# Nonprofit AI: Jet Fuel Trade-offs, Agentic AI, Work and Learning Styles

An episode of [Community IT Innovators Nonprofit Technology Topics](https://invisiblenetwork.ai/podcast/1536843269), 11 Aug 2026, 32 min.

Audio: https://www.buzzsprout.com/1029925/episodes/19633326-nonprofit-ai-jet-fuel-trade-offs-agentic-ai-work-and-learning-styles.mp3

Carolyn Woodard covers four AI stories with real implications for nonprofits this week, starting with Meta CEO Mark Zuckerberg's sprawling new manifesto on open source AI and the release of Muse Glimmer, a lightweight model anyone can run locally. Carolyn touches on why this matters for budget conscious nonprofits weighing vendor values, and why open source still requires more technical capacity than most organizations currently have on hand. From there, Carolyn digs into a story about Alaska Airlines using an AI tool called Flyways to suggest more efficient flight routes, saving fuel while keeping a human dispatcher in the loop on every final call. She uses it to revisit a three filter framework for thinking through AI's environmental tradeoffs: whether the benefit you are getting is worth the cost. It is a useful model for any nonprofit trying to weigh AI's real impact rather than just reacting for or against it. Next, a refresher on agentic AI, prompted by a listener question after last week's episode. Carolyn breaks down the difference between old school recommendation engines, generative AI, and agents that can take multi step actions on their own, plus what those tools are actually called inside Microsoft, Google, OpenAI, and Anthropic products. She also unpacks two recent incidents where AI agents from OpenAI and Anthropic got loose due to human error, and what that means for how carefully nonprofits should scope permissions before saying yes to an agent, and urges you to check with your policy and your IT team if you are unsure about the parameters of anything you are allowing an AI tool to do. Finally, Carolyn pushes back a little on the popular advice that everyone should write their own first draft before bringing in AI. She argues it really depends on your task, your learning style, and how you think best. Resources Mentioned: • Zuckerberg manifesto pushes an open source approach on AI as Meta releases its latest model – AP via ABC News – https://abcnews.com/Technology/wireStory/zuckerberg-manifesto-pushes-open-source-approach-ai-meta-135519669 • The FAA Wants to Reboot the Nation's Airspace. This Airline Shows How It Might Work – NPR – https://www.npr.org/2026/08/10/nx-s1-5872752/airspace-reboot-alaska-airlines-flyways • OpenAI's Hacking Debacle Was a Human Mistake – WIRED – https://www.wired.com/story/openais-hacking-debacle-was-a-human-mistake/ • OpenAI, Anthropic AI Hacking Sprees Illegal? – WIRED – https://www.wired.com/story/openai-anthropic-ai-hacking-sprees-illegal/ • Should You Really Use AI to Write That First Draft? (And What to Do Instead) – TechSoup – https://blog.techsoup.org/en-us/posts/should-you-really-use-ai-to-write-that-first-draft-and-what-to-do-instead New every Tuesday. _______________________________ Start a conversation :) • Register to attend a webinar in real time, and find all past transcripts at https://communityit.com/webinars/ • email Carolyn at cwoodard@communityit.com • on LinkedIn • on reddit/r/nonprofitITmanagement • on the Community IT website Thanks for listening.

## Transcript

[0:22] Carolyn Woodard: Welcome everyone to the Community IT Innovators Nonprofit AI Podcast midweek check-in on Tuesdays.
[0:29] Carolyn Woodard: My name is Carolyn Woodard.
[0:31] Carolyn Woodard: I'm your host.
[0:32] Carolyn Woodard: I'm not an AI expert.
[0:34] Carolyn Woodard: I'm very curious about AI and nonprofits and the roles and directions that AI is taking and some policy questions, news items.
[0:44] Carolyn Woodard: So every Tuesday I come in and just kind of catch up on what's going on and how it particularly relates to nonprofits.
[0:52] Carolyn Woodard: So I have today one of my first breaking news things, just as I am recording this, Mark Zuckerberg published yesterday published a big manifesto about open source AI.
[1:06] Carolyn Woodard: And he announced a new free lightweight meta-AI model that anyone can run locally, something we've been talking about on this podcast for a month or so now.
[1:16] Carolyn Woodard: The idea that you could have your local version and that would be more environmentally friendly and it would deal with some of the issues with power imbalances and cybersecurity and just some of the yucky things about the AI industry and the big tech companies that nonprofits feel conflicted about.
