Name: Inside Track to the Unified View-Chat for Blackbaud AI
Uploaded: 2026-02-10T17:30:59.026Z
Duration: 56 min 20 s
Description: Inside Track to the Unified View-Chat for Blackbaud AI

Transcript for "Inside Track to the Unified View-Chat for Blackbaud AI": Welcome to the uni the inside track to the unified view. Today's session will specifically focus on the development, use, and future plans for chat for Blackbaud AI. And while AI chat is obviously not a feature within database view, the fact remains that as Blackbaud lays out plans for the development of the unified view, we also have an obligation to incorporate the latest technologies into Razor's Edge NXT. And AI, love it, hate it, feel indifferent to it, AI is undoubtedly a technology we will be reckoning with this year and into the future, both Blackbaud and our society as a whole. So this team that we have assembled here today focuses specifically on innovation and incorporating new technologies like AI and others into the system. So that as we get to the unified view, we are not simply recreating the most effective tools from the past, but we are also helping build toward the future. So let's go on this journey together. But before we do, some housekeeping notes for those of you who have just joined. Yes, we will email a link to the webinar recording to you. And the audio for the webinar is broadcast through your computer speakers. If you have any glitches, like you don't hear audio or the slides freeze and don't keep up with what the speaker is saying, generally refreshing your browser is the most efficient way to resolve those. For questions to the panel, to the individual speaking, use the Q and A tab. And in the docs tab, you'll see a lot of references for the content that we are discussing today. Also, I've got because this is a roadmap oriented session, I want to remind you that what we are talking about deals with future looking projects and developments for the software that we are working on right now and will be working on in the future. This is subject to change at any moment. So just be aware, as you make plans for the software, that you should be basing your decisions on features and functionality that are currently available. Okay. My name is Catherine Hall. I am a product marketing manager with Blackbaud. Many of you may not recognize this group that we've assembled for you today. All of us work on technology innovation at Blackbaud, creating ways to apply leading edge technology to concepts that in ways that create practical, effective solutions. We are all here to introduce AI chat to help you understand and begin to use the latest AI technology if you wish to. And let's see. I'm going go to the next slide to explain who will be presenting what part. So user experience testing, Doctor. Lorraine Curry is going to be talking about how we developed the initial use cases, use cases being the problems that we attempted to solve, how we decided on what we would focus on first. Then Leah will discuss how the Blackbaud ecosystem is evolving to support security, privacy, and consistency across products. Lindsay will cover resources where we've developed to help you use AI chat. And Carrie is going to bring it home showing you some of the new directions we plan to take chat. And before we I turn it over to Loreen, I want to set up a conversation with her with the premise our innovation team begins with, that you are not looking for software features just for the sake of having more features. We begin by trying to understand the problems that you need to solve. And only when we think we understand this do we consider the role that AI technology may or may not play in helping solve some of these problems. So our goal is to create better tools that will help you coordinate efforts to raise funds for the causes you care about. And how do we do this? We are always looking for the magic intersection between your goals, as expressed to us, what you need to accomplish in terms of mission work, and what we could technically create with our software. So to drill down further into how we determine what to build, I'm going to turn you over to Lorena. Thanks so much, Catherine. And yeah. Hi. I'm Laureen. I'm a senior principal researcher. And what I'd love to do is just to give a glimpse into how our research has supported our approach to Blackbaud AI chat. So as Catherine talked about, we we know from our research that people don't want AI for the sake of AI itself, and a core part of our responsible AI approach is to start with the problem. So I think it's really important for us to share what's happening behind the scenes as we innovate responsibly with our focus on trustworthy AI and share some of the research insights that's really helped us prioritize what customers need from AI enabled services, which continues to inform our approach as we innovate intentionally. And as I walk you through some of these needs, I would love you to keep in mind that this is a journey that we're going through with you in terms of research, and we'll continue our research to ensure that we're exploring other use cases to evolve the experience of Blackbaud's AI chat. So a big part of our innovation journey is about understanding where social impact professionals are at now. So to be able to meet our customers where they are and our Blackbaud Institute research involving over 550 fundraising professionals provided insights about their fundraising performance, revenue changes, challenges facing the sector and the use and adoption of AI tools. It highlighted that usage is prominent in fundraising organizations with 82% of all respondents saying that they use AI and whilst this may be the case since the technology is still relatively new, we needed to dig in further to understand more about how you really need to use AI in the context of the Blackbaud's ecosystem. So some of the key insights highlight our customers are thinking about AI and how it can supplement, not substitute. So our customers have told us that they need AI to support staff and the work that you're already doing. And from a product perspective, this means that we needed to design experiences for human in