Name: AI Applications and Safeguards for the Future-Forward Finance Office
Uploaded: 2026-02-10T19:31:04.844Z
Duration: 1 h 28 s
Description: AI Applications and Safeguards for the Future-Forward Finance Office

Transcript for "AI Applications and Safeguards for the Future-Forward Finance Office": Hi, everyone, and welcome. My name is Caroline, and I'll be kicking things off today. Thank you for joining us for AI applications and safeguards for the Future Forward Finance Office. Before we dive in, let's quickly cover a few housekeeping details here. For engagement tools, on the right hand side of your screen, you'll see the control panel with chat, docs, polls, and q and a. These are your tools to interact with us throughout the session. In the docs tab, you can download today's webinar, but we'll also send you an email of the webinar after the recording of the session. For polls, all poll questions will appear in the polls tab. Each poll stay open for a few minutes so you have time to respond there. If you're looking to earn CPE credit, responding to the polls is required. The Polls tab will disappear once the polls close. For chat and Q and A, feel free to chat with us and other attendees during the webinar. If you have questions, submit them in the Q and A box in your control panel. We'll answer as many as we can during the session and at the end. For any streaming issues, if you experience any problems there, just try refreshing your browser. This usually resolves the issue quickly. We also have Ali, one of our representatives, on the line today. Ali is available to answer sales specific questions and learn more about your organization's needs. Connecting with Ali is easy, usually just a quick five to ten minute chat or call, and this is just the fastest way to get information information on how Blackbaud can support your organization. After your conversation, Ali can connect you with an account executive for next steps often as soon as the next business day. If you are a current customer and have a question about your software, we recommend using the chat or calling the support number provided here. And for CPE credit today it is available for the session. You must answer all three poll questions to receive the CPE credit, and certificates will be sent out within two weeks of the webinar over email. If you have any questions or issues, please reply to the email or contact this email here. And with that, let's get started. I'm excited to hand things over to, Paul, Mike, and Dave. Caroline, and we're thrilled to be here today. We've got quite a lot going on in our title, AI. I know I've probably heard it scientifically speaking used over a million times the past month and a half. I don't think that's gonna stop anytime soon, and we don't have all the answers. But what we're here to do today is to talk about generative AI, how it's changing nonprofits from a finance and operational perspective, and and what you can do to make sure you integrate it in a sustainable and and thoughtful way into your organization. So I'm pleased to be joined by my co presenters today. First, I'll introduce myself then kick it over to Dave and Mike. My name is Paul Prezziati, and I'm a partner with Johnson Lambert. Little bit about our firm. We're a national CPA firm that provides audit, tax, internal audit, and other consulting solutions to the nonprofit community. Myself, I colead our national nonprofit audit practice. And in addition to my day job, spend a lot of time doing presentations such as these and serving on various audit committees and and boards of directors. Dave, I wanna kick it over to you. And for for those of you who can see, Dave has the best title out of anybody on this webinar. Dave? Thank you, Paul. It's it's something to live up to. That's for sure. Yeah. Like Paul said, my name is Dave Fuge. I'm the chief innovation officer here at Johnson Lambert. Been with Johnson Lambert for a long time, about twenty years. And in in my new title, which I acquired a couple years ago, I transitioned from chief information officer to chief innovation officer. And I just like to joke that that just meant I got more responsibilities, and all the other responsibilities of CIO and and now all the new fun ones, which is really to help make technology work for us and for our clients and to drive appreciable benefit to the firm and value to our clients in using those technology pieces. I live and breathe AI. I've been working with AI in some capacity for the past five or six years. That has significantly changed, recently, and, I am super excited to talk to everybody about all the new developments and how we can think about using AI responsibly in our organizations. So thanks for letting me come and present to you today. Mike. Okay. Thanks, Dave. Equally excited to hear from Dave, but also just to let you know, I'm Mike Gellman. I'm coming to you from the Washington DC area. For my whole career with nonprofit organizations, mostly in public accounting, now with fiscal strategies for nonprofit, we're a consulting firm sitting alongside the C suite, sitting between a CFO and a CSO, Chief Strategy Officer on sustainability. And capacity issues, Dave, you're going to see that we have a solution here for capacity finally for nonprofits. And I think it's really one of the biggest hurdles that we get through. Like Dave and Paul, I split my time in thirds, about third of consulting. Third of my time is teaching. My home base for teaching is at Georgetown University here in Washington DC. I teach for a lot of national organizations like ASAE and the United Way and National Urban League and such and different other component organizations. And my wife and I try to spend a third of our time being active volunteers and I think it's really important when we sit on as many boards and committees as we can and to participate. But I think you learn a lot by sitting on these boards and giving back and then when you're talking to your current boards, it's good to let them know that you've served that volunteer role. Sustainability education for nonprofits is an all volunteer effort. We kicked that off when COVID hit and we've been posting every week. It's a site. We have podcasts, articles, resources, templates and things. And I'll shoot you out a link to that here in the chat and such. So with that, Paul, let's go back to you. Thanks, Mike. Alright. So let's get into this conversation and and get it kick started here. And, you know, I I think it's safe to say I'm not the only one. When I look back from where we were in February 2025 to February 2026, the game has changed in that. It really was more of a, quote, unquote, concept, a hypothetical when we spoke about generative AI at this time last year to flash forward to 02/10/2026. A lot of us are, a, using it in our everyday lives, and, b, it's becoming a part of our workflow at our organizations. And nonprofits have a lot to gain when you think about generative AI and incorporating it into your workflows. And like