309: Responsible AI Adoption for Grassroots and Justice-Focused Nonprofits with DeAnna Hoskins - What the Fundraising

Loading summary

Transcript22 lines

[00:00]
Deanna Hoskins
AI is not going to take people's job if you actually implement the governance structure that is necessary, that a human has to be involved in every step of the way. And I tell people this, especially when I was training my team. I first had to remove the fear and explain what it was. It's a tool. It's not a replacement. It's a tool. So that when Deanna has these ideas in her head, Deanna's not starting from a blank paper.
[00:30]
Mat
Hey, my name is Mat and I'm obsessed with helping leaders in the nonprofit space raise money and run their organizations differently. What the Fundraising is a space for real and raw conversations to both challenge and inspire you. Not too long ago I was in your shoes, uncomfortable with fundraising and unsure of my place in this sector. It wasn't until I started to listen to other experts outside of the fundraising space that I was able to shift my mindset and ultimately shift the way I show up as a leader. This podcast is my my way of blending professional and personal development so we as a collective inside the nonprofit sector can feel good about the work we are doing. Join me every week as I interview some of the brightest minds in the personal and professional development space to help you fundamentally change the way you lead and fundraise. I hope you enjoy this episode, so let's dive in.
[01:25]
Mallory Erickson
Welcome everyone. I am so excited to be here today with Deanna Hoskins. Deanna, welcome to what the fundraising.
[01:31]
Deanna Hoskins
Thank you, thank you for having me.
[01:34]
Mallory Erickson
Let's start with you just telling everyone a little bit about you and your work and what brings you to our conversation today.
[01:41]
Deanna Hoskins
Thank you Mallory. Again, thank you for having me. Deanna Hoskins, President CEO of Just Leadership USA Just Leadership is one of the national organizations that's founded by and operated by formerly incarcerated individuals. Our goal is to empower, educate and elevate their voices to drive policy change that impacts them and their communities and that continue to oppress and marginalize us. Prior to joining Just Leadership, I was Senior Policy Advisor up under the Obama administration for the Department of Justice. I oversaw all of the federal government's investment in corrections and reentry. And just recently I finished up a year long fellowship as a White house fellow, the first formerly incarcerated in its 57 year history. And I was actually assigned to work with the Secretary of the Army. The White House Fellowship actually takes individuals out of public sector and put them into the highest levels of government to understand how the inner workings of government and decisions are made. So I came to this conversation around AI in that White House fellow following the army secretary being in top secret meetings. And it was one particular meeting where the CEO of OpenAI, Anthropic, Facebook, TikTok were all meeting with the Secretary of the army. And we would debrief after those conversations. And I was like, what is going on? And he made a comment, he said, deanna, the military has been fighting a war on the ground for years. The war has moved to the air and we're not prepared with, thus why you see all of the media saying military has commissioned 36,000 drones and all of this rapid peace. So I started researching and what I discovered was that there goes the military, there goes the world. Right? The way we do our health care system came from the military. The way we preserve food in the grocery stores came from the military. So I started taking advantage of learning what AI was about in terms of where the world was going inside the Pentagon. And I realized at that moment, if the military is moving this way and thus the world is moving this way, where is that going to leave nonprofits that are in the community doing the work directly that's impacting people? If we weren't having these conversations around capacity building, how do you fund it? How do you train how to even understand it? And when I look at the people who are incarcerated that we want to empower, we're talking about transforming and preparing. If they already don't have access to the computer and now we're moving into this AI world, how much larger is that divide and separation of reality going to be? So I just started digging my boots in, started understanding AI, started taking trainings, became a professional SCRUM master to understand sprints and how things are developed, to coding at Columbia Justice Lab that was focused on directly impacted people. Building tools, not so much because I want to build a tool. I'm just a person who needs to understand the inner workings so that when I sit down with tech people who are experts, I can understand their language and I can talk about how it's going to transform to our communities.
[05:08]
Mallory Erickson
Okay. One of the things I really appreciate about what you're saying is I feel like in our sector so often we define good or bad or what's working or what's not working based on short term implications. Right. So we're like, okay, if we do this, then it gives everybody access to this right now. Or if we build this thing, it's going to help increase our fundraising results this year. And I feel like something I'm constantly trying to sort of press on the sector around is we have to rethink what is quote unquote working mean. And if you can raise more this year because of a tool or because of a particular strategy that's actually going to decrease your fundraising capabilities in three years because you actually used guilt and manipulation and all these things that we know work for one year and then have massive retention issues like, does it, did it work? And I think with AI, this brings sort of additional considerations that, that are so important because of the speed of AI's ability to quote, unquote, solve a pain point. Right. But is it helping us do more of the bad stuff faster? Or is it really addressing root cause issues with a lens towards long term impact and implications, downstream impacts, which I feel like in a world of so much uncertainty right now can feel so overwhelming to the nonprofit leader. And, and it's also leading to a lot of resistance around innovation because of fear that there's a harm that they might not be able to predict or anticipate that they could be contributing to. So that was a very long way of asking, like, how do you balance those truths? Right. There's certain types of risks that we could potentially mitigate or certain types of sort of long term impacts. But I also feel like we're in this world with so much uncertainty around the implications of this technology that I certainly, as conscientious as I try to be in my own work, I do fear sometimes that there's an unintended consequence that I could be missing. And how do you sort of like, you know, balance that thinking?
[07:27]
Deanna Hoskins
So, no, thank you for that. And I think your question has several parts to it because I think we all went straight to a tool. We went straight to let's implement. Let's build a tool that does this better without the foundation of education, training, alleviation of fear, right? Because everyone, when you say AI, I always say this. If you're over 40 and you say AI in a room, people automatically go to ChatGPT and I'm like, okay, it's more than that, right? It's around the innovation, the utilization of the tools we use every day that can solve pain points up. Are you even maximizing how you utilize your Google Calendar to communicate with your team? Right. Do you have a dashboard within your nonprofit that you don't have to host all these meetings? You can actually go say, what is my policy team doing? Where's my policy on tracking this? Is my communications team prepared to send out the given Tuesday notices? What are the assets for that? It's in a dashboard that's Part of innovation, of efficiency. You've now reduced the number of meetings you're having to call your team together. I need you to give me an update on this. I need to give that. How do you start off small for efficiency so that you have the brain space to think more scalable and more innovative. Right. But you also brought up something Mallory, I think people are missing. You can implement all the tools today, but if you are not strategically thinking forward, that tool is going to be outdated in six months. Right. So in order to keep up, you have to keep elevating, but you can't elevate without the proper education. Then there's another part of harm we're not even talking. I look at a lot of philanthropies investment, right? There are some funders out here who are dumping 800 million in AI adoption. One where's the training and capacity? Because you're thinking there's a level playing field, everybody should adopt smaller grassroots nonprofits are already behind the eight ball trying to do the work. You have EDs