[1:36] Carolyn Woodard: So it's worth knowing about if you're nonprofit budget consciousness and thinking about all these issues around AI tools.
[1:44] Carolyn Woodard: Of course, it's Meta.
[1:45] Carolyn Woodard: So as always, weigh this against your usual vendor values checklist if you have that.
[1:52] Carolyn Woodard: If you don't, I talked about that a couple episodes back.
[1:55] Carolyn Woodard: So you can check that out.
[1:56] Carolyn Woodard: I'm going to keep an eye on this.
[1:58] Carolyn Woodard: Um,
[1:59] Carolyn Woodard: So far, my understanding is that again, so far, the uh do-it-yourself AI tools and models, the open source is requires some technology, some technical background.
[2:13] Carolyn Woodard: So for I don't know, 90% of nonprofits, it's out of the realm of possibility that you would really need a very techie person to be able to set it up and run it.
[2:23] Carolyn Woodard: Um, maybe that'll change.
[2:24] Carolyn Woodard: I have not looked at this tool yet, so I don't know exactly what's in it.
[2:28] Carolyn Woodard: I'm sure my nonprofit techie people out there are looking at it already and will be having some hot takes um in the upcoming couple of days.
[2:37] Carolyn Woodard: So if you haven't seen this news item yet, I'll have a link to some things about it in the show notes, but you can also just Google it.
[2:44] Carolyn Woodard: It is out there, it just happened.
[2:46] Carolyn Woodard: All right.
[2:47] Carolyn Woodard: So the first thing I wanted to talk about this week was the issue that, you know, air traffic controllers, there aren't enough of them for the airways, and that they do a lot beyond just the in the moment directions, you know, to keep the planes away from each other as they're landing.
[3:08] Carolyn Woodard: They also um, you know, they tell the planes what flight routes they are gonna take.
[3:13] Carolyn Woodard: They give them the information, here's how you're getting to where you're going.
[3:17] Carolyn Woodard: Um, so
[3:18] Carolyn Woodard: Alaska Airlines, the story was about this tool that they use called Flyways AI, uh, built by a company called Airspace Intelligence, which not surprisingly has a bunch of defense contracting, you know, contracts.
[3:31] Carolyn Woodard: But what this tool does is it crunches real-time data on air traffic, weather, wind patterns, et cetera, and it suggests the most efficient flight routes.
[3:40] Carolyn Woodard: Alaska Airlines uh reports saving tens of thousands of hours in the air, uh, millions gallons of fuel per year.
[3:48] Carolyn Woodard: Now, this is what they're reporting.
[3:50] Carolyn Woodard: There's no, I looked into you, like, could I actually find that there are some numbers that you can compare, um, but I couldn't find anything on that, of course.
[3:58] Carolyn Woodard: Um, the FAA has uh contracts with Airspace Intelligence to do similar predictive analytics to the whole US airspace under a program called SMART.
[4:12] Carolyn Woodard: The interesting, well, there are lots of interesting things about this.
[4:15] Carolyn Woodard: We talked before about, you know, if you're the frontline um people accepting phone calls about your shelter, like can you have an AI that can answer the most common questions so that you can focus on the difficult questions?
[4:28] Carolyn Woodard: And how do you control for that so that it isn't the AI making decisions, the human in the loop, the human is still making the final decisions, has the editorial um, you know, final say.
[4:41] Carolyn Woodard: And so the way the story is presented is that the dispatchers, you know, they already have, you know, like they're looking at five screens at once, and they're looking at the the um the air routes that the planes have inputted that they want to take to, you know, this is Alaska Airlines is going from, I don't know, Seattle to New York.
[5:03] Carolyn Woodard: And uh in the past, they had to make those decisions themselves, and you know, that's what they're trained to do.
[5:10] Carolyn Woodard: So this tool gives them suggestions, it prompts them, taking all these multiple things into effect, into taking these multiple things into account, which an air traffic controller, you know, they can synthesize, like the the weather plus other routes, plus you know, fuel and et cetera, efficiency, but the AI is really good at doing stuff like that, taking all of these inputs and putting out this suggestion, but then it is safeguarded in that it's just a suggestion.
[5:42] Carolyn Woodard: The air traffic controller has to decide which route to put that flight on.
[5:47] Carolyn Woodard: Um
[5:48] Carolyn Woodard: I just thought that this was interesting for another reason, besides the human in the loop.