anything, you have to have a thoughtful strategy for how you're gonna build a culture surrounding adoption, how you're gonna have the safeguards. Said another way, you can't just throw a clot in the mix and expect it to work. You're gonna have to have a thoughtful strategy for how you do that, and that's what Dave, Mike, and I are gonna talk about today. So you can see here we have four critical pillars for adoption. I I do apologize this font is a bit light. We will get this edited and and send out and update it to you. But the pillars are described a little later in the presentation in that they're this, Exploring the potential of AI for nonprofits, what it might be, thinking about data, having a strategy for that data, ensuring it's adaptable, ensuring it's clean, ensuring it could be integrated across multiple platforms, then thinking about the cultural side of things and building a culture where AI is inclusive, we have safeguards surrounding AI, and then lastly, thinking about everything from policies and procedures perspective. As a compliance guy, I love policies and procedures, and Dave has a lot of those that we can discuss with you as a group. Alright. So let's get into the the conversation about AI here. And, again, when we think about nonprofit organizations, right, we can we can pretty safely say for the most part, they're always light on a few things in terms of having the necessary resources and using those resources to be able to provide the the mission in the most effective and efficient way. And that's really what we're gonna talk about here in in terms of enhancing these areas with AI. And and, Dave, wanna get your thoughts here. I mean, these are just a couple of concepts. But when you think about enhancing communications in AI, what immediately comes to mind for you? Yeah. I think this is a this is a a great place to start, Paul. And I wanna maybe backtrack for just a second. Paul hit on this, and I wanna just spend a little bit more time talking about it before we jump into some of the use cases because I think what's happened and this is just such a fast moving area. There are so many new things that are coming out all the time and advancements, and it's just so difficult to stay on top of all of these these things. I read a lot, of publications coming out from the the popular AI labs, like Anthropic, Google, OpenAI, etcetera, and and what they're working on and the challenges and and the things that they're solving at their organizations as sort of a a bellwether, an an an understanding of what's possible. And like Paul alluded to in the very beginning, when we started out in 2025, most folks were looking at these tools as, you know, I hate to be reductive here, but a fancy way to Google something. Right? And a a much more efficient way to to to Google something. And, really, a lot of these frontier labs have been using this tool these tools in more comprehensive ways. And I think this phase change that happened over the course of '25 and in '26 created something, that a lot of industry folks call overhang, meaning the capacity for where the tools are today versus what we use them for because we think we know what we can use them for today. And so while we talk through some of these examples, I think it's important to consider this overarching concept that what you thought AI was capable of doing yesterday, has been replaced with what it is actually capable of doing today. So when we talk about things like donor outreach, right, automating and personalizing solicitation letters and emails at scale, most people are gonna jump to thinking, oh, it can write these emails for me. And I what I wanna help broaden the scope for everyone on this is to think about not just drafting the email. Drafting the email, sending the email, updating that email in a source system, generating replies to people that send those messages back to those outreach emails, performing a job function from end to end, a process from end to end, not just one little piece of that. And I think that's where the overhang is today. A lot of organizations have moved beyond just having it do one little piece. So it's integrated in my email and it'll rewrite my emails for me. And that is absolutely helpful, but the capacity of these tools to do more for an organization is there. It's just the org most folks have not gotten to a place where they can realize that benefit. So content generation. Right? Draft compelling narratives and summarize research for grant proposals. Absolutely. Yes. That that is something that can be totally done. What about researching and finding those grant proposals first, then drafting the content that's required, and then facilitating the submission process? That might sound, you know, a pie in the sky, but the tools are capable of performing these multistep actions is really what I'm trying to to get get across. Let's talk about marketing real quick. Generate engaging on brand social media content or press releases. That is absolutely doable today. Another way to expand the use of AI, what if you could develop marketing plans for specific people that you're looking to market to and make them individualized campaigns? Rather than one larger broad campaign, you could leverage AI to run a thousand campaigns. So while we're thinking about this, these are just sort of ways to be sort of expanding the use of AI to gain an understanding of how some of these tools can be used beyond what we think about them or what we thought about them in 2025. So these things are are possible today. What I just described to you is possible today using tools that are available to organizations right now. That's one of the things that I wanna make sure that comes across in this talk today is the the capacity for you to use these tools to affect real change and do real work in your organizations is absolutely there. It doesn't take away from what we're gonna talk about in a couple slides, which is the requirement for governance and and, data accuracy and all of those fun things. But I wanted to maybe level set here and talk about what's possible right now. Right now, it's beyond the chat window. We have moved into actually completing business tasks, and I think that's a real game changer for many organizations. So what does all this really result in? You know, the AI is able to do these things for you, you know, greater efficiency, better engagement, and ultimately, we're looking for increased revenue. So just as a primer and through the lens of fundraising and communications, let's start to think about these AI tools as more than just a chatbot. Let's, Dave. Mike, any thoughts from your perspective? Yeah. I want to piggyback one thing on Dave and see if you agree with me. And we've actually had this conversation. It's coming back to the capacity. You know, all through my career we've seen technology changes that has increased our capacity, but this is a sea change for capacity. If you look at the average nonprofit organization, they've got 70 of