who are doing work and not acting can't be an ED in the way that a ED of a goodwill or ED of a YWCA that has a national presence, a national footprint and a national framework. Right. So when you do that one, you're not educating to remove the fear and the resistance. Because the narrative out there, AI is going to take people's job. AI is not going to take people's job if you actually implement the governance structure that is necessary that a human has to be involved in every step of the way. And I tell people this, especially when I was training my team, I first had to remove the fear and explain what it was. It's a tool. It's not a replacement. It's a tool. So that when Deanna has these ideas in her head, Deanna's not starting from a blank paper. Chat GPT is your thought part. You can dump an idea in the chat and help chat fluff it out. Chat is not the implementer. You then should take what you learned in chat over to Claude to say, I want to actually strip this down in a way that is implementable in a non profit with limited staff. So now you've trained your team on how to prompts to get a response that they go in and put in the human tool of what is giving you a path. Now is this path possible in my organization, what is the low hanging fruit that I can show my team? Movement, productivity. But then if you're a service provider now, you need the harms because you need policy and governance and guardrails on how your team use it and then what they can put in it, and then building a sandbox that is contained within your organization only. But you have to have policies. And as the leader, you got to be willing to take risks, but you also got to be willing to be held accountable. That's why your governance is important when you have personal identifying information. And so what I'm seeing, Mallory, is philanthropy is talking about these tools. They are not talking about capacity and training. And I nowhere see anybody talking about governance. Who's going to own. It's almost like HIPAA and electronic health records. You have to have a HIPAA law within AI, within your nonprofit in some way that protects not only personal identifying information, then there's something called SOC 2, your financials. If you have a third party, which I am a smaller nonprofit, my financials are handled through a third party. I have a vendor policy of what my vendor has to have in place so that all my information is not dumped into an AI automated tool. So that's part of what we're missing. That's the foundation before I can ever start building.
[12:30]
Mallory Erickson
Yeah, I so appreciate that. And it's interesting now being on sort of the technical side of, you know, some of these conversations with organizations and sort of like, navigating, you know, this space. And I think something I talk about a lot from our fundraising AI work. And just sort of my orientation to the work I do with the sector in general is like, the point is not to get some AI, right? Like, we will have people come to practically and be like, I need to get some AI. Can I, like, get your AI? And I'm like, what?
[12:58]
Deanna Hoskins
What?
[12:58]
Mallory Erickson
What do you mean you need to get some AI? No, I'm like, the point is not to have AI, right? The point is to solve some of the challenges that you're experiencing. Experiencing. And so really starting from that place of understanding your pain points and your problems, and then starting to investigate the different solutions to those, some of which may include AI in another tool, some of which might be an AI native tool. Right. And so it's interesting because it's similar to actually how I built practivated. And this goes back to some of what you were talking about before. That really made me think about kind of like my early journey, which was not, you know, I grew up in the Bay Area. I've spent my entire life near Silicon Valley, and I have avoided technology my entire time, much to the dismay of my parents, you know, especially my dad. Who would be like, don't you want to go work for, you know, so and so, like, your siblings? And I would be like, no, like, this is my calling. This is where I belong. And my husband jokes that I still never tell people that I'm a tech founder. But I had been trying to solve this problem, problem around fundraiser readiness and what it takes in our brains and bodies to be more confident and emboldened in our relationships with donors. I had been trying to solve that problem with workshops and speaking and podcasting for years. And it wasn't until generative AI. I saw the power of generative AI and saw the power of personalization at scale that I thought, what about instead of personalization at scale for the donors, personalization at scale for the fundraisers, could this build the muscle memory and confidence and skills that we are really needing in this sector? And that's when I started to be able to get curious about what might be possible if I built this thing and the other thing. And I'm very curious for your thoughts here, and you're absolutely welcome to disagree with me. One of the things, you know, because my background had been in the fundraising AI space, all through that lens of response and beneficial AI, I realized, you know, sort of surprisingly, that I had to start building the actual tool to a certain extent. Not selling it, not sharing it, but starting to build something in order to actually illuminate for myself all of the ways in which it could be problematic. Right. It wasn't until I had my first donor avatar conversation and got feedback that I could see. See the bias, Right? And I could see that I'm neurodivergent. Right? So I don't have donor conversations the exact same way as somebody who's neurotypical. And the feedback that I need is different. And so. But it wasn't until I started to build that I started to see all the equity potential, equity issues, potential bias issues, ethics issues. They started to illuminate themselves. And then that's when I could take those into consideration. Consideration as I continued to build. And so I'm just curious because I really hear what you're saying around the governance piece and all of that needing to be in place before you can really build. I'm curious for folks who are feeling so overwhelmed with, like, how does this even work? And we can't even imagine what might be irresponsible, like, any recommendations you have around how they take one foot forward in one area while they're taking another foot forward in another.
[16:18]
Deanna Hoskins
So thank you for that because I think you really touched on something really important nobody pays attention to. You're not going to understand bias if you totally rely on the tool 100%. Right? But you also are not going to recognize bias till you learn how to use the tool for what you want. So that goes back to my original. We have to train and build capacity, right? So that people could see the actual biases in these tools that are going to come up. And then you understand how to even prompt the tool to say, remove the various biases in different things of it goes to the training of how you even prompted to get what you want. Right? So I think that's really the part because it only can give you information that has been dumped in it. And I want to give you an example. I was at a tech conference conference, my first ever tech conference, and I'm new to this world. Well, Deanna, why are you here? Where do you really fit in? But I needed to understand these experts who were building avatars, building chat bots, doing all these things. And I remember the one girl brought up to me and she said, if you don't understand, if you are already biased yourself, you're not going to understand the biases that is giving you because that appears normal to you. But she also said, we don't train people. So when you actually implement your tool or you're utilizing your tool, you have to understand what you don't want it to have. So she had created something called Chat GPT, Black Chat gbt. And the reason she created Black Chat GPT, because again, chatgpt only could give you the information that's been fed that is learned from the various people across the country. Right? She realized and she put a question in chat to give, I think she put a question in there, something like, Give me the 10 most popular black people. And the people that it brought up. It was really amazing how if we don't program information correctly, it can almost eliminate a population or a culture of people because it's only getting who's putting in information, right? So if she had Never created black ChatGPT on top of ChatGPT, ChatGPT would not even highlight after President Obama, Harriet Tubman, Frederick Douglass and those different people. So she started to see where history was being eliminated if we're not intentional in keeping a human element in it. So what I tell people, and this is going to sound crazy to people, you should be having a conversation with Chat. And when Chad is wrong, you should be of course correcting chat to say, well, that's not correct. Because in 