the loop, transparency, and control. In terms of collaboration, this meant that we had to give real consideration around the role of human input. So for example, you might see this showing up as the ability to review a history of interactions and by the very nature of of how you can use Blackbaud's AI chat to brainstorm strategies, for example. Structured support, so this was all about guidance, not guesses. So our customers have told us that they don't want generic AI magic. They want structured support, clear recommendations and context aware assistance that reflects your organisation. And accessible entry point, so many social impact professionals have expressed value as seen in terms of AI and fundraising, but the use cases aren't always, you know, imagined. And this is where Blackbauds have a responsibility to really show these use cases and and take you on that journey. So Blackbaud AI chat is a key example of an accessible entry point providing a way for social impact professionals to explore ways to utilize this technology as part of your solution. And impact over novelty, so any AI enabled service needs to support fundraising outcomes. This isn't about use of AI for AI's sake. And finally, they're reducing cognitive load, so social impact professionals needed help navigating complex donor data and scaling stewardship, which is where you might see functionality such as being able to surface high potential donors based on engagement history and available data. So when we talk about an outcome driven approach that we take, what this really means for our teams is that we need to start with the problems our fundraisers are experiencing. And the problems that grounded our team's thinking around Blackboard AI chat included time and resource constraints, so limited ability to segment donors, backlogs, and stewardship, too many donors per staff member, lack of time to adapt campaign strategies. This is a consistent theme. So what this meant is the AI must save you time, it must support outreach responsibly, and lack of confidence in what to do next. So teams were often, in our research, unsure how to most effectively tailor outreach resulting in what we were kind of seeing as this ladder rinse repeat approach to campaigns and outreach, and we're seeing staff uncertain what messaging might resonate with their constituents. So what this meant was that AI must provide actionable context specific support. Donor motivation and connection gaps, donors can feel disconnected from cause impact and and social impact organizations can sometimes struggle with time to personalize communication. So what this meant was that AI must support strengthening donor connection, not depersonalize it. And then finally, demand for transparency, control, oversight. So from both user research and customer FAQs tracking, customers want oversight. They want to review and edit content, so AI chat must be controllable, editable, reviewable, and governed. So the above problems highlighted raised key use cases to focus in on at this point, including identifying top prospects to contact, understanding key information about prospects, and brainstorming fundraising strategies and and generating donor engagement content. So just get getting a bit more specific based on the outcomes I just talked about, this then gave our team insights needed to consider the initial priority use cases for Blackboard AI chat, focus around areas such as campaign management, fundraising strategy, donor insights, and reporting. So, for example, these priority use cases included supporting you with content for donor outreach, recommending donors and prospects for outreach, summarizing and highlighting important data on record, brainstorming and recommending fund fundraising strategies through to answering product how to questions and generating lists based on available criteria. So, again, it's important to highlight that we'll continue to explore use cases through research, but these were all strong signals to support our teams in focusing in on these specific areas initially. So we've looked at the use cases that were important to you and how research has informed our approach, you know, why Blackbaud AI chat specifically? So across multiple streams, our research our research highlighted the organizations needed a conversational entry point into AI because social impact professional teams reported being overwhelmed with dashboards and complex tools, difficulty knowing where to start with AI. And the so what here is that AI chat creates a low a lower barrier, intuitive way to ask questions, get help, and explore possibilities. You also discussed with us that you needed a trusted way to navigate donor data. So our customer said, you know, I don't know where to find what I need. I don't have time to dig into data. An AI chat supports this by offering explanations of donor insights, quick pull summaries, ideas on engagement approaches. And finally, you talked to us about the need for human centered communication and messaging support. So based on insight from communication studies and donor life cycle insights, AI chat helps to draft thank you, suggest stewardship ideas, tailor messaging to donor stage, keeping tone mission centered. So the aim of focusing on these outcomes identified in research was to support reducing mental load. So AI chat becomes, you know, the first place to ask for help in RDNXT, faster decision making, so through more efficient, quicker ways to view patterns in your data, and more personalized donor engagement supporting you in ideas to craft messaging. In terms of security, keeping AI use within RDNXT versus use of external tools, so it's that connected system. And in terms of the research journey ahead, our teams will focus on continuous improvement and continue to monitor credibility and trust, and engagement to focus in on those critical human elements of Blackbaud AI chat to support our customers in achieving value and ensure trust in the service. And so far, in early research, some of the sentiment towards Blackbaud AI chat amongst customers has included that the support is appreciated. It's seen as promising for identifying new prospects