 Well, hello, everyone. Welcome to Executive Directors Chat. Today, our topic is working with the world-class AI, which we know is artificial intelligence grant writing assistant. So I know many of you have come to our ED chats, and we're always on the Zoom platform where we can see your face. We're going to do it a little different today, because I believe that there'll be a lot more interaction. And I think this is going to be a better quality webinar. This is webinar style instead of the Zoom meeting style. So I want to show you how you can engage today. But those of you, I know someone said this is the first time here. This is your first time here. We would love for you to use the Q&A feature. That's question to answer feature to ask your question for our speaker. I know many of you are going to put your questions in the chat. I'm sure we'll be able to grab those. We will email you the replay today. He's not going to show any slides. He's going to do a demo straight from the website. So you'll get the video replay within four to eight hours, probably tomorrow. If you need the closed caption, go ahead and type on that CC button at the bottom of your Zoom screen. And if you learned something cool today, which I'm sure you will, would you share it on your social media and make sure you hashtag TechSoup. So I'm going to get ready to move out of the way. My name is Arita Simon. It's your first time here on the webinar producer here at TechSoup. But today our speaker is Phillip with one L. Dang. He is here. He's the CEO and co-founder of Granible.co. Now I want to mention something that Phillip shared with me because it enlightened me. He is not from California out to my valley where they create apps and all these things. He has been working in the nonprofit sector and he built the prototype for Granible about 15 years, right? About 15 years ago. And so this is going to be an amazing webinar. I do want to mention if you want to ask Phillip a question live, use the raise your hand option and I will bring you on. I will unmute you so you can ask him a question live. I'm going to move out of the way and allow Phillip to take over. Welcome, Phillip. Thanks so much, Arita. And hello everybody from seemingly everywhere. It's very cool to be with you all. Yeah, so to start out, I just wanted to share that this feels like a really familiar space for me because as Arita mentioned earlier, I've worked in the nonprofit sector for most of my entire career. So that spans about 15 years. I didn't come up with the prototype for Granible 15 years ago. I was a teacher in the Marshall Islands, if anybody knows where that is, I worked with grant seeking organizations for that entire time. I've been an executive director myself. I've run grant seeking for a large land trust. And also I've been a professional grants consultant as well. So throughout that process, I picked up lots of tricks like I'm sure many of you have and. Experienced how the system is, is broken, how it's important, how it's inequitable. And all of these conclusions led me to start playing around with building software in 2020. I really leaned into it when I first learned about GPT three, which a lot of folks don't know actually was released in late 2020. It didn't generate much attention then, but I thought it was quite a profound shift. And I felt like it was only going to get better. I started trying to teach myself to build software, which was pretty tough going until I met my co-founder here in Richmond, Virginia, where I live. My co-founder's name is Robert, and he is a stellar software engineer. So we put his technical ability together with my understanding of the grants ecosystem. And that's how we build Granible. And the idea is that as many of you probably know, grant seeking is a fairly well-structured, sometimes it's repetitive, very derivative process, where a lot of our time goes towards finding bits and pieces and then reformatting to suit different funders. So what we saw was an opportunity to make that a lot easier using generative AI technology. So today, what I'm hoping to do is rely on Aretha's help to make sure that your questions are coming up and that we can talk about everything at the intersection of philanthropic work and artificial intelligence. In a few minutes here, I'll do a demo of our software, but I really, I'm not coming into this thinking about it like a sales pitch at all. I hope if you work on grants that Granible could be useful, but I do some of my best work through my newsletter, which I write every week, which is about what I'm seeing at the intersection of technology and philanthropic work. So 15 years doing purpose-driven, mission-driven work and now going on three years of being a technologist, I see this, I believe I'm a storyteller by training and by nature. And one of the things I'm trying to do is help describe what I'm seeing as trends in technology and what it means for the nonprofit sector, why it's important. And so a lot of times people use conversations like this to ask me about different types of AI, what are the trends, who are the players, techniques, risks, benefits, and we can speculate about the future together as well. So I really do encourage people to add some Q&A and give me some feedback as to how I'm doing so that we can make this discussion as useful as possible for all of you. So Aretha, are you seeing anything that we should flag right now? No, not right now, doing great. Just people coming in, still coming in. So if you'd like to proceed with the demo, I'm sure there'll be questions then. Okay, cool. So I'll share my screen now. And can you all see, sorry, I'm going to do the whole thing where I got to move all my Zoom boxes and chats around. So I'm going to close my chat. So Aretha, can I rely on you to just flag any questions for me? Absolutely. And are you seeing grantable? Yes. Okay, so this is what a new grantable account would look like. So it has no folders or files. And so what you would do if you were starting up is you would start by uploading some previous grant proposals. So these two are fictionalized proposals that I've created. They're based on real grants, but I've just changed all the information. So you just upload them as you would, and there's nothing else that needs to get done here. Now the content from your previous proposals is here. You can look at it