[5:54] Carolyn Woodard: Like, don't worry if you're flying Alaska Airlines, uh, they're still making the decisions.
[5:59] Carolyn Woodard: Um, but I just thought it was interesting.
[6:02] Carolyn Woodard: We had talked about the three filter decision making process of, you know, does are what are you asking the AI to do?
[6:12] Carolyn Woodard: Where are you?
[6:14] Carolyn Woodard: So, how bad is your use of energy and the data centers?
[6:17] Carolyn Woodard: Uh, is your grid really stressed?
[6:19] Carolyn Woodard: Is your water under a lot of stress?
[6:21] Carolyn Woodard: And then the final part of the three filter model tool is are you getting a benefit from using the AI that is equal to or outweighs the environmental impacts that you're probably having as a way to think about it, right?
[6:42] Carolyn Woodard: So instead of just rejecting all AI, thinking about, and this comes from you know, environmental organizations that are really excited about using AI for a lot of things, you know, like animal, like endangered animal tracking or, you know, deforestation or you know, complex weather patterns or looking at um data from uh satellite images, like all of those things that AI is pretty good at, but then also feeling really conflicted that it's bad for the environment.
[7:13] Carolyn Woodard: And so trying to build into your decision-making process those benefits.
[7:19] Carolyn Woodard: And so this I thought is like kind of scratching that surface of, well, if the AI, you're using electricity and water to run this AI, that's pretty complex.
[7:29] Carolyn Woodard: So it's taking into a lot of these inputs and churning out this complex decision-making suggestion.
[7:37] Carolyn Woodard: Um, and but you're saving on jet fuel because you have all of these flights that are now going in these more efficient ways and taking into account weather and their route.
[7:51] Carolyn Woodard: So you could see where there could come a point where in fact you're saving enough in the jet fuel, which is also really bad for the environment, that it makes sense to use the AI tool to be more efficient in your planning of the route.
[8:06] Carolyn Woodard: And as I said, I couldn't find, I mean, I'm not a scientist at MIT and I don't have access to Alaskan Airlines models and where their data center is and what in how much it's actually costing to run the AI.
[8:19] Carolyn Woodard: But
[8:20] Carolyn Woodard: I think it's an interesting question to think about, maybe in a smaller scale for your nonprofit.
[8:26] Carolyn Woodard: Um, because, you know, I'm sure Alaska Airlines is looking at this, of it's worth it to pay it because we are saving on the jet fuel.
[8:34] Carolyn Woodard: Um, but for your nonprofit also, you could be looking at this at your scale of, you know, if I use this tool and I gain back two hours in my day, for example, that I used to have to spend, you know, making this spreadsheet.
[8:51] Carolyn Woodard: And now I can just, I've made a you know, widget in my AI tool and it can I put in the CSV files and it makes this spreadsheet for me.
[8:59] Carolyn Woodard: If you get two hours of your day back for that, I mean you could go get a massage or do yoga or something, which is also very valid, you know, fight and burnout right there.
[9:10] Carolyn Woodard: But um, if you had two hours to do human stuff at your nonprofit, what would you use that time for?
[9:19] Carolyn Woodard: And if that maybe the two hours that you get back isn't it's maybe you're using the AI not for productivity gains for yourself or your own day-to-day work.
[9:32] Carolyn Woodard: Maybe you're using this AI for something your mission does, something your nonprofit does.
[9:38] Carolyn Woodard: And you use the AI for a project to you know combine some new data that you had, but you weren't able to analyze before, or you combine census data with your population data and you identify, you know, 500 more kids that you could reach with a scholarship or something like that.
[9:59] Carolyn Woodard: Does that gain offset the cost to the environment, to your electricity grid, to um, you know, your way of working?
[10:11] Carolyn Woodard: So I don't know.
[10:12] Carolyn Woodard: I just I find it fascinating to think about.
[10:14] Carolyn Woodard: And I think for each organization, like that's gonna be a very local answer of how how can we quantify or qualify the benefits that are possibility, right?
[10:27] Carolyn Woodard: They're future benefits to our nonprofit of the way we operate, the things that we do, uh, versus the current electricity, environmental, and water damage that we may be causing now.
[10:41] Carolyn Woodard: And uh it's just something you're gonna have to work through and see is it a potential that you can get to?
[10:50] Carolyn Woodard: And is it a potential that is it an opportunity?