and cultivating donors. Summaries provided when viewing prospect records are valued and the ability to search across the database was highlighted as a key strength. So everything we built and everything we build is is shaped by what we're identifying that customers need and what will support you in advancing social good, and we'll continue to explore use cases ongoing in research. So thank you very much for your time, and I will now hand over to our principal product manager, Heather McLean. Wonderful. Thank you, Laureen. And you can see that it was this incredible body of research that has really founded the beginnings of AI chat and now really focusing on how we are going to implement that within Raiser's Edge NXT to specifically help fundraisers. So what you're seeing on this slide is how we're thinking about AI and RE NXT really at a system level, not as isolated features but as support across the full life cycle of fundraising and donor management. When we step back and looked at how you actually work in Razor's Edge NXT from prospecting and moves management to stewardship reporting and performance, we started to see a common pattern. The challenge, like Laureen said, wasn't the lack of data or tools. It it was really the cognitive load of pulling everything together, deciding what matters most, and and knowing what to do next. So instead of starting with where can we add AI, we started with where do fundraisers and the admins like you who enable them spend most time interpreting information, stitching context together, or second guessing decisions. That's why our priority use cases span across all aspects of donor management and fundraising from prospecting and pipeline pipeline optimization to engagement, stewardship, reporting, strategy. It's a lot to do. And there are the moments where intelligent assistance can generally help, and and that's what we're trying to focus on for first. And it's important to be clear that this isn't about replacing your expertise. It's about really leaning into it as professional fundraisers and those that enable them. We're grounding these assisted workflows in black bots, best practices, our products such as our predictive models, and our institutional knowledge of the incredible people that work here. Combined with your judgment and your understanding of your donors and your mission, the goal is to reduce friction between insight and action to help you move faster from what am I seeing here, how should I interpret this, to like what do I do next. And whether that's prioritizing outreach, refining a move strategy, or understanding performance without running five different reports, that's where we want to be. So it's important to say that this is a journey. What you're seeing today in the product is not the end state. It's really the foundation, and we're intentionally building toward more and more sophistication over time, laying in deeper insight, more customization, stronger recommendations, and and tighter integration as both the technology evolves and we learn how you are actually using it in the real world. It was impossible to understand how chat was going to be used in all of the various scenarios across all of the organizations that we serve around the world. So our prioritization on what we'll focus on next will continue to be driven by your real usage, the outcomes that we're driving and and real feedback and and not novelty. And and Lindsey Ricks, my colleague, will share a bit more details on how we do this later and your role in that. As new capabilities become viable and we see where assistance is truly earning its place in those workflows that you love, we'll evolve these experiences along alongside you. So over time, what this creates is a more connected experience in Razor's Edge NXT where assistance shows up inside workflows you already trust, respects the way that you work, and helps you focus more of your time on relationships and impact. So that's the direction we're heading, practical, human centered assistance that earns its place in your daily fundraising work. But, of course, none of this work will happen if it's not built on that foundation of trust. And so these capabilities don't just magically appear in the product. They're enabled by deliberate platform level decisions around security, privacy, and responsible AI. So I'm going to turn over to my my product partner from our platform team, Leah, who helps make these experiences possible behind the scenes. Wanted her to come and share and walk you through how we're approaching responsible AI at Blackbaud and how we're ensuring these capabilities show up in your product in ways that are transparent, governed, and aligned with the way you work while still advancing really quickly as technology evolves. So, Leah, over to you. Great. Great. Thank you so much, and hello, everyone. So nice to meet you. I'm Leah Wilson Duff. I'm senior, technical product manager for AI platform. The easiest way to think about, sort of what AI platform is is if the rest of my wonderful colleagues you'll be hearing from today are talking about the body and colors and seat and trim level of the car, my team, the AI platform team, are responsible for building the engine and all of the stuff that that the wonderful, awesome, like, utilities and everything that make up the rest of the car, are sort of powered by. So my team is making sort of decisions that are foundational to all of the ways that my other, product colleagues are creating the features and capabilities that y'all are gonna be using to help improve all your day to day work. So I wanted to talk about the how we sort of start from a foundational level of trust and privacy. Trust and privacy and and decision making are core to how we approach building that engine internally. First and foremost, as I'm sure most of you are very likely aware, we are a an opt in organization. That was really critical to us because at Blackbaud, what we wanted to make sure is that organizations are able to decide their own comfort level with generative AI features. At the January, we launched in the sort of global opt in. What we realized is that organizations had differing levels