the resources tied up in labor. And for the first time we can take that 70% of our resources and see a day, Dave, where we could be doubling and tripling the output of what they're doing by using the AI tools because AI is becoming and think of all the time we've spent time, you know, just back on basic research, okay, fact checking, doing things that now we can use AI. And in the past it's always been, oh, can we sneak in another staff position? Do we have a little bit of room on our labor line, which is 70% of our budget to repeat once? So I just see this as a huge transformation. And organizations that don't jump on this will be behind. Do you agree with that statement? They'll be able to almost like virtually double their staff because the. capacity will. go up. Easily. Yes. Easily. I think doubling is a conservative estimate, to be honest. And yeah. that's a pretty semi shocking statement. I mean doubling anything, you know, increasing the capacity or anything 50%, 100% kind of how you look at it, just mind bending where in the past we would settle for a 5% or 10, you know, operational efficiency change. I can't you know, how we use this and Dave, you just said, you know, we're on the cusp of how we'll be using this. It's changing week to week, actually, month to month, year to year. It's going to and we you have to get on, you know, out in front of that, Paul. No doubt. And where we'll be in February 2027? I I don't like making predictions, but I can say with certainty it'll be different a year from now than what it is today. And so I I think it's one of those things you're gonna need to continuously adapt on. Dave and Mike have already touched on on a bit of this and, you know, the power and potential for automation both from a reporting perspective and and data entry. Where I wanna take the conversation a little bit is the positive. Right? Because it's natural to hear this and say, is this going to take away my job? My stance on this is it's going to change jobs. And I think especially for nonprofits, it's gonna change it for the positive in that. I I challenge each and every one of the 900 plus of you on the webinar today to think about how much of your day is spent doing monotonous administrative tasks. Even for me, and I I probably shouldn't say this, but I'll ask my firm to put on their earmuffs, I probably have too much administrative tasks in my day. But imagine a world where let's just even say that administrative time is reduced by I'll use a small number, 15%. What does that mean? It means you can focus your time on where you can provide the most value. Would you agree, Dave? Yeah. A 100%, Paul. That's absolutely right. I think it just opens up more capacity for more valuable work that people want to be doing, most people want to be doing, and take away the stuff that they they don't necessarily want to be doing. Right? I think that's that's ultimately where a lot of this tool these tools are gonna drive that efficiency 100%. The the the thing I also wanna make clear on here is I I'm a technology person. Right? And I'm speaking from a technology seat. When I say things like these tools are available to you as an organization today, I think it's important to understand that I'm not talking about your developers. I'm not talking about those that have a specific programming skill set. I am talking about these tools are available to affect actual work and do actual work at an organization based on your ability to write language that is human understandable. And I think that's the other really big aspect of this. Many other technologies out there that when they came into the fore required you to have a set of developers to fully make use of the tool. This is unique because it's interacting with people in a way that is with natural language. Meaning if you can write it down, the system can understand what to do and how to do it. And I think that's the real game changer here as well. These tools are powerful and can do all of these automations and these sort of things, but it's the fact that anyone, you don't have to be a developer, can wake up in the morning and think, I wanna do this today and use an AI tool and make that happen. Right? There are some risks that in is gonna be involved in that, and I'm sure that we're gonna talk about that. But I wanna just drive the point home that it is not for a set of developers to use AI in your organization only. These are tools that are accessible and available to anyone to make these sorts of improvements. This isn't this isn't constrained by your ability to understand code or software or anything like that. I think that's really powerful here too. And Dave, think you can add that it's not a threat to people's positions. It's actually an opportunity. Because before AI, people were constrained by their job description and the systems they were operating within. Now they can get back 50% of their day and then expand what they can do. So I think the effective speeches I've seen, you know, to staff is that utilizing this tool is not a threat. It's actually an advantage for you to advance both yourself and your organization. 100%. That's my core belief as well, Mike. I mean, I look at this as a multiplier, not a subtractor from an organization. Exactly. And I think. that's really what what one of these tools, can do for you. We we have this slide about boosting programmatic impact with which I think is important as well. The service delivery capabilities of these tools are absolutely there. They can analyze large datasets, define find trends. They can predict, community needs more accurately based on a whole bunch of different sources. They they are very good Swiss army knives, if you will, at being able to tackle a whole bunch of different challenges that your business or your organization might be might be encountering. So, yeah, they they are absolutely hungry for data and able to organize and summarize that data in a way that helps you make organizational decisions. Alright. So we've got our first polling question here, and I'm gonna kick it to Blackbaud to push that out again. Remember to respond to the polling question in order to get CPE credit. And I I promise you a lot of you are probably asking, you're telling me what's possible. How do we do this? We will get to that in just the next section. I I promise you we will give you real life examples. Okay. And can we see the polling question here? Yeah. It looks like the poll's up. Okay. Great. Okay. And our polling question is, do you use AI in your daily tasks at your organization and frequently, occasionally, or never. What do we think we're gonna get here, gentlemen? I'm I'm hoping it's not a never for the winner. I'm gonna guess occasionally, but I feel like we're getting more towards that frequently. Yeah. We'll talk about why frequently is becoming more and more important, but my guess is gonna be occasionally. I think that's what most folks are are are gonna say. Alright. I actually think that's a good sign because I think