[10:54] Carolyn Woodard: Could your nonprofit be doing something more efficiently, more effectively, um, more holistically?
[11:01] Carolyn Woodard: Uh, could you do some things that you've always wanted to do, but just never had the manpower to be able to put all of those numbers into your program design, for example, or to your analysis or evaluation.
[11:12] Carolyn Woodard: So things to think about.
[11:14] Carolyn Woodard: You know, we often have the kind of what about is um when we think about and talk about using AI, uh, a lot of people will come back, if you have concerns about it, people will come back immediately with, well, you're using electricity anyway, like you have the lights on, you're flying to a conference, so you're using a, you know, you're using jet fuel, which is really bad for the environment.
[11:36] Carolyn Woodard: You are, you know, you have steak at your gala or you know, whatever it is that's wasteful.
[11:41] Carolyn Woodard: Uh, you're using a laptop that was made by, you know, child labor somewhere with terrible use of um, you know, cobalt or whatever.
[11:51] Carolyn Woodard: I guess where I'm going with this is there's always trade-offs.
[11:55] Carolyn Woodard: And I think that's the point of that third filter in the model is thinking about the opportunity that you have, you may have within AI to help you build your capacity or sub in some capacity for you that you're under-resourced and understaffed.
[12:16] Carolyn Woodard: Um, and would that use of the AI, like I said, is that offset?
[12:23] Carolyn Woodard: I mean, even this is a question about jet fuel, right?
[12:26] Carolyn Woodard: If the AI is helping the airline use less jet fuel, that's a good thing.
[12:32] Carolyn Woodard: So just the balance is never all on one side or all on the other side.
[12:37] Carolyn Woodard: It's always you have to weigh it together the consequences and your current actions.
[12:42] Carolyn Woodard: All right.
[12:43] Carolyn Woodard: So that was, I guess, a little deep dive into airlines.
[12:46] Carolyn Woodard: Um,
[12:47] Carolyn Woodard: I wanted to do a quick kind of explainer, refresher on the different types of AI that there are.
[12:54] Carolyn Woodard: Last week I talked about agentic AI and how the AI agents had gotten out, they'd gotten free.
[13:00] Carolyn Woodard: And uh I got a question from someone about, well, tell me again, like what are agents and am I using agents?
[13:06] Carolyn Woodard: Can I make an agent without even knowing it?
[13:08] Carolyn Woodard: And then it would escape and it would be my fault.
[13:10] Carolyn Woodard: So uh quick refresher.
[13:12] Carolyn Woodard: Um, I talked about something way back several months ago when we were talking introducing this podcast, um, the old AI, right?
[13:21] Carolyn Woodard: So if you use a recommendation engine, uh like Amazon would tell you, like if you liked this, you you know, might like these similar products.
[13:30] Carolyn Woodard: Uh, Netflix, again, giving you recommendations, um, GPS giving you routing advice.
[13:37] Carolyn Woodard: Um, and
[13:38] Carolyn Woodard: Of course, all of these things, like they've been around for decades and they all run on data centers.
[13:44] Carolyn Woodard: But don't, you know, not to put a fine point on it, but we've been using data centers for a lot of stuff.
[13:49] Carolyn Woodard: If you're looking at videos on YouTube, right, that all goes through data centers.
[13:53] Carolyn Woodard: That's why we can do that.
[13:54] Carolyn Woodard: Um, but no one called it AI at the time.
[13:57] Carolyn Woodard: And um, you know, the
[14:01] Carolyn Woodard: My understanding of how this flyways uh tool that Alaska Airlines is using is very similar to that GPS.
[14:08] Carolyn Woodard: You know, it's just taking a bunch more, it's very, it's more complex than a map in your car and where you're going to, but it's a very similar application.
[14:17] Carolyn Woodard: And that is, you know, it's predictive and it's uh analyzing a bunch of data points, but it's um it's not generative.
[14:28] Carolyn Woodard: So then the next step, of course, was this generative AI, which Chat GPT was launched.
[14:34] Carolyn Woodard: I don't know, it was several years ago now.
[14:37] Carolyn Woodard: And if you think back to what ChatGPT was like when it launched, and we all thought it was so cool.
[14:42] Carolyn Woodard: And now we'd be like, oh, that was like baby gen AI.
[14:45] Carolyn Woodard: But it can use natural language, you can ask it a natural question.
