this?”
    • “Here is my email, rewrite it very concisely – like Hemingway.”
  4. Computational.
    This is what it sounds like. AI can do computations, including coding, much faster than we can.

    We use AI to build applications and crunch data. The days of this use case are early going but the horizon is very clear.

    I would encourage everyone to watch three minutes of Jensen Huang’s interview at the World Government Summit. “It is our job to create computing technology such that nobody has to program and that the programming language is human. Everybody in the world is now a programmer.”

    The following is a very recent example where we used AI to sift through thousands of grant opportunities to identify the 3% that were relevant to our research. Historically, this would’ve been a two-day project for an intern, or we would’ve needed someone to write some code with many conditions.

    Example: A large international foundation publishes all of its grants into a spreadsheet containing 5000+ records. We asked AI to help us find only the most relevant examples and within a moment it returned a new spreadsheet with 140 highly relevant records.

  5. Learning.
    In any given year, our team traverses 40+ causes across 30 countries, which means we need to become subject-matter-literate very quickly. Acknowledging that we need to be on the lookout for hallucinations and bias, AI is a handy copilot and jumping off point.

    Examples:

    • Summarize [title] book in under 1000 words. What are the three big things I need to take away from this?
    • (Recently building a case for early intervention)
      Is there any research or evidence to make the case that for every $1 invested in early childhood intervention or prevention, $x is saved?
    • Teach me about the origins of the Donor-Advised-Fund (DAF). After you do that, identify some recent trends for me, and try to attribute those trends to any sources that may be helpful for me to read.

Finally, while this is just a quick survey into our everyday use cases, I would also point everyone to our friends at Fast Forward who are at the forefront of AI for Humanity. Follow Fast Forward to learn more about how innovative AI is being used to impact literacy, health, climate change and more!