and read it like you normally would. You can see organizational overview and so on. When you're ready to work on a new proposal, there are two ways you can go about that. You can upload the file just the same as before. So this is a real application from the Seattle People's Fund. So you can do that and the document will appear once it's uploaded. Just like you normally wouldn't. You can go about answering the questions, you know, by, by hand, you can say innovate NOVA, which is the fictional thing. Or you can actually go to a question like here. Where it asks for the overview and you highlight the part that you want help to address. You do command D, which turns the question into a block and that's what gives you your AI superpowers. So just to demonstrate what it's capable of, you can just go ahead and click generate. And it will start to draft an answer based on the proposals you uploaded before. So that's just one way that you can go about it. But once it's done, I'll show you how you can actually have more control over what it's writing and fine tune the outputs. So it's being quite thorough here. And as you see, when you make blocks, actually, it populates a nice little table of contents here, which is a nice way to navigate through long documents, especially. Okay. So what happened was Grantable looked through those two documents that I uploaded earlier, found snippets that felt it felt were useful for describing for answering the question. It grabbed those sources and put those here. So you can see these are the sources that it used. And then it wrote the draft based on this content. So one of the things we get asked a lot is how is chat GPT different from Grantable? And what we're doing is using your content and the GPT writing ability. So the way I describe it is it's a bit like chat GPT is a restaurant where if you go out to eat, you order a dish, it'll be prepared with the restaurant's ingredients. Whereas Grantable is a bit like a home chef that comes to your home and cooks the food using your ingredients. So that's why these numbers are from the documents that you've uploaded. It should have your stories. It should sound like you. You can also give it instructions. So you could say something like limit to 200 characters. I know a lot of people have quite a bit of experience trying to get things to line up with character limits and length limits. So we'll see how this does. So it's a little over 300. But what you can do is then kind of pare down. Sometimes, you know, we have, this was actually a pretty large question to get down to 200 characters. But hopefully you can see the example there. Another thing you can do is you can highlight sections and then click this revision button. You can have it, for example, make it shorter. And it will give you a new version of whatever the selection is. You could just do one paragraph, one sentence. You can give all kinds of different instructions here. You could say, make it funnier, make it more serious, whatever. And it will keep tracking all your options until, oops, I can't get past this. Help button. Let me change my screen for a second. Oh, that's actually an interesting, let's just do another one. So once it comes up with a shorter version that you like, you can insert it. And you can see that it's replaced the original one with the shortened version. The other way you can start is you can just do a new file. So that's a blank page. A lot of times we copy and paste RFPs from other sources. So let me just, I think I have one bookmarked here. So you can just, you know, this is a fellowship application that I worked on a while ago, but you can just imagine that if that's your RFP, you can also just paste those questions in. And then again, you just create the block. I would recommend actually searching for your content. So you can, let's see, could do, I guess problem doesn't hold review. There you go. So you can go to here and manually select your source. So what I did there was highlight, control C, like copy, and it grabs that source. And then it will send its all say limit to 100 characters or let's say words. And then you can go compose. So that's pretty cool, right? So you took the overview, gave it a character limit and then have it compose. And then once it's done, I mean, a lot of people ask like, Oh, does this mean grant writers are irrelevant? No, not at all. Once, once you're here, it really is just a regular writing experience. Again, deleting, adding fine tuning, however you like. One of the things that I will just close this demo with is to say, when you're done, you of course can export the work to other formats like doc X, or you can just copy and paste into the forms. But once you mark the block complete, it kind of cleans it up. It gets rid of the tools, but it also adds it to your blocks here. And what that does is it really levels up the AI's ability. The next time you encounter a similar question to grab exactly the right pieces to help you draft answers. So as you go, it gets better and better and your or your information stays fresh and it stays in one place. I know a lot of people have, you know, worked on the master document, the boilerplate document, the, the spreadsheet where you put all the copy paste cells, but it's very hard to keep that stuff fresh or find it after a while. So one of the really helpful things is that grantable allows you to search across all of your work and it keeps your blocks in particular stored like this so that you can, you can tag them, organize everything and then reuse it really seamlessly. For those of you that work for multiple different organizations, if there are any of you out there, you can also easily switch between workspaces just by toggling like that. So I think that covers most stuff. I do see some questions that are popping up. Aretha, do you want to direct me to some for. Absolutely lots of questions. I'm going to go with the ones that came in the chat and then we're going to go to the Q&A because I didn't want to miss the ones in the chat. Tracy asks, are there, are the application fields auto populating? So many of the grant applications are on portals. That grantable does not auto populate to the fields yet. That's actually a pretty challenging technical problem. As all of you probably know, there are 8000 million grant portals out there and they all have different, you