of of comfort across the board with generative AI use sort of in general and then maybe specific to their organization. And that's one of the things that we really keyed in on and and realized was gonna be really fundamental to how we approached all of our sort of engine building is we wanted to make sure that it was opt in by design. And so as an organization, you have to choose to opt in to these generative AI features. And part of the way that we arrive at those decisions is we have internally something that we call the responsible AI council. We make sure they all have, like, cool, matching, outfits and everything like that, and we show up. And and, really, what that's all about is gathering all of the brightest minds at Blackbaud in security, in legal, in data ethics, in product, in engineering. And all of those people on the cutting edge of the understanding of technology in general and generative AI come together to basically have a council of talking about how do we make these decisions and how do we make them in really, like, transparent, ethical, and understandable ways that we can communicate out to all of our customers. And from the very beginning, privacy is our foundation. When building the engine, we make sure that we have a mind on how privacy works. That means that we are redacting and cleaning data. We have safeguards built into the fundamental nature of all of these things. They're at the forefront of how we even built out all of the capabilities that our product teams are using to bring generative AI features to all of y'all. So what does that look like? One of the fundamental ways that, that we approach, sort of privacy and security in the LLM space is the technology that we are leveraging. So think of the engine powering the engine, so to speak, includes architecture level configurations. Now what does that mean? It's essentially we the concrete that we are using to build the foundation of our LLM engine, of our product platform for AI, have built in configurations and safeguards for things like hate, things like bias, things like all of those specific things that could end up having, like, really sort of negative outcomes insofar as interfacing with LLMs. We make sure there are safeguards at the highest possible level in the sort of engineering architecture. The and in addition to that, we continuously are monitoring and evaluating all aspects of LLM performance. So in this space, there are things like quality that we measure over time, the amount of hallucinations that are happening, the all of the the the technical details, like context that's provided to the LLM and the way that it then uses that to provide quality outputs. And we also have user feedback, the thumbs up or thumbs down, for instance, on the AI chat itself. All of that is brought together to be continuously monitoring and continuously improving. We are using enterprise grade models. The to stay on, like, the cutting edge of how LLM technology evolves. And staying on the cutting edge of how the technology functions means we get to stay on the cutting edge of of security and privacy standards in the LLM space as well. And lastly, like Heather and Loreen and Catherine have been talking about, is we really are focusing on keeping people at the forefront of how we are building all of these things. People. Right? And to that to that end, accessibility is a mindset. We realized that in building, the initial stages of, of AI chat that there were some accessibility standards that we wanted to hit. So the AI chat meets our SkyUX accessibility standards as well as the, WCAG standards, which is sort of the industry leading accessibility standard for things like screen readers, for instance. And what we wanted to make sure is that at a platform level, the tools that my team bring to the rest of the organization to build the tools for y'all to use, that we were making these decisions from the very beginning, thinking about user needs and people, human beings, both internal and external. So as we collaborate, I just wanted to let you all know we are thinking about privacy, data, accessibility, and really building a really, really great platform for all of my internal colleagues to use to bring to all of y'all external colleagues. So thank you all for listening. Hopefully, that wasn't too nerdy, too techy, but I just really wanted to give everyone an insight into what I'm thinking about building the engine here at Blackbaud. So I'm gonna hand it off next, to, my colleague Lindsay to talk about getting the most, from Blackbaud AI chat. Hi, everyone. I'm Lindsay Ricks, one of the technical writers working behind the scenes on chat for Blackbaud AI. In my role, I also create a lot of the supporting content that helps you use chat with confidence. I know many of you are power users of Razer's Edge NXT, and you care deeply about accuracy, clarity, and trust in the tools you rely on. So today, I wanna walk you through a few resources that are designed specifically to help you get more out of chat and to make sure your voice directly shapes where we take it next. Let me start with something you can use right away, the chat with Blackbaud AI quick start guide. If you haven't seen it yet, it's a short practical walkthrough of how to work effectively with chat, the kinds of questions it handles well, ways to refine your requests, and examples you can adapt for your day to day work. The goal is to help you save time. Instead of trying to reverse engineer how chat thinks, this guide gives you a clear path so you can get high quality answers faster. We're also working on something that reflects a significant investment of time and effort, a chat for Blackbaud AI prompt library designed with your workflows and needs at the forefront. This resource is currently in progress and will be available soon. And just to level set, an AI prompt library is simply a collection of well crafted example prompts you can use to get high quality results from chat. Instead of having to figure out the perfect way to phrase a question on your own, the library gives you ready made starting points that help you can use that you can use as is or to adjust to fit your