1918, what slavery, you know what I'm saying these different things. Jim Crow laws were in flat back. So in fact, slavery was not abolished at the end of slavery because we have Jim Crow, we have black laws, and then it will say, you're correct, but this is how we have to ensure our voices and the things that we know are biased in it. Because instead of always trying to extract, you got to remember you're putting information in too. So even if it kicks you something biased, if it kicks you misinformation, and even if it's around your stuff because it's learning you, if you have a page and I tell everybody, please have a paid subscription and set your guardrails right, Even your personal chat that knows you, you're paying for has guardrails. It will bring up information misinformation and you have to say, no, I have a master's degree in criminal justice and remember I did. Oh, you're right. So you have to almost engage with it. And I know it's weird, it's your friend at this time. And we don't do that. Right. We take what is giving us and extract it as correct. Instead of saying, hey, that appears a little biased because African Americans may not benefit from this in the same way more economically advantaged communities would do to this. And you start giving it facts. Your educatement too. Again, it's almost when I do criminal justice reform, and I have to remember I'm having conversations with policymakers, other people who only can bring the point of view they've had experience from. Yeah, same thing with chat. People are inputting stuff. If people who haven't had your experience have not been overwhelmingly inputting information that is getting sourced to the bottom. So we have a responsibility as well to challenge it and put in correct information.
[21:24]
Mallory Erickson
Yes. Hey, I want to tell you about a podcast I genuinely love, the small nonprofit podcast. It's hosted by the incredible Maria Rio, and it speaks directly to the realities so many of us are living inside every day. Maria, tell us about your show.
[21:43]
Maria Rio
Hi, I'm your host, Maria Rio. I moved to Canada as a refugee, use charitable services myself, and have been fundraising for over a decade. I love exploring ways our sector can grow, change and make the world more equitable. The small nonprofit podcast is where we do just that. I believe small nonprofits are where the most transformative work happens. But we've been set up to failed by systems that expect us to do more with less. Every week I sit down with leaders, fundraisers and change makers who are doing things differently. They're building sustainable funding without compromising their values, creating power instead of dependence and pushing back on the broken systems we've all inherited. You'll walk away with actionable strategies you can use, not just inspiration. So if you're ready to stop scrambling and start building something that lasts, you're in the right place. Let's get into it.
[22:35]
Mallory Erickson
If you're leading or fundraising at a small nonprofit and you're craving work that feels aligned, sustainable and human, go listen to the small nonprofit podcast.
[22:45]
Mat
Search the small nonprofit podcast on Spotify
[22:48]
Mallory Erickson
or Apple Podcasts to listen. Okay, I think that is so important and I know, I know we're just wrapping up, but I just wanted to like click on a few things that you said because so my family speaks Spanish also and it's been really interesting when we speak to these tools in Spanish versus English and the different. We've tested it a little bit where and like image creation will do the same prompt in English versus in Spanish and wildly different responses in a ton of different ways. Right. And so I think, you know, what you're saying is really important and it's also making me wonder and we're not going to solve this today, but you know, there's a lot of criticism in this space around what they call wrapper products and practivated isn't this, but essentially that layer on top of an LLM layer, like what you were talking about with Black Chat, G B T. But what you're saying to me actually is that in a lot of ways those wrapper products are, can be really important because they're, they're identifying the heuristics and the ethics and the bias mitigation to ensure that the outputs are aligned with your values, your experience, your. And so I just really love that kind of reframing because I think that's really important and really important to understand. I think with the right type of heuristics and guardrails and controls, you actually can have a tool that is much less biased than a human. Like that's something we work really hard on. And practivated is like as a coach, I am going to have my bias if I listen to somebody give a, you know, a ten minute donor conversation. And I have my ways right. That I can't control because they are my lived experience. But actually through how we control the platform, we can ensure that people's feedback is less biased than any human coach would be. So you just, you have me thinking about so many things. Okay, tell, tell everybody where they can go to connect with you to learn more about your AI work and then of course about the organization.
[24:50]
Deanna Hoskins
Thank you. You definitely can reach out to us at our website@jlusa.org and there's a contact link there. You can email us or you can email me at deanna d e a n n a l usa.org and I'm loving having these conversations because we are literally moving to a tool. And really I want people to think about whatever population you're serving, whatever your passion is, what does that look like for that community and that culture. And that's where your voice has to speak. As these tools are being developed, as you're interacting, it only can adopt the culture and give the response if the culture is inputting information as well. So really, you know, when we circle back to the purpose of the call and really focusing on philanthropy, philanthropy has to focus. The tools are important. But if I don't have the capacity, if I don't have the education and understanding, the tools are just going to be there. And again, after the trendy thing of AI becomes normal, funders aren't going to fund training and capacity. So we have to get it now because they are widening the net of disparity, thinking that communities and organizations are already there and they're not. So we have to start at ground zero and then we have to use. You said something when I came into this conversation, the quick and easy solve this problem in a year. This is again, you can build a tool today that's going to be outdated six months. So how do you invest multi year in not only the capacity, the training, the governance, but then the scalability and building Because I always say Apple iPhone, never changed their name, they're just on version 18. Yeah, it's the same phone but they got to keep updating it because guess what, the old ones, you know, my kids tell me all the time your phone's outdated. I'm like, it still works, but I don't have superior picture taking like they do because the tool is actually outdated.
[26:59]
Mallory Erickson
Yeah. Well, thank you so much for joining me and for all the work that you're doing to support the sector.
[27:06]
Deanna Hoskins
Thank you, thank you for having and thank you for having this important conversation so we can elevate it.
[27:18]
Mat
I hope today's episode inspired or challenged you to think differently. For additional takeaways, tips, show notes and more about our amazing guest and sponsors, head on over to Mallorykson.com podcast and if you didn't know, hosting this podcast isn't the only thing I do. Every day I coach, guide and help fundraisers and leaders just like you. Inside of my program, the Power Partners Formula Collective. Inside the program, I share my methods, tools and experiences that have helped me fundraise millions of dollars and feel good about myself in the process. To learn more about how I can help you, visit MalloryErickson.com PowerPartners Last but not least, if you enjoyed this episode, I'd love to encourage you to share it with a friend you know would benefit or leave a review. I'm so grateful for all of you and the good, hard work you're doing to make our world a better place. I can't wait to see you in the next episode.
[28:18]
Practivated Announcer
Fundraising is hard. Every donor conversation carries pressure and most fundraisers are expected to just figure it out through trial and error. That's why we built Practivated, the first ever AI powered donor conversation simulator made just for fundraisers. It's a safe, judgment free space to practice your pitch, refine your storytelling, and build the confidence that drives real results for your mission. Because conversations move missions forward, with Practivated, you and your team can practice anytime. Get real feedback instantly and walk into donor meetings ready, not rehearsed, but prepared. See how practice changes everything? Try practivated today at www.practivated.com and start building confidence one conversation at a time.