[14:50] Carolyn Woodard: It's using this predictive text where it like goes out to the data center with the packets of the words that you've put into your prompt or query, and then it's coming back with you, and it feels almost instantaneous that it's answering your question right away.
[15:04] Carolyn Woodard: And it's synthesizing a lot of things.
[15:07] Carolyn Woodard: It can do online research for you.
[15:09] Carolyn Woodard: It can look up URLs, it can do um, you know, when you put in a Google search, you're gonna get that AI suggested, uh, is this what you were looking for?
[15:19] Carolyn Woodard: This seems to be what you were looking for, type of thing.
[15:22] Carolyn Woodard: So that is Gen AI, I think is a lot of what we talk about in this podcast is how you can use Gen AI tools to do your work at a nonprofit.
[15:32] Carolyn Woodard: And then there's agentic AI.
[15:34] Carolyn Woodard: So it's been commercially available for a while.
[15:37] Carolyn Woodard: Um uh there's, you know, there's always this large Microsoft nonprofits conference uh every year.
[15:44] Carolyn Woodard: And two or three years ago that they were talking in that conference about all things agentic, like agents are just going to do everything at every nonprofit, which maybe eventually, but not right away, clearly.
[15:55] Carolyn Woodard: Um, but
[15:56] Carolyn Woodard: It's kind of the next step up when you're comfortable chatting and doing prompts and doing um AI tasks where you're in control of what it's doing, is um you can make these agents that can make their own decisions.
[16:11] Carolyn Woodard: You give them the parameters and then they do the thing and they are free to find the best way that they can do that thing.
[16:16] Carolyn Woodard: And they, I don't want to say quote unquote, they think about it, but they, you know, working with their model, they are able to um carry out actions.
[16:27] Carolyn Woodard: And then even more, another step up is you can chain them together so you can have different agents that do different pieces of a process, and then you put them all together and they do the process from start to finish.
[16:39] Carolyn Woodard: You can build into those agents that they have checkpoints where they have to check with you and they can't go ahead unless they have gotten a human yes-no answer, go, no, go.
[16:50] Carolyn Woodard: Um,
[16:52] Carolyn Woodard: And these agents in the tools you are probably already using uh are called different names.
[16:59] Carolyn Woodard: So they're not always called an agent.
[17:00] Carolyn Woodard: So in Microsoft, uh Copilot does call them agents, they're built in Copilot Studio.
[17:06] Carolyn Woodard: Um they there's a governance tool called Agent 365, very uh similar to many other Microsoft uh products there.
[17:17] Carolyn Woodard: Uh
[17:17] Carolyn Woodard: In Google, they're called Gemini Agents and they run on the Gemini Enterprise Agent platform.
[17:24] Carolyn Woodard: In OpenAI, uh they're called GPTs with actions or ChatGPT Workforce Agents.
[17:30] Carolyn Woodard: Uh
[17:31] Carolyn Woodard: If you're using Anthropics Claude, um they are built using Claude Agent, or you can build them in Claude Cowork if you have that as part of your license.
[17:41] Carolyn Woodard: Um
[17:43] Carolyn Woodard: The question of uh, you know, is it practical, is it secure, uh, could you build one without knowing it?
[17:49] Carolyn Woodard: Um when your AI tool prompts you to allow access to a folder or your email or your entire SharePoint, you know, you can say no.
[17:58] Carolyn Woodard: And if you've noticed in the last couple of weeks, even just since there's been these rogue agents running around, uh, and I think actually maybe since the Mythos um rollout uh issues where you know the US government said, oh, that's too powerful, you can't roll it out.
[18:15] Carolyn Woodard: Um, that that version, there have been, it feels for me as a Claude and a Gemini user that it feels like there are more checks on it.
[18:23] Carolyn Woodard: So it will ask you more.
[18:24] Carolyn Woodard: Like, do you want to give me access to this folder?
[18:27] Carolyn Woodard: Do you want me to uh do this thing?
[18:29] Carolyn Woodard: So you there are more points at which you have to allow it to do the thing.
[18:33] Carolyn Woodard: And as I said, you can always say no.
[18:35] Carolyn Woodard: Um, you know, especially if it's something that seems kind of far reaching.
[18:39] Carolyn Woodard: Um,
[18:39] Carolyn Woodard: But anything that you're unsure about, like if it's saying like I want to access everything in SharePoint, like probably you should be thinking about that.
[18:46] Carolyn Woodard: Like, oh, what does that say in our AI policy?