it was south of occasionally not too long ago, Dave. Very true. Okay. Just a couple more seconds, then we'll close the poll. Alright. Can we go ahead and get results for that poll? For those that can't see the poll, if you're looking at the chat, on the right hand side that says chat messages and poll, there's a little red dot over top of poll. If you click that, you're gonna be able to see, the polling question there. I see some folks in the chat not able to see that. But if you can look at the poll at the top, the red dot, press that, and submit your answer. Okay. And looks like the leading response is occasionally pretty much what we thought there, which is good. And and maybe if we do this webinar this time next year, we'll get to frequently. Alright. Let's talk data, and and data really is the backbone for how you're gonna make this work from a technical perspective. Dave, I'm gonna let you talk a little here about overview. You know, when we think about foundational data and AI, where do we start? Right? Because that's that's the core of everything. Yeah. One of the the most famous phrase phrases, garbage in, garbage out. Right? If your AI systems are relying on data that is flawed from the beginning, it's not gonna be able to be operationalized for you. You're not gonna be able to gain benefit from that if the data that it's being used to help form its opinions and form how it should be doing the strategy is in is inherently flawed. So, really, it does a true AI implementation across an organization really does hinge on high quality data sources for the AI to be able to learn about your organization, learn about the the the task that is at hand, and leverage the high quality data that it can trust to then help perform a job function. So, you know, we've we've listed that here in a couple different ways, but generative AI systems are really only good at as their source material. One thing I'll just say on the source material piece, and we'll dive into this a little bit later, is the source material might be a dataset that you have at the organization, but it might also be organizational context that is not documented. And I think that's another piece that we want to uncover here. We might have great systems like Blackbaud that keep a lot of data organized about our our people or our contributors or members of the organization. We have a lot of good information there. But what a lot of organizations don't have and where this might impact their ability to use AI effectively is some of the unwritten things about how that data plays into organizational activities. We'll get into that in a little bit, but just wanna throw throw that out there too that the source material doesn't just have to be a foundational dataset. It could be process documentation. How does this work or what are the guardrails I need to be aware of when I'm working with this data? Or who do I go to talk to to help me understand this problem or how to resolve this issue? And so it's those pieces of data as well that I think are, important as we talk about implementing AI. And then lastly, on the overview, it's just thinking about this as adoption in two phases. Looking for immediate productivity with human oversight and overview is important, and then migrating from that sort of experimental phase into a more operationalized and consistent deployment of these AIs. We'll talk about that in just a minute. Dave, and that's a good place to start. Right? And and I think you phrased it nicely in two different phases. Right? What can we do immediately and push out with some training, but maybe not the most extensive versus what's gonna take more time and efforts? And, you know, I don't wanna assume anything. So let let's start with the more immediate productivity and talk about some of the tools you're seeing used. Yeah. I think this is a great this is a great point. We're and, again, I'm not trying to jump ahead in our slides. I do wanna mention this, though. We talk about immediate productivity. It's about getting peep the people the tools in their hands to make these changes, to start working with them and developing that AI fluency across the organization that we're gonna talk about in a little bit. But some of the tools that I have really been impressed with recently are the ones that are more desktop focused. While there are we've talked about the different model, providers where you can go to their website and chat with the model on the website. There's a new class of models and a new class of products that are being brought to the the desktop essentially, bringing the AI to you onto your desktop to inspect and run operations across your computer. One of the most famous ones right now is by the company called Anthropic, their model Claude, and their desktop version of that model called Claude Cowork. And when I say desktop version, I wanna be clear. I'm not talking about running the actual model on your computer. It is a application that you install locally and you interface with the Claude model on your desktop. And why these are big game changers and why these are things that for these sort of immediate productivity gain is that they facilitate and can do multiple things across your computer for you, things that you might not, be ready for yet, and I and and we are not ready for yet. But just to give you an example, if you wanted to have Claude set up and reorganize all the folders on your desktop, it's one of the most common use cases that they throw out there. Can they get access to a desktop or get access to a downloads folder and perform some sort of cleanup operation for you? I want you to rename or tag all of these in a specific way, read all of them and figure out what they are and then tag them and and organize them accordingly. That's great. A cloud can also then go onto your browser and open up a browser tab and perform actions inside of a browser tab on your behalf. It's like the computer running a tab of Chrome in in your computer. All of these have security concerns, and I would caution you not to give Claude a whole bunch of credentials to go run with. But what I'm what I'm trying to lay out here is that this immediate productivity and the tools that are available today are ones that are ready for people to begin experimenting with. I see a question here from Colleen. Is is Claude Cowork HIPAA compliant? I would I am not ready to say that. In fact, anytime you use one of these models, whether it's Claude or OpenAI or Gemini, it's really important from a security perspective that your organization has entered into agreement with these vendors, you know, at least I should say from my experience, I would want to enter into an agreement from with these vendors that has protections around data privacy, confidentiality, and security. So I am we if we were to use a model, we would not be using a free and available model. We would be paying a small fee for access to one of these models so that we can get that assurance from the model provider that the