Summary6 min read

Podcast Summary: What the Fundraising

Episode 309: Responsible AI Adoption for Grassroots and Justice-Focused Nonprofits with DeAnna Hoskins

Host: Mallory Erickson
Guest: DeAnna Hoskins, President & CEO of Just Leadership USA
Date: July 21, 2026
Duration Summarized: [00:00–27:05]

Episode Overview

In this impactful episode, Mallory Erickson sits down with DeAnna Hoskins, President & CEO of Just Leadership USA, to unpack the challenges and opportunities of AI adoption for grassroots and justice-focused nonprofit organizations. DeAnna draws on her unique background—from federal policy advisor to White House Fellow and grassroots leader—to illuminate how nonprofits can responsibly navigate the fast-evolving world of AI. The conversation addresses not just the efficiency gains and risks of new technology, but goes deeper into issues of equity, capacity, training, governance, and the critical need for intentionality to ensure technology serves—not further marginalizes—vulnerable communities.

Key Discussion Points & Insights

1. The Military, Technology, and the Nonprofit Sector [01:40–05:07]

DeAnna’s Journey into AI:
- Shared her entry point into AI from White House fellowship working with senior military and tech leaders.
- Noted that many societal innovations (healthcare systems, food preservation) have their origins in military practice, and observed the rapid AI adoption in military as a bellwether for broader societal change.
- DeAnna:
  
  "If the military is moving this way and thus the world is moving this way, where is that going to leave nonprofits...?"
Implications for Nonprofits:
- She became concerned about the widening digital divide, especially for marginalized populations (e.g., formerly incarcerated individuals).
- Emphasized the urgency for nonprofits to build capacity, funding, and training for AI, warning that those left behind in this transformation risk further marginalization.

2. Rethinking ‘What’s Working’ with AI & Technology [05:07–07:27]

Mallory Challenges Short-Term Thinking:
- Raises questions about what 'working' really means—does a quick AI win actually serve longer-term strategy and ethics?
- Stresses the need to avoid using technology solely for short-term outcomes, especially if those methods undermine trust or harm relationships later.

3. Building the Right Foundation: Education & Governance Before Tools [07:27–12:30]

Start with Training and Capacity:
- DeAnna highlights a common misstep—jumping straight to tool adoption without the foundational work of training, education, and capacity-building.
- Governance as Non-Negotiable:
  - Advocates for policy frameworks akin to HIPAA (for medical records) and SOC 2 (for financial information).
  - DeAnna:
    
    "AI is not going to take people's job if you actually implement the governance structure that is necessary that a human has to be involved in every step of the way." [00:00 & 07:46] "You have to have policies. And as the leader, you got to be willing to take risks, but you also got to be willing to be held accountable."
- Risks of Philanthropic Funding:
  - Critiques funders who throw money at AI adoption without funding the requisite training and operational changes necessary for responsible use.

4. The “Right” Approach to AI – Solving Problems, Not Chasing Trends [12:30–16:18]

Mallory on Purposeful AI:
- Warns against “getting some AI” simply for the sake of it—urges organizations to start with the challenge they’re trying to solve, then evaluate if/where AI fits as part of the solution.
- Shares her experience building personalized fundraising training tools and how active development surfaced unexpected bias and equity considerations.

5. Tackling Bias and Creating Equitable AI [16:18–21:24]

Learning through Doing—and Correction:
- DeAnna stresses that bias in AI is only revealed if users learn and engage with the tools.
- DeAnna:
  
  "If you don't understand, if you are already biased yourself, you're not going to understand the biases that is giving you because that appears normal to you."
- Shares example from a tech conference: Black ChatGPT was developed because standard models defaulted to incomplete or biased answers about Black historical figures, underscoring the need for intentional data input.
- Ongoing, Interactive Correction:
  - Coaches organizations to regularly challenge, correct, and “educate” their AI tools, reminding listeners that "you’re putting information in too"—active human involvement is essential.