work. Now the reason we're creating this library is pretty simple. Prompts are powerful. The way you ask a question can completely change the quality of the answer you get back. We don't want anyone using Razor's Edge NXT to feel like they're staring at a blank screen trying to guess what chat needs. And this isn't going to be a giant list of generic prompts. That's not helpful. It's designed to spark ideas and show you new ways chat can support your everyday work. Everything from cleanup tasks, audits, reporting to more creative or strategic thinking, when you're planning ahead or solving complex problems. These examples give you a clearer picture of what's possible. They're practical high quality prompts that can help you go from can chat help with this to yes, and here's exactly how to ask for it. So really, the library is meant to reduce guesswork, expand what you can do with chat, and give you a faster path to better outcomes. You probably noticed that as you move around Raiser's Edge, you'll start to see helpful suggestions you can select in the chat. These aren't random. They're designed to help you quickly ask the kinds of questions that make sense for where you are in the product. And once you're in the chat, some responses will include links so you can see exactly how chat for Blackbaud AI arrived at an answer. The way that way, you can verify the information, confirm the source, and feel confident that the response is accurate. All of this is there to help you ask better questions, get clear answers, and trust the results you're seeing. And the last thing I want to highlight, and honestly, is the most important, is the feedback option right inside the chat. When you send feedback from inside the conversation, we get the context. We can see what you asked, what chat gave you back, and what you were trying to accomplish. That's gold. It helps us pinpoint where responses need to be sharper, where the product needs to be smarter, and where the experience should be more intuitive. And just to be clear, any personally identifiable information, things like donor names, contact details, or anything sensitive, is automatically redacted before it reaches us. So we get the insight we need without ever seeing your organization's private data. So please keep it coming. The good, the bad, the confusing, every bit of feedback you give us inside chat truly helps us to improve it for you and for every other user who relies on it. So that's a quick look at the tools available now, what's coming soon, and how we're working to support you better. At the end of the day, we want chat to feel like a partner, something that helps you move faster and frees you up to focus on the work only you can do. Next, I'm gonna hand it to Carrie about where we're going next. Great. Thanks, Lindsey. For this portion of the presentation, I'm actually gonna stop sharing slides and jump into the product itself. So give me a minute while I update the screen sharing. Here we go. So what I am showing is still in a test environment, in a development environment. It's something that has not been released just yet. It is an example that lives within Prospect Insights Pro today, but I wanted to share it as an example of the direction we're moving in. So although this specific example is within Prospect Insights Pro, you will start to see similar functionality, similar patterns in other areas of Razor's Edge NXT. So what I'm showing here is the undergiving queue, which highlights individuals who have high predictive the model suggests that they have a high likelihood to give a major gift to your organization. They're assigned to a fundraiser. They're in portfolio. But when we look at their previous years of giving, they all fall below that predicted target gift range. And right now, we have the recommendation of make sure you have an opportunity in place for this individual. With this new feature, we would take that a step further. And when you go to add an opportunity, we'll use the models from ProspectInsights Pro along with, some of your environment settings, such as fiscal year settings, to prefill some of these fields for you. Here, we have a little, eye icon, which gives you more information about how we're pre filling these fields and why. The ask amount is based on that target gift range as well as the expected amount. Again, that deadline is going to be based on your fiscal year settings, and all of this information is still editable. So if you go in and you decide we actually want to target a different ask amount or put a different plan in place, you do still have that flexibility. So I'm gonna go ahead and just fill out some of these required fields for the purpose of this demo. Okay. And then from here, once you click save, you'll see this new success message appear. And it asks if you'd like to create an action plan for this opportunity. If you click on that link and if you have enabled the generative AI functionality, so going back to what Leo was discussing, then it will start to create that action plan, and it's calling endpoints for data related to that prospect, to the opportunity, related to best practices for fundraising. So it's pulling together all of this information, and then that data is, used by the large language model, the LLM, to generate a series of suggested actions for outreach for this individual. So you'll see here, we now have, initial outreach with a date in the future. Part of this is intended to be proactive. Can we put a plan in place to help you get to that ask amount for that opportunity? The talking points, again, are going to be personalized and tailored for that individual. So here, it's referencing a recent gift of $25,000 to the capital campaign, which we can see reflected here on the prospect in sites pro tab. And then a series of additional actions, ultimately, to help with that cultivation to get you to that final asked amount. Now, because all of this is delivered within the chat experience, you can also go back and forth to refine the plan. You might say, hey, I have an event coming up that I would like to include in this series of actions. I wanna make sure I invite them to that event. Or maybe