data is not being that we send to it, is not being used to train their model, is not being read or or interacted with by the employees at the organization. So good good governance around the the third party that's gonna potentially have access to this data is absolutely important. Little bit of a tangent there, but I thought that question was really, kind of in line with what we were talking about. So we talk about these tools. I am talking about them in the context of we've already set up and established a a relationship with these vendors that ensures the privacy, confidentiality, and security of the data that we're using to send with them. I would not advocate for any organization to use the public chat GPT if you have private information and you're concerned about disclosing that publicly. So I feel like it's important to talk about that. But for immediate productivity, getting those agreements signed and in place and for a nominal fee, you can expand the power and the ability of your people on your team to play with these tools and figure out what works. Once you've sort of started down that road and begun to get that fluency, you can move down into a more advanced integration where AI is doing more for the organization by connecting directly to foundational systems like Blackbaud or like other systems that you might have in your organization that can actually function in calling all those those systems autonomously. Again, that's a much more advanced deployment, but I think it's it's important to understand from a foundational perspective what these tools are capable of doing, and getting that in the hands of your people is gonna be really important, to establish those skills. Yeah. And what Dave just said overlaps quite a bit with the section on culture and the. section on safeguards. I know a lot of you are gonna have thoughts, but in summary, yes, your employees should not be entering private donor data into a browser with Gemini. That's. a big no. Right, Yeah. Dave? Yeah. Unless you've got that agreement, you have an understanding on how that data is gonna be handled, the free version of any of these things, we should not be doing doing any donor data in there at all or or any private data in there at all. And Dave talked a little bit about the future, right, for what these tools can be used for an advanced perspective on in terms of fundraising, in terms of donor communications. That's that's really just the tip of the iceberg. But we can't have any of this without clean, solid internal records. And, Dave, my perspective here, this is where the real time and effort is gonna take place. Yeah. Yeah. Absolutely. Getting this stuff cleaned and and ready for an AI system to leverage is absolutely, where most organizations are gonna spend a ton of time. I think the the benefit of this we're talking about this in the context of AI, but the benefit of this goes to our people as well. If people know that the data that they're working with is accurate and up to date, they are able to produce more and rely on that data more just like a machine is. So while we're talking about this through the lens of of using AI, I think it goes throughout every operation that might interface with the person as well. It's kind of interesting too. Some of the most recent things that I've seen is companies actually using AI to help them clean their data so that that can then be used by AI. So, you know, it it seems like a daunting project, but there are new ways for for us to get this accomplished with with new tools and and much faster and cheaper than before. So it's absolutely critical though for for this data to be if we're going to be relying on this data, for it to be accurate and up to date. I think that also goes toward a culture of data governance and and and cleanliness. If everyone is pulling in the same direction about why this data needs to be up to date and accurate at all times, the better outcomes you're gonna have. Right? So this is why this becomes very, very important. Okay. And and going a little deeper into the data strategy here, both from technology readiness and also thinking about legal and and ethical perspectives. Mike, would would love your perspective here. Right? This sounds like an area where if you're going down the AI path, you need to consult with your attorneys. What do you think? Yeah, Paul. And I'd like to come back to my two c's. We talked about one which is everything here is driving capacity, which is huge organization. The other C is collaboration. And the opposite of collaboration is that when it comes to AI, Dave, I'd love to get your comment on this, is you're not operating on your own, on your use. Right. Before you use it, during your using, and after the output comes back that you're through a collaborative culture, Paul, right, that we're going back and forth. And this goes even outside the organization with legal counsel, with, you know, consultants, especially in the development area. People are around there. Also, and I think it's really big on AI, we're using it in the lobbying area. Know how misused, you know, when we talk about in development, it's huge in our lobbying communications and our call to actions. And so that each step of the way, that governor on this is making sure there's a collaborative culture and that you're bringing in these different parties to check. And that won't put a governor on We're going be using these tools but you're checking. And I think you can actually make it more efficient, Dave, that when you've got to check before you jumped into a certain AI task. Yeah. Especially if we're using the the data that we have at our organization to help train a model, even more so to make it more accurate and customizable for your organization, that's absolutely a a place where we wanna make sure that we have done all of the right data hygiene things. Right? Like, so anonymizing the data, getting consent, like you kinda alluded to, Mike, and and getting that making sure that we're adhering to all these privacy principles that we might have, general privacy principles, general best practices around how we use data and and and use that to help inform decisions going forward. So so, yeah, I think it's really important that we are upfront about the use of training models. We don't shy away from that. And and, certainly, you know, from my experience, I would love to be transparent wherever and I am transparent wherever possible around our use of these tools. And I think most organizations are going to be in that same boat as well. But I think just letting folks know where this is happening, you know, you don't have to advertise, but not shying away from those conversations is gonna be really important too. And to do that, you need to make sure that you're performing and using this ethically and with the right sort of data security approaches that we just discussed. Dave, can I give a quick example I think people relate to. is it was just eye opening to me? In a departmental meeting five