Community IT Innovators Nonprofit Technology Topics

Nonprofit AI: Vetting AI Tools for Nonprofits; What Are AI Agents?

• Community IT Innovators • Season 7 • Episode 14

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 18:25

Nonprofit AI: Vetting Your Tools and Understanding AI Agents

In this midweek AI check-in, Carolyn Woodard explores how nonprofit leaders can make informed, safe decisions in an increasingly crowded technology landscape. The episode begins with a review of a new AI Tools Safety Guide specifically designed for the nonprofit sector. This resource evaluates popular tools based on the criteria that matter most to mission-driven organizations: data privacy, security, and ethical responsibility.

Carolyn also demystifies the concept of AI Agents. By breaking down the hierarchy of AI—from simple bots and assistants to more autonomous agents—she explains how these specialized tools can eventually handle the busy work of repetitive tasks, such as cross-referencing spreadsheets or organizing files. 

Whether you are just starting to draft your organization's AI usage policy or you are looking for ways to streamline your internal workflows, this episode provides practical guardrails for navigating AI adoption at nonprofits.

Featured Resources

  • Tool: AI Tools Safety Guide for Nonprofits Visit Meet the Moment A searchable directory of AI tools evaluated for trust and safety by reputable nonprofit technology experts. It’s an ideal starting point for organizations with strict data handling needs and limited research time.
  • Template: AI Acceptable Use Policy Download from Community IT IT is better to establish clear principles now than to wait for time to make a perfect policy. This template helps you communicate expectations to your staff and board regarding the use of generative AI.
  • Community Discussion: Nonprofit IT Management Join the Reddit Community Have a specific question about an AI tool or a repetitive task you'd like to automate? Connect with Carolyn and other nonprofit professionals on our dedicated subreddit to share insights and ask questions.


_______________________________
Start a conversation :)

Thanks for listening. 


Carolyn Woodard

Hello and welcome to the Community IT Innovators Technology Topics Podcast, midweek nonprofit AI check-in. I'm your host, Carolyn Woodard, and I am the outreach director for Community IT. And as I say every week, I am not an AI expert. None of us are, so we're going along on this journey together and hoping to create community and share resources and provide some explanations and answer some questions on nonprofits and AI.

Carolyn Woodard

So if you do have questions you'd like answered, you can reach us on Reddit r slash nonprofit IT management. You can contact me through our website, communityit.com. Just use the contact us form and uh put in your message there that you have a question you'd like answered on nonprofit AI.

Carolyn Woodard

Today I'm gonna give us an interesting new tool that I have come across. Um it's an AI safety guide for nonprofits, and I will share the link in the show notes. It's from Meet the Moment, which is a startup uh hoping to help guide nonprofits through the AI jungle, basically, uh started by Josh Peskay and Kim Snyder, who do the AI courses through uh NTEN and TechSoup. Uh so they're very reputable within the nonprofit technology space. Uh they do have courses available. I think the ones on TechSoup are on demand. If you're a member, I'm not sure if they're free. If they aren't, they're very affordable.

Carolyn Woodard

But this tool I wanted to share is on their website. It's the AI Tools Safety Guide for Nonprofits, a searchable directory of AI tools evaluated and rated for trust, safety, privacy, and responsibility. Make informed decisions about which AI tools are right for your organization at the moment. They have 16 tools that they have evaluated, but they are doing more every day.

Carolyn Woodard

And what I like about this uh guide is that they clearly lay out their criteria. Um, they clearly designate uh the rating fees and paid tiers, uh, and they just doing the research, which is you know, who has time for that? So they do these evaluations and it's uh specific to nonprofit organizations with data handling needs and you know uh budget constraints and time constraints. Uh, every tool is evaluated against criteria that matter to nonprofits, data privacy, security, terms of service, and ethical considerations, which is something we've been talking about quite a bit on this midweek check-in on nonprofit AI. Um, they have the ratings.

Carolyn Woodard

um, they have a policy generator as well that's available here. So you can uh generate an AI usage policy for our organization if you haven't done that yet. We do have a template on our website as well for AI acceptable use policy, but you know, you can also Google it and uh come up with a policy. Uh, it can be as complex or as simple as makes sense for your organization. I would say that if um I would get started and wait to make a perfect, not wait to start, because I guarantee you people on your staff are already using AI, and so you want to make sure that you're uh um explaining what your principles are, even if you don't have a full-on policy yet, but uh how you want to be using AI where you don't want AI to be used, and make sure that you're communicating that clearly with your staff and your leadership and your board.

Carolyn Woodard

Um, and the other thing I like about this AI safety guide from um Meet the Moment is that it is being updated all the time. So uh that is a very helpful uh thing as well. So I will share that link with you in the show notes. Um, that was just a nice uh resource that I wanted to share. Um if you've used it, please get in touch and let me know how it worked for you.