[18:49] Carolyn Woodard: Do I need to check with my IT team?
[18:51] Carolyn Woodard: Um, you know, before you grant it access to everything that you have access to, just make sure that that is within the parameters of your policy, which hopefully you have a policy by now, but if you don't, I would still say, you know, check with someone, run this by someone who, you know, does your security at your nonprofit because it's just good to check.
[19:10] Carolyn Woodard: Make sure that you're using the AI tool in a secure way.
[19:13] Carolyn Woodard: Um,
[19:14] Carolyn Woodard: You can work with your team to develop agents.
[19:19] Carolyn Woodard: Uh, one of the guidelines, guardrails is to make it as narrowly scoped as possible.
[19:25] Carolyn Woodard: So don't give it free range over everything.
[19:28] Carolyn Woodard: You know, um, don't expect it to be uh, I don't know, Jarvis in the Marvel movies.
[19:34] Carolyn Woodard: Um, but you can make it a very specific agent that does a very specific, narrowly scoped thing.
[19:41] Carolyn Woodard: You can use a pre-made agent that already does that one thing, uh, and you'll find those in your tools uh in the AI tool of choice, if it's Claude, Copilot, Gemini, it'll show you some pre-made tools that you can use.
[19:57] Carolyn Woodard: Um, and so those are already, you know, they've been created already, so they've been vetted, they do a thing.
[20:04] Carolyn Woodard: Um, they're not um they're not the cutting edge of agents that are you know able to make a lot of their own decisions.
[20:14] Carolyn Woodard: So before you hit create, um, you should be sure what permissions you're actually granting that agent, whether you can audit the permissions later, if you can revoke them later.
[20:24] Carolyn Woodard: So again,
[20:25] Carolyn Woodard: If you're not sure about these things, um, check in with your IT team.
[20:29] Carolyn Woodard: You can even ask your AI tool, like, I'm getting this prompt.
[20:33] Carolyn Woodard: What does that mean?
[20:34] Carolyn Woodard: What do you what is it going to allow access to so you can understand better?
[20:39] Carolyn Woodard: But I would definitely run it by a person after you do your research, maybe run it by the person at your organization who runs IT or runs security and just make sure that you're using it correctly if you're not sure.
[20:52] Carolyn Woodard: I did talk about the Open AI and Anthropic incidents with their rogue agents last week.
[20:58] Carolyn Woodard: There, I think there was another one that happened since then or has been revealed that happened earlier, but they were just had to disclose it.
[21:06] Carolyn Woodard: Um, and all of those agents that quote unquote got out were really advanced models.
[21:13] Carolyn Woodard: And uh for
[21:15] Carolyn Woodard: I think in both cases it was human error that allowed them to escape their sandbox.
[21:20] Carolyn Woodard: And what they were doing was they were testing the cybersecurity.
[21:24] Carolyn Woodard: Was that agent able to get out?
[21:27] Carolyn Woodard: And um, yeah, there were some things that were left open, and so the agents were able to get out onto the open internet and go to these other sites and try to uh find some information for them, uh, which they weren't supposed to do.
[21:39] Carolyn Woodard: And in both of those cases, it was the human error, which you know, there's humans involved all along here.
[21:46] Carolyn Woodard: So it's not the last time that we're gonna see some kind of human error that made that allowed an AI tool to do something that the humans didn't expect or just hadn't locked down as much as they thought they had or had wanted to.
[22:00] Carolyn Woodard: Um, so
[22:04] Carolyn Woodard: I think it's not um it's not something that people have to think about at nonprofits if you're just getting your feet wet.
[22:13] Carolyn Woodard: I mean, it is good to be cautious.
[22:15] Carolyn Woodard: If your AI says, hey, can I see all of your SharePoint?
[22:18] Carolyn Woodard: you know, just if that were your assistant, would you let them?
[22:21] Carolyn Woodard: Like if that were a person, would you say, sure, look at my files?
[22:25] Carolyn Woodard: Can you check my email while you're in there?
[22:27] Carolyn Woodard: Uh so just think about what you would want them to be able to see and the files that you would want them to be able to access and use your human judgment.
[22:37] Carolyn Woodard: And if you're not sure about something, again, that's the time to check your policy, check with other people at your organization.
[22:43] Carolyn Woodard: Um,
[22:44] Carolyn Woodard: I think one of the risks that a lot of nonprofits are currently, or the IT tech teams that nonprofits are currently, you know, thinking a lot about is how AI allows individuals to go do things individually.