or six of us were sitting around and what we discussed was the prompts that we were issuing into And so it was a short conversation. We came down and we changed our prompts and then the data coming out, of course, works much better. But also we're checking it from a safety point of view, from ethical and all that. Yeah. And so that discussion came from each of us going out and doing all this work to discussing the prompts that led to even better output from AI. Yeah. Absolutely. And I think one of the other the other tips that I'll give folks on on this call is to really think about the outcomes and talking with AI about the outcomes that you're trying to achieve. And building those expectations from the start, like Mike talked about the prompts, are are really helpful to tell the AI what to do and building a part of that prompt to say, here's what the expect expected output should look like. And I don't wanna go too far down the rabbit hole, but one of the things that we're seeing is the potential bottlenecks are gonna occur at the human review part of this. Right? And building in people's ability to review what the AI has generated. So getting your your AIs to think about how to make it easy for people to review what they've done so that it's not a bottleneck on review, you're able to see these benefits and actually have people over top of them, able to review them without it being this this, you know, tidal wave of of AI content coming at them. Right? It's easy to review and and and work with. And Dave, you know, we actually found that our prompts were leading to a bias. Oh, And it. forced us to redo our prompts. You know what I mean? And there's people are so quick to blame AI for something. But if you actually look back at the prompts it was leading AI down a rabbit hole. Yes. AI so now we're spending very easy. target, I'll add. With that in. mind, gonna advance us to our next polling question. And, again, wanna make sure you respond to this polling question for CPE credit. While everyone's looking at that polling question and the polling question is this, how many applications do you use that interact with AI? While everyone's responding, and we'll give you just a minute here, did wanna touch on one other thing, and that's the capacity side of things. You know, gentlemen, my sense is we laid out those two options. Right? Option one is the ready made tools. My sense is most organizations, and I'm never gonna say all, but most can probably incorporate the ready made tools with thoughtful lessons learned we're gonna talk about in the next two sections without, let's say, making significant alterations to to their organization, their capacity. Option two, Dave is living and breathing this. It takes a lot of time. And I would just. caution that if you wanna get to option two, I think you first need to take a step back and say, do we have capacity, and do we have the right skill sets to get this implemented? Yeah. Without a doubt, Paul. It takes a lot of time and effort to rethink processes in the context of of AI and using them even semi autonomously at scale requires considerable amount a considerable amount of time. Okay. Just one more second here. Alright. It looks like the answer that we have is one to two applications that you use that interact with AI. It sounds about right to me. You know, if you if you pulled our employees and what they're using now at Johnson and Lambert, I I think that's probably right. I again, I also think if we change this next year, it might be a little different. Alright. Culture. So as everyone who knows that works in a nonprofit organization, culture is something that has the utmost importance, and it's always shifting. And I I think a lot of us on this webinar saw a big shift during the pandemic when maybe we went from, you know, if not always in office, being a lot in office to to that ultimately changing. And operating in a hybrid environment, it can work very well. It's just different than an in office culture. And that same principle, Dave, that applies here to AI. Right? Yeah. Exactly. I think I think what we're looking at is, to be blunt about it, these tools are great if you use them. Right? And I think getting that and driving that adoption across an organization to see these benefits is gonna take some time and and help to get people used to these tools, some change management, some education. But what I've discovered in sort of our journey on this and and in advising others on it is really repetition and using the tools as they best fit for the people that are working in those roles. So non prescriptive. There's suggestions. There's examples, but allowing people in teams to figure out how to use these tools and understand the the best use cases for them and where the the issues are with using AI, where it's good, where it needs help, that sort of stuff actually actively comes from using the tools. So making them available is okay but monitoring their use and understanding how folks in your organization are either adopting or not adopting them is really key here. You wanna make sure that if you wanna see this sort of AI adoption that you're you're hoping to see, to see those operational benefits, you're cultivating a culture that is focused on allowing AI to help the people, help people in your organization achieve more. Right? And that's through consistent daily use of these tools. So, you know, the goal, as this slide says, is AI fluency. Well, what what is AI fluency? And it is the ability to understand how to use this technology. And important to understand, this is not technical coding. It's how do you get the most out of using an AI tool, when you want it to complete a task? Do you know how to write well defined instructions? Before I even jump down this, I wanna maybe take a zoom out real quick and say, I read a quote about this and I thought it was really accurate. It basically said, you know, if you're a good manager or if you know how to break problems down into small digestible parts, if you know where people that are gonna be doing this task are gonna get hung up, there's gonna be questions, there's gonna be issues, and you anticipate those and tell them where to how to resolve them. If you are explicit about what the expected outcomes are supposed to be, what they should look like, what they shouldn't include, how to validate that they've done this thing correctly. If you can write that for a person, you're gonna be able to do that for AI and that's really what AI fluency is all about. It is getting the machine to do the thing that you want it to do and it being able to understand where it could go wrong. Right? So we want people with AI fluency the AI fluency to be a a cornerstone of this. So this is summed up here. Great. Prompting is delegation. It's all about human the same way that you would tell a person to be successful, using AI is the same way. You want the AI to be successful, it needs the right information to make that happen. So the best way to do that is to