you say, you know what? I want to have these actions in place for the future, but I don't wanna come up with the talking points until we get a little bit closer to that ask amount. I want to fill in those details on my own later. But after you refine and you go back and forth, it asks if you would like to add these steps as actions. And if you say yes, it will do so. So now it's pulling together all that information, along with the refinement that you did to create actions. And each of those actions will then be connected to that specific opportunity. Within these, you can also click to review any of the individual ones. Maybe there are specific edits you wanna make just here in this interface, and any of those suggested talking points or notes would be included in this action as well. You can click a button to add all of those actions, and you'll see I do now have notifications that those actions were assigned to me. I can click view opportunity, which will pull up that opportunity I just created. And you can see here all these actions, that were created as part of that plan. So, again, the intent behind this feature is to help get to that ask amount for that specific opportunity and to take a more proactive approach to make sure you're putting those actions in place ahead of time. And you have them available to you then in the work center where you can review upcoming actions. Like I mentioned, right now, this feature is only within Prospect Insights Pro, and we haven't released it yet. We're going to start releasing in waves shortly. But still, if you're adding an opportunity from other pages, it wouldn't be this experience just yet. So because we're focused on major giving use case within Prospect Insights Pro, we have the predictive models here within the experience. You would only see this when adding opportunities, from any of the queues within that Prospect Insights Pro experience. So I have one more thing I do want to share, but it's a little further out, and so I need to go back to the slides. Here we go. So you thinking through that workflow. You've gone through the process. Maybe you reviewed several undergiving prospects, and you've, added those opportunities and those series of actions. But a lot of them are maybe six months or even a year into the future. What we're working on next would be an update where you can prepare for upcoming actions. So, for example, as a fundraiser, maybe I come in here and I can see what are all the actions that are assigned to me for a given time period. Maybe I'm coming in to prepare for my day or for my current week. Now, again, what I demoed included suggesting talking points and content that's personalized for that prospect, but it was all based on today, what the information is at that given time. But six months from now, maybe something has changed about the prospect. There's new context that you want to consider. So from this view, you would be able to click a button to regenerate context. Again, it would pull in all of the latest information to make sure that when you're reaching out, you have the most up to date information possible. You can also come here to complete an action and add in your notes after, having that conversation with a potential prospect. One thing I do wanna call out again about this whole experience is the transparency aspect. So you'll see AI called out whenever generative AI is used. And, again, going back to what Leah was sharing, if you have not enabled the generative AI functionality, then you wouldn't see these features at all. The other direction we're moving in with this example is taking some of those best practices that Lindsay was sharing. So when you click a button to create the action plan, behind the scenes, we are adding in all that information into the prompt, that we know will be most successful for creating that series of actions for you. And then that's what I had to cover for today. Catherine, yeah, I think you may be up next. I am going to come back on stage and respond to a couple of the questions that we saw in the Q and A that came up over and over again. And if we have time, I'm gonna call up some of our different presenters onto the stage to respond to questions. First off, all of the functionality that we were talking about or most of the functionality we were talking about until Carrie started speaking is being developed for all customers of Rasers Edge NXT globally. So there are some responses that you will get that are different based on the software or the modules, the capabilities that you have in your system. So if you have PI Pro, Prospect Insights Pro, you will see different responses possibly than if you have Prospect Insights. But in general, all of these capabilities, the AIHI capabilities before Carrie started speaking are going to all customers and are available now. If you have questions, you're not seeing it in Lindsay's quick start guide that is linked through the docs menu. It goes through troubleshooting steps. So there is a generative AI or global AI opt in that has to be accepted and that has to be done by a certain level of admin at your account. And I am gonna now call Heather up on the stage to respond to a couple questions that she had called out. And just hit the go on stage. There we go. Alright. Was the question the the question about hallucinations? Is that the one you. wanted me to cover? Alright. And and feel free if if Leah or anybody else has anything to add to come up and and layer on what what I'm sharing. I don't wanna get super technical in the weeds, but there's a lot of ways that we try to prevent hallucinations. Like, as we started with, it's like we also just start with, like, is AI the right tool here? If the answer is, like, very deterministic and there is only one right answer, Like, sometimes we don't even launder that through generative AI. You might experience that answer in chat, but it it's so deterministic that we've just called to retrieve that data, and you're actually just receiving it in chat. So first, it's like, is AI the right tool for the job? Then, like, we only want to, like, provide the right context to the LLM to to make recommendations only what they