[22:59] Carolyn Woodard: And it's not always the case that the IT team can, and or nor do they want to shut down your access of your AI tools to be, I mean, that would make them not very useful to you.
[23:11] Carolyn Woodard: So it's a balance and it's always good.
[23:14] Carolyn Woodard: We've been saying this for a while, but I think the transparency and openness and there are no dumb questions.
[23:21] Carolyn Woodard: Like, try to keep creating an environment at your nonprofit where people can ask if they're unsure because this is all brand new, it's moving really fast.
[23:31] Carolyn Woodard: The security may be different next week than it is today, of what I'm telling you.
[23:35] Carolyn Woodard: So keeping that dialogue out in the open and making it a very safe place for people to ask their questions and knowing who to ask at your organization is very important.
[23:46] Carolyn Woodard: In fact, I just recently was seeing today or yesterday this um advocacy for an AI owner at your nonprofit or foundation or association, that there would, this is a job title, that someone is like the AI kind of guru for your organization that's keeping an eye on the policy, that's um handling the training and AI literacy, and just, you know, it's a big task.
[24:12] Carolyn Woodard: Uh,
[24:12] Carolyn Woodard: It looks deceptively simple because you just ask it a question.
[24:15] Carolyn Woodard: But if you think about all the change management, all the things that could go wrong, uh, having someone in charge of that at that executive level is kind of a good idea.
[24:23] Carolyn Woodard: And so you're starting to see these job titles more and more people are hiring for them.
[24:28] Carolyn Woodard: So, another question about the agents that I just thought was so fascinating.
[24:33] Carolyn Woodard: I don't have an answer to it, but I saw an article about who is liable, right?
[24:38] Carolyn Woodard: In those rogue agents questions, or if, you know, for some other AI agent issues too, even if it's doing what it's supposed to be doing at your organization and that thing is against a policy like GDPR or HIPAA or something like that, so that you could be held accountable for it, uh, it's doubtful that they'll hold those agents accountable for it.
[25:00] Carolyn Woodard: It's still human in the loop.
[25:02] Carolyn Woodard: Like you're the agent isn't gonna go to jail.
[25:05] Carolyn Woodard: But it's a big gray area, like they don't actually even really know uh the laws have not completely been written yet of how does it apply if it's an agent that's taking that action.
[25:16] Carolyn Woodard: So something to keep an eye on if you work at a nonprofit that works in legal issues.
[25:20] Carolyn Woodard: I'm sure you already are thinking about this.
[25:22] Carolyn Woodard: So um keeping an eye on it.
[25:25] Carolyn Woodard: And then um I wanted to close up with some more thoughts on how to use AI when you are using AI at your job, when to let it write for you, when to let it think for you.
[25:42] Carolyn Woodard: Like, is there a is there a possibility that you would stop thinking because you were using AI, which is coming up a lot in education.
[25:51] Carolyn Woodard: So if you're a nonprofit that works in education, you've seen this.
[25:54] Carolyn Woodard: People are talking about it all the time.
[25:56] Carolyn Woodard: But I wanted to kind of land on this idea that so the argument is that you should write your own draft.
[26:05] Carolyn Woodard: Um, even if it's messy, even if it's, you know, not at all what the polished finished project is gonna be like, the act of writing the draft is important to thinking.
[26:15] Carolyn Woodard: And it's when unique framing can happen, it's when you bring what you know to that draft, and that you shouldn't hand that step over to AI because everything then downstream is gonna be shaped by what the AI proposed, and you're very likely to go along with it.
[26:33] Carolyn Woodard: Uh,
[26:33] Carolyn Woodard: AIs have bias built into them.
[26:36] Carolyn Woodard: They make decisions that aren't transparent about what word to use, what uh metaphor to use, how to write what they're gonna write.
[26:44] Carolyn Woodard: And so there's this, you know, proposal out there, or I guess judgment or opinion that um you should, you should do the first draft and then you use AI maybe to refine it.
[26:57] Carolyn Woodard: I want to push back on that a little bit.
[26:58] Carolyn Woodard: I think in my career and in my experience, there are so many different learning styles.
[27:04] Carolyn Woodard: There are so many different jobs.
[27:06] Carolyn Woodard: Even within your job, you do at a nonprofit, you probably do a hundred different things a day.