do iterative collaboration. You want it to not just expect what what everyone calls a one shot prompt, meaning you say one thing at one time and expect it to be completely right. What you are trying to do is ultimately iterate with the machine to get to the right answer and then once you get there, you document the things that you that the steps that it took to get to that answer and you can save those and then reuse those in the next in the next iteration of that. We want specialized and targeted training. We wanna develop these pathways that allow people to understand the different ways that their departments can use the tools and how to get the most out of them. And I think the training component is so critical, and that's where a lot of the thought is gonna go into. If we're gonna go down this path, the path two that we talked about before, the rollout, it's it's paramount that you have examples. You have a forum for people to collaborate. Mike, because we've seen this in other areas for nonprofits. And just throw aside AI. Let's talk about tools or new team members. If we just throw them to the wolves, I would venture to say this chance of successor are zero or near zero. And and I think this applies to AI too. Would you agree? 100%, Paul. And Dave hit the nail on the head. You can cultivate culture. Culture does and make it better. And the way I like to look at culture here, Paul, is we so often talk about we are enabling bad behavior. What we want to do is encourage good behavior that's through collaboration and encouragement. And these are like simple words. Why are we using these simple words when it comes to AI? This is how we do the iterative process. We demonstrate that working together through our prompts, back to that example, the trial and error that goes on and how we can increase our capacity and performance. And so we're shifting the culture to enabling good behaviors. And the one thing, Dave, I think that's really where organizations slip here, and Paul, I know we're going to get this into this in next section, I've got to say it here, is that they allow solo actors to act. And then all the governors and the risk management and all the things go away. So the culture has to be collaboration. We have to cultivate this collaboration and we share this information in real time. Dave, wouldn't you agree with that? Yeah. I think that's that's really important. I think giving people the freedom to explore and find new ways to use the tools is important. But to Mike's point, bringing that information back and sharing it so that you're fitting within your governance framework and that you've you're you're all around the guardrails that are that need to be there. You know, I think having an environment that allows people to explore and try new things is great with the right governance and guardrails over top so that they can't do anything that would be harmful to the organization. I think, to Mike's point, sharing what they've learned through the use of these tools is exceedingly important. Sharing that across the organization. Yes. And the other thing here that, again, is is is so very important and also too, this goes back to what we talked about before that this this is akin to other things nonprofits have adopted over the years, leadership sponsorship. Said another way, tone from the top. I can guarantee you if your senior leadership is not supporting and not promoting these tools that they will not get used, and they will not get used effectively. You'll just have people you doing a scenario, Clog gone wild or Gemini gone wild. I I think you need the tone from the top to say, we know these tools are new. We knew know this is gonna be a period of experimentation, and that's okay. But, Dave, I think that has to come from the highest levels. Yes. Absolutely. Leadership participation and setting the tone absolutely has to has to happen without a doubt. Alright. We've got another polling question here. Again, you wanna make sure you respond to this in order to get CPE credit, and we'll push that out to you right now. And the poll is this. Does your organization's leadership champion AI? I I wanna know you are in a safe space when you respond here, so please respond honestly. Your organization's leadership will not see the results of this poll. I'll give you just a couple of more seconds here as we're getting short on time, but the last section might be the most important. It's one Mike and I love about internal controls. And, yes, we do have to have internal controls for AI because we don't want that Terminator two scenario. Alright, Dave? Absolutely not. If anyone's listened or heard about something called Motebot or Clawdbot or OpenClaw, the, the concept that Paul is talking about is absolutely real. So if that if no one's heard of that, I I encourage you to go look up and see what's happening with that. Yeah. And we're not gonna steer the webinar this way because we have nine minutes, and that is a ninety minute topic for discussion. that's a whole separate conversation. That's right. That's. right. Alright. It looks like somewhat is our leader here. Okay. Last section, and this one might be the most important one. AI is a tool that is being used now. It's gonna be used more than ever. And safeguards are important when we consider this, and it's also important to make sure we have the right governance policies. Again, you know, I don't have to tell you this. These slides speak for themselves. Really, what I wanna get into is is the meat and bones here. Let's talk about algorithmic bias, Dave. Where. where should we start? Yeah. I think Mike had a real really great example about that. When we're talking about two different kinds of AI and I know we're running short on time, I don't wanna spend go down too big of a rabbit hole here, but we're talking about LLMs. Right? When we buy an LLM access from Claude or or OpenAI or from Google, we don't have a lot of visibility into the training of that model. Right? We don't see how that model is what data was used to to get the the responses that it's it's able to give to us. But there are other models that might be tuned specifically on the data at your organization and understanding where the data at your organization may be biased. Or if you're using data from one of these providers, how do you understand the the the data biases or things that might actually might be in the the training data that is used to create these models? And most of the providers now, Quad or Anthropic, Google, and OpenAI, all create something called model cards, which you can review to see some of the overarching tests that they run to mitigate bias, but you have very little control on that. So I think understanding when you're looking for these tools to make these sorts of decisions for you, it's kind of opaque in that you don't necessarily know the data that it was trained on. You have to actually ask it for its justification in some of these. Model bias is real and can basically, it can it can cause