need, right, to be able to do that because there's things like recency bias and there's things like giving it too much data to do the task. We want to be sort of really mindful about what data we serve up to reduce the risk of hallucinating because of the amount of data. There's also things just in places like safety prompts and system prompts where you give very stern instructions that they're not to make up anything that's not provided to them. And then we have evaluations sort of across the life cycle of our development and sort of predevelopment before you even see a feature where we're running prompts and evaluations for things like hallucinations at a really, really, really high scale to, like, understand where there could be issues that we need to optimize before it's even ready for you to see. And then once it's something's in production, we're still constantly monitoring through some of the tools that Leah talked about to make sure that things are remaining stable and safe and not hallucinating. So while it it can't be perfect, and that's why you see the message in in chat that says, like, these things are not perfect. Please check them. We do always try to provide a way for the end user to do that. So that's why you'll notice, like, when you ask a data question, we don't just respond with a number and say, well, how much did I raise last year? Oh, 1,200,000.0. That's how much you raised. We know that that's nuanced, right, and there's a lot that goes into answering the question. And that's why we always have leaned into lists to say like these are the gifts that match the criteria to answer your question, or these are the constituents, or these are the actions, or these are the opportunities, and where possible linking you to that underlying dataset so that you can look at that list and see the actual filters were applied to provide the answer. Also, something we've noticed is in the logs is perceiving that there is a hallucination because it was not super clear that the context of the conversation in chat has actually changed. So you may still think you're asking a question about a previous constituent, but the context for the chat has actually changed. And so some of that is back end work to to make that context switching cleaner, and and some of that is looking at how we can improve the user interface to make it clear to use context of which record you're in. So definitely not perfect, but about an area for improvement. I don't know if anybody else has anything to add, but those it's a it's a loaded question with lots of things that are helping to reduce that because we know like, I hope it's evident through everything that we've talked about is building your trust in using these tools is is paramount. So constant improvement in these areas, hopefully you'll see. And I wanna point out another question that has come up is people wanna know more about data privacy and security when they're using Blackbaud AI or AI chat. And again, pointing out that there is a link in the docs tab to the trust center. Look for the AI questions. And there are about 21 different questions and responses from our security and our product team on responses to those questions. Yes. That is the best source of truth to get those answers. And as things evolve through our trust committee and some of those groups that Leah talked about that are working behind the scenes, they keep that trust center updated with the latest and greatest news. Any other questions that any of the panelists wanted to call out? Alright. Hopefully, we've we've done our oh, Leah. Good. Oh, Someone I I just came on the because I saw I I answered one of the questions specifically in the q and a, but I I thought I would just say to everyone since I saw it sort of pop up either as a specific question or maybe as a subset of buried in another question. But just to just to make clear, I know there is lots of the LLM space is generative AI and LLM, which is large language models. That's the underlying technology that makes up sort of what what is referred to as AI. The part of the one of the things that people hear with regards to AI is things like training models. The Blackbaud is not training models. The we are using enterprise grade models. By that, I mean, OpenAI, which they're the chat GPT company, and Anthropic, they're the cloud company. And Google also are, like, the big industry leaders in models. The models are to go back to my engine analogy, if if we are building the energy or if we're building the the engine as AI platform within Blackbaud, the models are like the spark plugs. They're what actually make the thing, like, fire up and be able to respond as an LLM. And we are not training models at Blackbaud. Training requires lots of very, very, very, very precise and sophisticated sort of setups and configurations and is an industry in its own right. And not that we are incapable of doing so, but in partnering with enterprise partners like OpenAI and Anthropic, we have the ability to leverage, like, the scale that they can bring to that with regards to things like safety and and and security and privacy. So we are not training models. We are not training models at Blackbaud or using information to train models. We have enterprise partners who are sort of providing those models, and we're using those. So I just wanted to to say that. If anyone is talking to their internal nerdy folks about LLMs, Blackbaud is not trained models. We are using sort of off the shelf models. And, Heather, can you respond to the question about whether or not Blackbaud's AI chat in RENXT saves the user data somewhere after the transact the the interaction. Yeah. And Leah can can stay up here as well because, again, there's there's. a partnership there. So keeping an audit trail of the logs is important for a multitude of reasons. By the time that someone like me as a product leader is analyzing logs and feedback, everything is redacted. So I'm not seeing any of the PII from your system. We're not we don't have access to any of that at the at that level. So we can speak to, like, how that's protected. But, like, when. you're giving us feedback or when you're, providing