Carolyn Woodard

Uh I also wanted to take a moment in this podcast to talk about what an AI agent is. I've got some questions about that um over the past couple of weeks, uh, that it's not all that clear. So we're trying to clarify.

Carolyn Woodard

We've already talked in one of these podcasts about the difference between public AI and enterprise AI. So if you're paying for a subscription, like you yourself might not be paying for that subscription, it may be through your company or your school. But if you're logged in to your company, Microsoft or your company Google Workspace or the Claude subscription for your company or the Perplexity account that's for your company, and you're logged in, you're paying for a license, someone is paying for a license, you you can check the terms and conditions, but in general, you will be in a more secure environment. It will only be using the, it will not be uploading your questions and interactions and files that you upload and data that you're using to its larger public language models that it is using. So it will not have access to your private documents and your private data to your company, uh, private to your company.

Carolyn Woodard

So that's a totally different. We'll do a different podcast on that of whether you should be using your company uh co-pilot or Gemini to plan your vacation. That's a maybe it's uh depends on your uh policies and acceptable use. Um, but just you know, like anything else that you would do on your company computer, know that uh other people at your company are going to be able to see it potentially. Um, and just be careful, you know, follow those terms and conditions for not impersonating someone or creating, you know, images or documents that are harmful or harassing. Um, all of those will be covered by your general acceptable use policy, maybe in your staff hand employee handbook. So um check with that before you embark on any nefarious um tasks.

Carolyn Woodard

But uh to go back to the question of what is an AI agent, so you may have been hearing this term, AI agent, um, and uh it's uh it can be a little bit complicated, and of course, in technology, everybody uses a lot of lingo and talks about um, you know, things in the most complicated way.

Carolyn Woodard

But I think the way I've come to understand it um most easily is that an AI agent is kind of like what it sounds like. So it's an AI tool, like a mini tool almost that is built to do a specific task.

Carolyn Woodard

Um so it's not a generalist, uh, it's not, you know, like Chat GPT where you can put in any question and get any answer. Um, it may be designed to do a very specific thing at your nonprofit. And it can be designed by AI. Like you can go into an AI prompt in Copilot and say, hey, Copilot, I want to do this thing.

Carolyn Woodard

Like I want to connect up this spreadsheet to a SharePoint file and then have it published to this other SharePoint folder that my colleagues can see it in. And then that might be a task that an agent, Copilot, could help you design an agent to be able to do. And then once the agent can do it, um, it can just do it.

Carolyn Woodard

So uh you need to be careful, it can be extremely helpful and useful. It can automate uh some repetitive tasks or tasks where you know you're just doing the same thing. It doesn't really take a human to connect up this spreadsheet to this other place you want it to appear. Um, so you can build an AI agent that can do that for you.

Carolyn Woodard

The capability of the agent is of course using the same multimodal capacity of generative AI. So AI agents can process multimodal information like text, voice, video, audio, code, and it can converse with you, it can reason, it can learn, it can make its own decisions, it can learn over time. So the more it does the task, uh the more it can tell you about the task that it's doing. It may even suggest improvements. You may build your agent in such a way that it improves itself.

Carolyn Woodard

Agents can work with other agents to coordinate and perform more complex workflows.

Carolyn Woodard

Some of the things that the agent can do, of course, it can do quote-unquote reasoning. It can use logic and available information to draw conclusions, make inferences, and solve problems, just as you know the chatbots are doing for you if you're help if they're helping you write something. Um, they're drawing conclusions and uh figuring out which group of words should come after that group of words.

Carolyn Woodard

You can build AI agents with strong reasoning capabilities. They can help analyze data, they can identify patterns, they can make informed decisions based on that evidence. The agent can take actions or perform tasks based on decisions, plans, external input, or internal input. These can include physical actions. In the case of embodied AI, probably not a lot of nonprofits are going to be building those little robots that can do their own thing.

Carolyn Woodard

Um, but one of the things that an AI agent can do is trigger other actions. So it might trigger you to say, hey, I've completed this draft that you needed, you need to edit it. So you can build an AI agent that would trigger you. It could trigger a different AI agent to take over the task and complete something much more complex.

Carolyn Woodard

AI agents can observe, they gather information about the environment and the situation that they're working in. That might be your files, it may be your database, it may be for nonprofits. You know, you can think of lots of uses where you might build an agent that can look through your grants that you've given or grants that you've received and find patterns. Um, look for maybe the people that aren't getting grants, that should be getting grants, like finding patterns where maybe you're finding cases of bias that you didn't even know was part of your operations or your programs and alerting you to those. So you can build an AI agent that would find something like that.

Carolyn Woodard

Um, and it can do it over and over. It can do repetitive um tasks for you.