things unintended actions to occur based on prior history. Mike, am I off base here, or am I hitting on this this the same way you. would yeah. I I think you're there, but I think on the flip side of it, Dave, and I know this is going to be a hard pill to swallow, is that bias needs to be internally managed Totally. also. And I think you have to step back and we've always learned from our mistakes, right, Dave? Yeah. And. you step back and you look and you come up with a conclusion, you know, which was helped by AI and say, why didn't that work? And it's like, well, that was the conclusion we were hoping for. You know? And so you do that through, again, back to my favorite C word, collaboration, that we step back and say, Hey, is this what it says or is this what we want it to say? Yeah. And then, Paul, like you said, sometimes we need to reach out and get a third party, another set of eyes. This comes back to internal controls, right? To me internal controls are Murphy's Law. We check and we keep rechecking and we keep rechecking. And the minute we stop checking and we let solo actives, someone who gets a set of data out and they start making all these decisions and no one else has said, hey, does that really make sense? Meanwhile, that data might have saved them one hundred hours. Right? AI was able to generate in the seconds. And they're just but we always gotta be you know, a little bit of Murphy's Law, we always gotta step back and be Paul, you always say it best. I mean, we have to be slightly, you know, suspect of everything that comes out, right, until we prove it's good. Yep. Yeah. And the other one that's listed here, I'm not gonna spend a lot of time on it because we touched on it earlier in this presentation. I'll just give you two pieces of perspective because there are a few questions on this. Data privacy, as Dave mentioned, absolute must that any vendor you're using, you have some sort of data privacy agreement with. And then number two, I do think it's important that you make sure you have legal review of your policies from that perspective as well. Likely, a technology consultant can assist you there, but just making sure you have all your bases covered. Last couple of slides relate to each other in terms of responsibility and audit trails and human in the loop. Let's talk about this for the conclusion to our conversation today. We should never ever, and I'm gonna use definitive terms, have AI making final decisions without human verification. Accurate, Dave? 100%. That's where we where I am today. And until I see otherwise, that's what I'm gonna recommend for for the for the time being. Yes. Without a doubt. Humans need to be looking at whatever is being produced. 100%. And from a policies and and governance perspective, is this a should this be in a policy? Should this be in your internal control documentation? Where where do nonprofit organizations start when they say, I wanna use AI, but I don't have any policies and procedures? I think starting with the most basic policy that's out there in high level terms on how these would happen, all outputs need to be reviewed before being submitted. You know, the work that AI does is not able to be pawned off on AI. You are responsible as as an individual for the work that's being done here to verify that this stuff is accurate. Yeah. I I I think that's just at a basic level, let's get some understanding around the organization about what we expect people to be use how we expect people to be using these tools. And for most, it's just a a simple policy document. And the policies, they need to be augmented by what I call working rules because policies are more formal and they're sluggish. They're slower moving. But. our working rules and our practices and procedures get pulled back to internal controls, that the checks and balances are in there and we can always move a practice and procedure into a policy and augment it. Thanks, Dave and Mike. Got a last polling question here. If we can assist you in talking about matters related to AI, nonprofit sustainability, or other internal control related matters, please don't hesitate to respond in that poll, and and we'll make sure we get back to you. We have a couple of minutes. We we've hit on a few of the bigger questions that we have here, and most of them relate to data privacy. Dave, are are you familiar with the concept of digital twins and data realis? Oh, like, data rails. Yeah. Yeah. Data rails. Yeah. And digital twins. Digital twins basically being a digital representation of a system or a product that is is able to be simulated in a in a digital environment. It looks like Rick submitted that. And, Rick, I would love to follow-up with you on your specific use case there, and and we can talk more more about that if you'd like. One thing while we're waiting for some questions to come in, did wanna point out the folks at Blackbaud have a great integration already with one of the tools we've already talked about, Claude. And I wanted to highlight that. It's available on Blackbaud's app store where you can actually hook Claude, one of these LLMs, up to the data inside of Blackbaud and have Claude help you search through the data stored there, leverage it for marketing campaigns, slice and dice the data that you'd like. It it's a really cool and powerful integration. So for the folks that are interested, I would highly recommend looking at that. It's a way to help give your agent access to the really reliable data that you have in some of your systems like Blackbaud. So wanna make sure everyone saw that as well. I see some. comments about not using free tools. Just to underscore that, do not use the free school free tools with your data. Make sure that you're entering into a Teams or enterprise agreement with OpenAI, Claude, or Google. All of them have programs. I I I shouldn't say that. Most of them that I've looked at have programs for nonprofits, and you can make use of those those terms, those advantageous terms from those organizations. So, yes, make sure you're getting an agreement signed before you do anything with that data. All right. Dave, Mike, I want to thank you both for your time. I know I learned a lot today, and there's still a lot to learn in the world of AI. Our team at Blackbaud, any any closing thoughts before we have everyone sign off? Yeah. Thank you, guys. That was great. Just to kinda go over some of our sources, we have a few blog posts, resources, and then also today's slides posted in the docs tab. And also feel free to reach out to Ali if you guys are interested in hearing more about our solutions. I also just pushed out those two last poll questions, not needed for CPE credit. But if you do wanna respond, if you want someone to contact you, if you're just interested in being contacted just to work with us or to work with them, whatever you guys wanna do. So feel free to respond to those, and thank you guys for your time. Thank you. Alright. Thanks. Thank you. Have a great day, everyone. Thank you.