[27:12] Carolyn Woodard: And each of those jobs has different parameters, different uh amount, layers of busy work, different amounts of creativity, um, different amounts that you need to think about them or want to have to think about them.
[27:25] Carolyn Woodard: So I don't think it's fair to just say a blanket statement, everyone should write their first draft themselves.
[27:32] Carolyn Woodard: I think it really is going to be individual and like I said, come down to the actual task and the way your brain works.
[27:41] Carolyn Woodard: So if you want, or if you're the type of person that loves a blank page and you really use that blank page for thinking, whether you're making PowerPoint slides, writing your first draft, um, you know, looking at an Excel sheet, you want to fill in those columns yourselves, that's the way you operate.
[28:00] Carolyn Woodard: That's fine.
[28:01] Carolyn Woodard: Like no one should be able to um, I don't know, give you parameters around how that would be best for you to do.
[28:09] Carolyn Woodard: But equally, like if that's not your forte, you hate a blank page.
[28:14] Carolyn Woodard: You will have, you know, like your intern write you a draft that then you can edit.
[28:20] Carolyn Woodard: Why not let the AI write a draft that then you can edit?
[28:23] Carolyn Woodard: Because you have human discernment and human decision making, and you can also put into your prompts as your AI tool gets to know you and does the writing with you, you know, it's gonna learn that "me" layer.
[28:35] Carolyn Woodard: And it's going, you can question it about biases.
[28:39] Carolyn Woodard: You can um bring your experience to it when you're editing it together.
[28:44] Carolyn Woodard: Um,
[28:45] Carolyn Woodard: If you're the type of person that wants uh someone at work to bounce ideas off of, to brainstorm with, you know, AI can be a good tool for doing that.
[28:54] Carolyn Woodard: Um, and
[28:55] Carolyn Woodard: I don't think we should say to people like you should only use AI in these couple of ways.
[29:00] Carolyn Woodard: So I think it's really going to come down to individual preferences and you know, where it saves you time, where it saves you gives you some of your time back, you know, reclaiming your time.
[29:14] Carolyn Woodard: If you have to write, um basically there are only so many ways you can write a newsletter, right?
[29:20] Carolyn Woodard: So if you're constantly trying to figure out a new way to connect those two items together, you know, sometimes it's AIs can be pretty good at that, at finding just a slightly different way to say it.
[29:33] Carolyn Woodard: So I don't think that is a drawback.
[29:36] Carolyn Woodard: Like it
[29:37] Carolyn Woodard: Really, we are all individuals, we're all creating this "me" layer with the tools that we use.
[29:42] Carolyn Woodard: And um, I guess just hold the line.
[29:44] Carolyn Woodard: Like if you're using AI for something that's really helpful to you that helps you think in the way that you use it, like go for it.
[29:51] Carolyn Woodard: You should you should be um free to do that.
[29:55] Carolyn Woodard: But I think you know, we're a nonprofit so we're always thinking about is my brain saying this?
[30:00] Carolyn Woodard: Is it just habit?
[30:01] Carolyn Woodard: Is it really uh a deeper issue?
[30:04] Carolyn Woodard: Is it something systemic and keeping an eye out for those biases and assumptions that the AI is making?
[30:11] Carolyn Woodard: You know, I'm sure all of you are doing that.
[30:13] Carolyn Woodard: So um keep keep thinking about that, and I will share this uh link to a couple different when not to use AI tutorials talking about those types of um, I don't know, learning styles and work styles.
[30:27] Carolyn Woodard: So uh that's all I have for you.
[30:29] Carolyn Woodard: I have some resources, as I said, I'll put in the show notes.
[30:32] Carolyn Woodard: Um that's it for today.
[30:33] Carolyn Woodard: I will be back on Friday with a new podcast about um how to organize yourself to uh move from how to know that you should move and then how to migrate from Google to Microsoft, if that is something that makes sense for your nonprofit.
[30:51] Carolyn Woodard: Um, of course, you can also go the other way.
[30:53] Carolyn Woodard: It's like, oh, Microsoft isn't working for us anymore.
[30:55] Carolyn Woodard: We want to use Google.
[30:56] Carolyn Woodard: So it's not a value um judgment, it's just some of the things to think about and the change management, because that is a really big piece of it.
[31:03] Carolyn Woodard: So I'll see you on Friday with that, and uh, we'll be back here with some more nonprofit AI on Tuesday.
[31:09] Carolyn Woodard: Until then, take care.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.