6. Embracing Culturally Responsive & Biased-Aware AI Tools [22:44–24:49]

Mallory on Language & “Wrapper” Products:
- Notes divergent responses from AI tools when prompting in different languages (e.g. English vs. Spanish), highlighting the importance of culturally responsive technology.
- Questions criticism of AI ‘wrapper’ products designed for ethics and bias mitigation, suggesting these tools can sometimes outperform human coaches in reducing bias.
Value of Community Input:
- Both emphasize the importance of inputting a community’s values, culture, and lived experience, so that AI tools reflect and serve those communities authentically.

7. Philanthropy and Sustainable AI Investment [24:49–26:59]

DeAnna’s Closing Call:
- Leaders must center their community’s needs in all AI adoption decisions.
- Philanthropy should prioritize longer-term investments in capacity-building, training, governance, and iterative tool development—not just initial tool acquisition.
- DeAnna:
  
  "After the trendy thing of AI becomes normal, funders aren’t going to fund training and capacity. So we have to get it now because they are widening the net of disparity, thinking that communities and organizations are already there and they're not." [25:55] "You can build a tool today that's going to be outdated in six months. So how do you invest multi-year in not only the capacity, the training, the governance, but then the scalability and building?" [26:25]

Notable Quotes & Memorable Moments

On Fear and Misconceptions:

"It's a tool. It's not a replacement. It's a tool. So that when Deanna has these ideas in her head, Deanna's not starting from a blank paper."
- DeAnna Hoskins [00:00 & 07:48]
On Bias:

"You're not going to recognize bias until you learn how to use the tool for what you want ... If you're already biased yourself, you're not going to understand the bias that it’s giving you because that appears normal to you."
- DeAnna Hoskins [17:15–17:51]
On Accountability:

"As the leader, you gotta be willing to take risks, but you also gotta be willing to be held accountable. That’s why your governance is important when you have personal identifying information."
– DeAnna Hoskins [11:45]
On the Role of Philanthropy:

"Philanthropy has to focus. The tools are important. But if I don't have the capacity, if I don't have the education and understanding, the tools are just going to be there."
– DeAnna Hoskins [25:10]
On Rapid Tech Change:

"Apple iPhone, never changed their name. They're just on version 18. It's the same phone but they gotta keep updating it … because the tool is actually outdated."
– DeAnna Hoskins [26:30]

Timestamps for Key Sections

DeAnna’s introduction and military-tech-nonprofit connect: [01:40–05:07]
Rethinking impact and short-termism: [05:07–07:27]
AI implementation mistakes & the case for governance: [07:27–12:30]
Identifying the right problems for AI: [12:30–16:18]
Surfacing and mitigating AI bias: [16:18–21:24]
Language, community, and wrapper products: [22:44–24:49]
Closing thoughts on philanthropy’s role: [24:49–26:59]

Connect with the Guest

DeAnna Hoskins:

Website: jlusa.org
Email: deanna@jlusa.org

Summary Takeaways

AI is a TOOL, not a replacement; education and governance are essential.
Nonprofits must center community, equity, and long-term impact—not just efficiency—when adopting AI.
Bias is inevitable unless people are trained and actively engaged in correcting and shaping tools.
Sustainable funding for capacity-building and training is CRITICAL; don’t let philanthropy just fund a trend.
Community voices must be inputted intentionally so that technology serves every culture equitably.

This episode masterfully blends practical advice with high-level caution, offering nonprofit leaders both a sense of agency and accountability to shape their technology journey wisely and equitably.