all of this, like, all we're seeing is that you're searching for a constituent, and we. don't see the constituent's name. We just see there is a constituent name there. So there's lots of AI behind the scenes that's making that all happen impossible so that we're able to still respond to your feedback responsibly without needing to see the underlying log data. Yeah. Yeah. So the and and I I see I see a question or a a chat. It is the these these spaces and especially emerging technologies and continuously, like, continuously evolving technologies, it it tends sometimes to be a challenge to not immediately go, like, very, very deep on a technical level to talk about these things. So, unfortunately, to address I have to get I have to get a little techy to address, like, the questions about the data that's shared. So I saw yeah. I see someone, James, over here in the chat saying, if if we're building the engine, is it using open open AI and Anthropic as the model? So, fundamentally, what we what we have is the AI platform, and any organization that's leveraging LLM tools has a centralized platform like we have here that is running and the apologies for for everyone for going at this level of detail, what's called a proxy. So what we do, to Heather's point, is by the time the information gets to the anyone inside the company, all the all of the any personally identifying information or anything that is identifiable or proprietary information is fully redacted because the engine that we've built uses the the model technology, which is the underlying sort of foundation of the file of us operate is a predictive model that grabs the next most likely token to then generate a response. And all of the information that is going to those third parties runs through our proxy. So the information is exchanged once, not stored, and then comes back to us to interpret to provide a response. So the in in sort of the larger questions, I I saw someone had said that not in question whether we are training models. Some people do have the question that we are training models and just wanted to make sure that those peoples that people knew the specifics. The data exposed to OpenAI and Anthropic is exposed for the the generating of response that our proxy gets. We are using the technology, not like they're they're, there's no data being stored. We are keeping it, in our sort of proprietary systems where everything is redacted. So that's sort of the approach that we are taking is no one inside the company is seeing any sort of actually identifiable data. Because as Heather said, by the time it gets to any other internal user, it's all like, anything identifiable is is redacted, and that's done automatically. It's done as a rule, as a process when things hit our proxy. And the the technology underlying that proxy is, like, one of the core things that we built out as an AI platform team to make sure that everything is on, like, a really, really locked in from a security and privacy perspective. Amazing. See? It's a part it is truly a partnership between the platform team and then us being able to use all of those incredible capabilities to then translate that into driving hopeful outcomes and. value for you. So it is it takes a village including you guys giving us all of that wealth of feedback to help this be a learning system and totally improve over time. Can I summarize it in a way even a marketer can understand? It sounds like you are saying we don't use people's data to train our customers' data. We don't use customers' data to train the LLM. Their information never goes to Claude in an unredacted way. It never goes to Claude or is that that short story? the, yes. So the, the we are we are using we are using underlying technology, so that we don't need to reinvent the wheel. And, again, as I said, the these enterprise models are come off the shelf sort of baked in with a high degree of sort of security and and trust. And we leverage that technology in our own technology. in order to make sure that things are redacted, PII is redacted, and no information is exposed to these third parties. Donor data is not going to Microsoft. It's not going to Anthropic. It's just Yeah. reiterate because the questions keep coming. Sure. Yeah. Yeah. Yeah. Yeah. I think I think we're gonna have. to get, like, way more technical. if we go any deeper than that. So just know that there has been zero changes to your your privacy policy with Blackbaud as a result of introducing things like chat. So you can go to your service agreements with Blackbaud and and those things remain intact. In fact, there's been an addendum to it that includes our mutual responsibilities when it comes to AI. So hopefully that helped to answer the questions at a level that didn't break too many folks' brains on this call. And we haven't had time to get to the many, many, many questions that were in the Q and A. We're going to collect these questions after the session shuts down, and then I will work to get a response posted to the community to the broad groups of questions. It may not come out. It's not going to come out this week. It might come out next week. But just to round up the general nature of the questions and get responses out there, because we're all very well aware that there are a lot of Q and A questions and a lot of answers get pushed out and sometimes it's kind of hard to follow the Q and A. And the next I'm going push this slide out. The next session in the inside track to the unified view is going to be the second Tuesday in March. That is going to cover query and reimagined export. And so we're going to go back to the the web view development, the the kind of features that a lot of you were asking for off and on throughout this presentation. So I hope you'll join us for that and the the product team. Thank you very much for joining us today. We really appreciate it. Hope we were able to respond to some of your questions and open up some new ones. We also hope that you will explore this new technology, the AI chat that we have developed as the beginning, just the beginning, to a new way to interact with your database. So thank you very much for joining us today.