Carolyn Woodard

Um, it can help you with planning.

Carolyn Woodard

Um, it can you know collaborate, of course, uh with you. You can use uh AI as a thought partner, you can program it to do Socratic questioning of you, like why is this your goal, or why do you have that as your time frame, or um you know, what event are you trying to coordinate with? All of those types of questions.

Carolyn Woodard

And of course, AI is self-refining and it can uh carry out self-improvement and adaption, it can learn from experience, it can adjust its behavior based on your feedback, and it can continuously enhance its performance and capability.

Carolyn Woodard

So, what are the differences between AI agents, AI assistants, and bots? Um, I think I'll start with the you know, the kind of lowest level uh would be the bots, which are automating simple tasks capable of having simple conversations. You probably know if you've interacted with chat bots, um, either that are run through, you know, some like your airline that you're trying to interact with or customer service, that uh they can be very uh simple. They can be more complex, but uh often they're they're pretty simple. They may get distracted easily by your question. If your question is too complex, hopefully they're programmed so that in that case they call an actual agent who can help you. Or if you're trying to do something that's not typical, they might need additional help to do that.

Carolyn Woodard

They follow predefined rules, they have limited learning on what they can do independently from what you're telling them to do. Uh, and they're very reactive. They respond to triggers, they respond to commands. So when you're putting the prompt in, they're responding to that. They may prompt you at the end, like, do you want me to help you with this other thing? Uh that's because they've been programmed to find, you know, to ask you to stay involved with them and uh and stay engaged.

Carolyn Woodard

An AI assistant, um, the purpose is to assist users with a task. So it's not automating the task, it's being an assistant - in the name. Its capacity is that it can respond to requests or prompts, providing information and completing simple tasks. It can recommend actions, but in general, the user is the one who makes the decision about what happens next.

Carolyn Woodard

So it might prompt you, you know, with some bag that's on sale that you previously asked it to look for, but it wouldn't purchase the bag for you if it's helping you with shopping. Um, if you are using an AI assistant for your scheduling, for example, it might surface like, oh, there's a conflict, you can't have lunch with this person while you're getting on the flight for this, you know, catching your flight. Um, but it wouldn't like then go into your reservation and uh get you a later flight in the day or anything like that. It would prompt you to say, you know, do you want to change this? Can I help you change your schedule? Um, it is also reactive, it's responding to your programming and your requests.

Carolyn Woodard

An AI agent is kind of the highest on this hierarchy chain of being able to do things by itself. It is autonomous and can proactively perform tasks. Um, it can perform complex multi-step actions, it can learn and adapt, and it can make decisions independently. And its interaction is proactive and goal-oriented.

Carolyn Woodard

Um, so it has the highest degree of autonomy, it's able to operate and make decisions independently of your input, basically. Once it gets started on its path, it does its thing. And um you need to define its role, uh, its task that it's doing, specific instructions, descriptions of the tools that it has, and then you know, the other, if it's working with other agents, what sorts of things it's going to do.

Carolyn Woodard

If this all sounds super complicated, you know, we're at the beginning of this.

Carolyn Woodard

Um, but I will say that if you have kind of ongoing busy work, repetitive tasks at your organization, that is something that you know takes a person, staff member, you know, a couple of hours a month and they have to do it every month. It's like, well, we have to download this spreadsheet from this place, and then we have to check it to make sure that it's accurate. We have to input this spreadsheet from this other place, and then all of that gets turned into a spreadsheet or PowerPoint that we, you know, send to our board member. And that takes us like half a day every month because we have to check all of these pieces.

Carolyn Woodard

That might be a task where an agent, you might want to ask an AI about creating an agent to be able to do that task. Um, and you know, follow your guidelines, follow your policies.

Carolyn Woodard

Uh, we recommend for nonprofits that uh when you're using AI, that a human is your final editor, has the final eyes on it. So when you create an agent that does a task, you might build that into its workflow. That it's last, the last thing that it does as part of the task is alert you that you have something you have to review before it can go forward. Um, so those are kind of safeguards that you can build into the agent.

Carolyn Woodard

If you're interested, I urge you to investigate, uh, ask your AI tool that you're familiar with, and uh enjoy using about uh some of your repetitive tasks that you might have, and if an agent could be created to do that task for you, again, um making sure that at the end you're the last eyes on the task and that you you are the final editor. Um, but that's in a nutshell kind of the difference between just using AI tools. A lot of our tools have AI built into them, and creating an agent for nonprofit AI.

Carolyn Woodard

So just a little intro if you're interested, get in touch. Uh, look forward to hearing from you. You can hear from me on Friday with our regular podcast on different technology topics. And I'll be back here at midweek on Tuesdays with more on nonprofit AI. Take care.