Loading summary

Transcript22 lines

[00:00]
Deanna Hoskins
AI is not going to take people's job if you actually implement the governance structure that is necessary, that a human has to be involved in every step of the way. And I tell people this, especially when I was training my team. I first had to remove the fear and explain what it was. It's a tool. It's not a replacement. It's a tool. So that when Deanna has these ideas in her head, Deanna's not starting from a blank paper.
[00:30]
Mat
Hey, my name is Mat and I'm obsessed with helping leaders in the nonprofit space raise money and run their organizations differently. What the Fundraising is a space for real and raw conversations to both challenge and inspire you. Not too long ago I was in your shoes, uncomfortable with fundraising and unsure of my place in this sector. It wasn't until I started to listen to other experts outside of the fundraising space that I was able to shift my mindset and ultimately shift the way I show up as a leader. This podcast is my my way of blending professional and personal development so we as a collective inside the nonprofit sector can feel good about the work we are doing. Join me every week as I interview some of the brightest minds in the personal and professional development space to help you fundamentally change the way you lead and fundraise. I hope you enjoy this episode, so let's dive in.
[01:25]
Mallory Erickson
Welcome everyone. I am so excited to be here today with Deanna Hoskins. Deanna, welcome to what the fundraising.
[01:31]
Deanna Hoskins
Thank you, thank you for having me.
[01:34]
Mallory Erickson
Let's start with you just telling everyone a little bit about you and your work and what brings you to our conversation today.
[01:41]
Deanna Hoskins
Thank you Mallory. Again, thank you for having me. Deanna Hoskins, President CEO of Just Leadership USA Just Leadership is one of the national organizations that's founded by and operated by formerly incarcerated individuals. Our goal is to empower, educate and elevate their voices to drive policy change that impacts them and their communities and that continue to oppress and marginalize us. Prior to joining Just Leadership, I was Senior Policy Advisor up under the Obama administration for the Department of Justice. I oversaw all of the federal government's investment in corrections and reentry. And just recently I finished up a year long fellowship as a White house fellow, the first formerly incarcerated in its 57 year history. And I was actually assigned to work with the Secretary of the Army. The White House Fellowship actually takes individuals out of public sector and put them into the highest levels of government to understand how the inner workings of government and decisions are made. So I came to this conversation around AI in that White House fellow following the army secretary being in top secret meetings. And it was one particular meeting where the CEO of OpenAI, Anthropic, Facebook, TikTok were all meeting with the Secretary of the army. And we would debrief after those conversations. And I was like, what is going on? And he made a comment, he said, deanna, the military has been fighting a war on the ground for years. The war has moved to the air and we're not prepared with, thus why you see all of the media saying military has commissioned 36,000 drones and all of this rapid peace. So I started researching and what I discovered was that there goes the military, there goes the world. Right? The way we do our health care system came from the military. The way we preserve food in the grocery stores came from the military. So I started taking advantage of learning what AI was about in terms of where the world was going inside the Pentagon. And I realized at that moment, if the military is moving this way and thus the world is moving this way, where is that going to leave nonprofits that are in the community doing the work directly that's impacting people? If we weren't having these conversations around capacity building, how do you fund it? How do you train how to even understand it? And when I look at the people who are incarcerated that we want to empower, we're talking about transforming and preparing. If they already don't have access to the computer and now we're moving into this AI world, how much larger is that divide and separation of reality going to be? So I just started digging my boots in, started understanding AI, started taking trainings, became a professional SCRUM master to understand sprints and how things are developed, to coding at Columbia Justice Lab that was focused on directly impacted people. Building tools, not so much because I want to build a tool. I'm just a person who needs to understand the inner workings so that when I sit down with tech people who are experts, I can understand their language and I can talk about how it's going to transform to our communities.
[05:08]
Mallory Erickson
Okay. One of the things I really appreciate about what you're saying is I feel like in our sector so often we define good or bad or what's working or what's not working based on short term implications. Right. So we're like, okay, if we do this, then it gives everybody access to this right now. Or if we build this thing, it's going to help increase our fundraising results this year. And I feel like something I'm constantly trying to sort of press on the sector around is we have to rethink what is quote unquote working mean. And if you can raise more this year because of a tool or because of a particular strategy that's actually going to decrease your fundraising capabilities in three years because you actually used guilt and manipulation and all these things that we know work for one year and then have massive retention issues like, does it, did it work? And I think with AI, this brings sort of additional considerations that, that are so important because of the speed of AI's ability to quote, unquote, solve a pain point. Right. But is it helping us do more of the bad stuff faster? Or is it really addressing root cause issues with a lens towards long term impact and implications, downstream impacts, which I feel like in a world of so much uncertainty right now can feel so overwhelming to the nonprofit leader. And, and it's also leading to a lot of resistance around innovation because of fear that there's a harm that they might not be able to predict or anticipate that they could be contributing to. So that was a very long way of asking, like, how do you balance those truths? Right. There's certain types of risks that we could potentially mitigate or certain types of sort of long term impacts. But I also feel like we're in this world with so much uncertainty around the implications of this technology that I certainly, as conscientious as I try to be in my own work, I do fear sometimes that there's an unintended consequence that I could be missing. And how do you sort of like, you know, balance that thinking?
[07:27]
Deanna Hoskins
So, no, thank you for that. And I think your question has several parts to it because I think we all went straight to a tool. We went straight to let's implement. Let's build a tool that does this better without the foundation of education, training, alleviation of fear, right? Because everyone, when you say AI, I always say this. If you're over 40 and you say AI in a room, people automatically go to ChatGPT and I'm like, okay, it's more than that, right? It's around the innovation, the utilization of the tools we use every day that can solve pain points up. Are you even maximizing how you utilize your Google Calendar to communicate with your team? Right. Do you have a dashboard within your nonprofit that you don't have to host all these meetings? You can actually go say, what is my policy team doing? Where's my policy on tracking this? Is my communications team prepared to send out the given Tuesday notices? What are the assets for that? It's in a dashboard that's Part of innovation, of efficiency. You've now reduced the number of meetings you're having to call your team together. I need you to give me an update on this. I need to give that. How do you start off small for efficiency so that you have the brain space to think more scalable and more innovative. Right. But you also brought up something Mallory, I think people are missing. You can implement all the tools today, but if you are not strategically thinking forward, that tool is going to be outdated in six months. Right. So in order to keep up, you have to keep elevating, but you can't elevate without the proper education. Then there's another part of harm we're not even talking. I look at a lot of philanthropies investment, right? There are some funders out here who are dumping 800 million in AI adoption. One where's the training and capacity? Because you're thinking there's a level playing field, everybody should adopt smaller grassroots nonprofits are already behind the eight ball trying to do the work. You have EDs who are doing work and not acting can't be an ED in the way that a ED of a goodwill or ED of a YWCA that has a national presence, a national footprint and a national framework. Right. So when you do that one, you're not educating to remove the fear and the resistance. Because the narrative out there, AI is going to take people's job. AI is not going to take people's job if you actually implement the governance structure that is necessary that a human has to be involved in every step of the way. And I tell people this, especially when I was training my team, I first had to remove the fear and explain what it was. It's a tool. It's not a replacement. It's a tool. So that when Deanna has these ideas in her head, Deanna's not starting from a blank paper. Chat GPT is your thought part. You can dump an idea in the chat and help chat fluff it out. Chat is not the implementer. You then should take what you learned in chat over to Claude to say, I want to actually strip this down in a way that is implementable in a non profit with limited staff. So now you've trained your team on how to prompts to get a response that they go in and put in the human tool of what is giving you a path. Now is this path possible in my organization, what is the low hanging fruit that I can show my team? Movement, productivity. But then if you're a service provider now, you need the harms because you need policy and governance and guardrails on how your team use it and then what they can put in it, and then building a sandbox that is contained within your organization only. But you have to have policies. And as the leader, you got to be willing to take risks, but you also got to be willing to be held accountable. That's why your governance is important when you have personal identifying information. And so what I'm seeing, Mallory, is philanthropy is talking about these tools. They are not talking about capacity and training. And I nowhere see anybody talking about governance. Who's going to own. It's almost like HIPAA and electronic health records. You have to have a HIPAA law within AI, within your nonprofit in some way that protects not only personal identifying information, then there's something called SOC 2, your financials. If you have a third party, which I am a smaller nonprofit, my financials are handled through a third party. I have a vendor policy of what my vendor has to have in place so that all my information is not dumped into an AI automated tool. So that's part of what we're missing. That's the foundation before I can ever start building.
[12:30]
Mallory Erickson
Yeah, I so appreciate that. And it's interesting now being on sort of the technical side of, you know, some of these conversations with organizations and sort of like, navigating, you know, this space. And I think something I talk about a lot from our fundraising AI work. And just sort of my orientation to the work I do with the sector in general is like, the point is not to get some AI, right? Like, we will have people come to practically and be like, I need to get some AI. Can I, like, get your AI? And I'm like, what?
[12:58]
Deanna Hoskins
What?
[12:58]
Mallory Erickson
What do you mean you need to get some AI? No, I'm like, the point is not to have AI, right? The point is to solve some of the challenges that you're experiencing. Experiencing. And so really starting from that place of understanding your pain points and your problems, and then starting to investigate the different solutions to those, some of which may include AI in another tool, some of which might be an AI native tool. Right. And so it's interesting because it's similar to actually how I built practivated. And this goes back to some of what you were talking about before. That really made me think about kind of like my early journey, which was not, you know, I grew up in the Bay Area. I've spent my entire life near Silicon Valley, and I have avoided technology my entire time, much to the dismay of my parents, you know, especially my dad. Who would be like, don't you want to go work for, you know, so and so, like, your siblings? And I would be like, no, like, this is my calling. This is where I belong. And my husband jokes that I still never tell people that I'm a tech founder. But I had been trying to solve this problem, problem around fundraiser readiness and what it takes in our brains and bodies to be more confident and emboldened in our relationships with donors. I had been trying to solve that problem with workshops and speaking and podcasting for years. And it wasn't until generative AI. I saw the power of generative AI and saw the power of personalization at scale that I thought, what about instead of personalization at scale for the donors, personalization at scale for the fundraisers, could this build the muscle memory and confidence and skills that we are really needing in this sector? And that's when I started to be able to get curious about what might be possible if I built this thing and the other thing. And I'm very curious for your thoughts here, and you're absolutely welcome to disagree with me. One of the things, you know, because my background had been in the fundraising AI space, all through that lens of response and beneficial AI, I realized, you know, sort of surprisingly, that I had to start building the actual tool to a certain extent. Not selling it, not sharing it, but starting to build something in order to actually illuminate for myself all of the ways in which it could be problematic. Right. It wasn't until I had my first donor avatar conversation and got feedback that I could see. See the bias, Right? And I could see that I'm neurodivergent. Right? So I don't have donor conversations the exact same way as somebody who's neurotypical. And the feedback that I need is different. And so. But it wasn't until I started to build that I started to see all the equity potential, equity issues, potential bias issues, ethics issues. They started to illuminate themselves. And then that's when I could take those into consideration. Consideration as I continued to build. And so I'm just curious because I really hear what you're saying around the governance piece and all of that needing to be in place before you can really build. I'm curious for folks who are feeling so overwhelmed with, like, how does this even work? And we can't even imagine what might be irresponsible, like, any recommendations you have around how they take one foot forward in one area while they're taking another foot forward in another.
[16:18]
Deanna Hoskins
So thank you for that because I think you really touched on something really important nobody pays attention to. You're not going to understand bias if you totally rely on the tool 100%. Right? But you also are not going to recognize bias till you learn how to use the tool for what you want. So that goes back to my original. We have to train and build capacity, right? So that people could see the actual biases in these tools that are going to come up. And then you understand how to even prompt the tool to say, remove the various biases in different things of it goes to the training of how you even prompted to get what you want. Right? So I think that's really the part because it only can give you information that has been dumped in it. And I want to give you an example. I was at a tech conference conference, my first ever tech conference, and I'm new to this world. Well, Deanna, why are you here? Where do you really fit in? But I needed to understand these experts who were building avatars, building chat bots, doing all these things. And I remember the one girl brought up to me and she said, if you don't understand, if you are already biased yourself, you're not going to understand the biases that is giving you because that appears normal to you. But she also said, we don't train people. So when you actually implement your tool or you're utilizing your tool, you have to understand what you don't want it to have. So she had created something called Chat GPT, Black Chat gbt. And the reason she created Black Chat GPT, because again, chatgpt only could give you the information that's been fed that is learned from the various people across the country. Right? She realized and she put a question in chat to give, I think she put a question in there, something like, Give me the 10 most popular black people. And the people that it brought up. It was really amazing how if we don't program information correctly, it can almost eliminate a population or a culture of people because it's only getting who's putting in information, right? So if she had Never created black ChatGPT on top of ChatGPT, ChatGPT would not even highlight after President Obama, Harriet Tubman, Frederick Douglass and those different people. So she started to see where history was being eliminated if we're not intentional in keeping a human element in it. So what I tell people, and this is going to sound crazy to people, you should be having a conversation with Chat. And when Chad is wrong, you should be of course correcting chat to say, well, that's not correct. Because in 1918, what slavery, you know what I'm saying these different things. Jim Crow laws were in flat back. So in fact, slavery was not abolished at the end of slavery because we have Jim Crow, we have black laws, and then it will say, you're correct, but this is how we have to ensure our voices and the things that we know are biased in it. Because instead of always trying to extract, you got to remember you're putting information in too. So even if it kicks you something biased, if it kicks you misinformation, and even if it's around your stuff because it's learning you, if you have a page and I tell everybody, please have a paid subscription and set your guardrails right, Even your personal chat that knows you, you're paying for has guardrails. It will bring up information misinformation and you have to say, no, I have a master's degree in criminal justice and remember I did. Oh, you're right. So you have to almost engage with it. And I know it's weird, it's your friend at this time. And we don't do that. Right. We take what is giving us and extract it as correct. Instead of saying, hey, that appears a little biased because African Americans may not benefit from this in the same way more economically advantaged communities would do to this. And you start giving it facts. Your educatement too. Again, it's almost when I do criminal justice reform, and I have to remember I'm having conversations with policymakers, other people who only can bring the point of view they've had experience from. Yeah, same thing with chat. People are inputting stuff. If people who haven't had your experience have not been overwhelmingly inputting information that is getting sourced to the bottom. So we have a responsibility as well to challenge it and put in correct information.
[21:24]
Mallory Erickson
Yes. Hey, I want to tell you about a podcast I genuinely love, the small nonprofit podcast. It's hosted by the incredible Maria Rio, and it speaks directly to the realities so many of us are living inside every day. Maria, tell us about your show.
[21:43]
Maria Rio
Hi, I'm your host, Maria Rio. I moved to Canada as a refugee, use charitable services myself, and have been fundraising for over a decade. I love exploring ways our sector can grow, change and make the world more equitable. The small nonprofit podcast is where we do just that. I believe small nonprofits are where the most transformative work happens. But we've been set up to failed by systems that expect us to do more with less. Every week I sit down with leaders, fundraisers and change makers who are doing things differently. They're building sustainable funding without compromising their values, creating power instead of dependence and pushing back on the broken systems we've all inherited. You'll walk away with actionable strategies you can use, not just inspiration. So if you're ready to stop scrambling and start building something that lasts, you're in the right place. Let's get into it.
[22:35]
Mallory Erickson
If you're leading or fundraising at a small nonprofit and you're craving work that feels aligned, sustainable and human, go listen to the small nonprofit podcast.
[22:45]
Mat
Search the small nonprofit podcast on Spotify
[22:48]
Mallory Erickson
or Apple Podcasts to listen. Okay, I think that is so important and I know, I know we're just wrapping up, but I just wanted to like click on a few things that you said because so my family speaks Spanish also and it's been really interesting when we speak to these tools in Spanish versus English and the different. We've tested it a little bit where and like image creation will do the same prompt in English versus in Spanish and wildly different responses in a ton of different ways. Right. And so I think, you know, what you're saying is really important and it's also making me wonder and we're not going to solve this today, but you know, there's a lot of criticism in this space around what they call wrapper products and practivated isn't this, but essentially that layer on top of an LLM layer, like what you were talking about with Black Chat, G B T. But what you're saying to me actually is that in a lot of ways those wrapper products are, can be really important because they're, they're identifying the heuristics and the ethics and the bias mitigation to ensure that the outputs are aligned with your values, your experience, your. And so I just really love that kind of reframing because I think that's really important and really important to understand. I think with the right type of heuristics and guardrails and controls, you actually can have a tool that is much less biased than a human. Like that's something we work really hard on. And practivated is like as a coach, I am going to have my bias if I listen to somebody give a, you know, a ten minute donor conversation. And I have my ways right. That I can't control because they are my lived experience. But actually through how we control the platform, we can ensure that people's feedback is less biased than any human coach would be. So you just, you have me thinking about so many things. Okay, tell, tell everybody where they can go to connect with you to learn more about your AI work and then of course about the organization.
[24:50]
Deanna Hoskins
Thank you. You definitely can reach out to us at our website@jlusa.org and there's a contact link there. You can email us or you can email me at deanna d e a n n a l usa.org and I'm loving having these conversations because we are literally moving to a tool. And really I want people to think about whatever population you're serving, whatever your passion is, what does that look like for that community and that culture. And that's where your voice has to speak. As these tools are being developed, as you're interacting, it only can adopt the culture and give the response if the culture is inputting information as well. So really, you know, when we circle back to the purpose of the call and really focusing on philanthropy, philanthropy has to focus. The tools are important. But if I don't have the capacity, if I don't have the education and understanding, the tools are just going to be there. And again, after the trendy thing of AI becomes normal, funders aren't going to fund training and capacity. So we have to get it now because they are widening the net of disparity, thinking that communities and organizations are already there and they're not. So we have to start at ground zero and then we have to use. You said something when I came into this conversation, the quick and easy solve this problem in a year. This is again, you can build a tool today that's going to be outdated six months. So how do you invest multi year in not only the capacity, the training, the governance, but then the scalability and building Because I always say Apple iPhone, never changed their name, they're just on version 18. Yeah, it's the same phone but they got to keep updating it because guess what, the old ones, you know, my kids tell me all the time your phone's outdated. I'm like, it still works, but I don't have superior picture taking like they do because the tool is actually outdated.
[26:59]
Mallory Erickson
Yeah. Well, thank you so much for joining me and for all the work that you're doing to support the sector.
[27:06]
Deanna Hoskins
Thank you, thank you for having and thank you for having this important conversation so we can elevate it.
[27:18]
Mat
I hope today's episode inspired or challenged you to think differently. For additional takeaways, tips, show notes and more about our amazing guest and sponsors, head on over to Mallorykson.com podcast and if you didn't know, hosting this podcast isn't the only thing I do. Every day I coach, guide and help fundraisers and leaders just like you. Inside of my program, the Power Partners Formula Collective. Inside the program, I share my methods, tools and experiences that have helped me fundraise millions of dollars and feel good about myself in the process. To learn more about how I can help you, visit MalloryErickson.com PowerPartners Last but not least, if you enjoyed this episode, I'd love to encourage you to share it with a friend you know would benefit or leave a review. I'm so grateful for all of you and the good, hard work you're doing to make our world a better place. I can't wait to see you in the next episode.
[28:18]
Practivated Announcer
Fundraising is hard. Every donor conversation carries pressure and most fundraisers are expected to just figure it out through trial and error. That's why we built Practivated, the first ever AI powered donor conversation simulator made just for fundraisers. It's a safe, judgment free space to practice your pitch, refine your storytelling, and build the confidence that drives real results for your mission. Because conversations move missions forward, with Practivated, you and your team can practice anytime. Get real feedback instantly and walk into donor meetings ready, not rehearsed, but prepared. See how practice changes everything? Try practivated today at www.practivated.com and start building confidence one conversation at a time.