know, applications that they delineate their forms on the back end. So we can't always be sure what's what. And the worst thing would be to have the wrong answer paced into the wrong section and get cut off. So we haven't tackled that yet. Awesome. So Wanda, you had a great question, but I want to save yours to the end because I think it might require a demo. And Shayna, I hope I'm pronouncing your name right. I'm a social media community. And will it generate content based on that? So you can upload any kind of writing. Notice of funding. That would be from the grantor. It was like an RFP. Yeah. I mean, I think what you would be doing is within the RFP, if this were a notice of fun, it's, it's any text. it's any text. And then when you come to a question that is requiring a narrative answer, just highlight that, turn it into a block, and then Grantable can help you address the question. I hope that's helpful. Awesome. Deborah said, is there a link to use for AI? So a link to Grantable. Awesome. And I put that in the chat room as well. And Harmony asks, it's Grantable and open source and free for all. The same question, someone else asked in the chat too. It is free to start and you can use it without signing up for a subscription. Basically, when you create an account, you'll be given an amount of AI usage for free. And that doesn't expire, it's just there for when you want to use it. Once you use that up to play around or write a grant, it will require you to either buy more usage or subscribe. You can buy kind of the pay as you go plan. Or if you subscribe, you get a better deal on your AI usage tokens. And the reason that is is because when you send one of these requests, there's actually a huge amount of computation that happens on open AI side through the GPT language model. So the information that you're sending, the writing, the question, all these different components that we've put together, that all goes in, it gets processed, and then it comes back out with the output. So all of that actually is quite intensive. And so that's how we've designed our pricing structure. But even the paid plans, the first level is $24 a month, the power level is $89 a month. So from my experience, we've tried to make this really affordable, even at different size nonprofit budgets. When I ran a nonprofit, it was mostly just me for five years. So $500 a month software was just out of the question. But things like Slack, there's sort of office software that costs in the 20s, the 30s, the 40s and really adds value. That's what we're trying to build for the nonprofit sector. Awesome. And just to bring a reminder to those who just came in and asked us, it's being recorded, this is being recorded, and we're going to send you the video replay on tomorrow. Kayla asks, is grantable? Is it software or is it AI? Well, this is a good segue into sort of the more technical... AI, the best description that I've heard comes from a guy called Demis Asabis, who leads Google's DeepMind project, which is probably one of the preeminent AI laboratories in the world. He described it very simply as the pursuit of technology and software that is smart like human beings. So AI is a kind of software and grantable is software that incorporates elements of AI. So everything you see on the screen, all the buttons you can click, being able to type and navigate menus, all of that is software. The way we store your data on a separate secure database, that is software engineering. And then open AI has the GPT large language models. That's what gives it the writing ability and the comprehension ability. So we're sending information and grantable is communicating with open AI to do what you see on the screen. Awesome. Great answer. Kelly asked or actually made a comment, we've never done a grant application. Any suggestions for newbies? Yeah, hang tight and believe in yourself. Work with an awesome grant expert in your community. Our next phase is trying to create some intelligent tools to help with that first grant. How do you diagnose the work ahead of you? How do you lay it out systematically, identify all the little checkpoints? I applied for a grant once and there was the clear deadline. And then there was a criteria that said you have to submit this other like letter of intent by this previous deadline or you won't be considered. But it was a really small thing. So we put together the whole proposal, got there and they said, oh, did you submit this other? And I was like, oh my gosh, no. We had to really negotiate with the funder to please let us submit the full application. So it's really important to look through and pull out those little details. It's important to answer questions thoroughly and concisely while being compelling and emotional in the right places. That's what gifted grant professionals can help you do. So I don't think we're going to be replacing all of that contextual knowledge anytime soon. But as far as sort of laying out the tasks systematically, making sure that you're answering all the points within a question, it can definitely help. The last thing I'll just say is there really is that sort of good info in, good info out, bad info in, bad data out, dynamic here. It can, it will be as good as the writing that you put into it, the investment that you put into it. So if you have an annual report or other materials where you've written extensively about the work of your organization, you can upload that. Anything you as a person would use and say, this is a great paragraph describing our work, I would use that to write a grant application. Definitely you can add that and give that to your, the AI assistant as the template material to help you answer these questions. I do think it's easier to write in grantable, especially if you've never written a grant before. What I would recommend though is still to try to seek out an experienced grant professional to give you their perspective, even on the drafts that you're creating to see how they stack up to what is industry standard. But believe in yourself, a single no doesn't mean anything. So it definitely takes persistence, build those relationships with the funders. A lot of times I've seen that a first application is rejected, the applicant gets good feedback from the funder. That in and of itself establishes the relationship and the next cycle they apply again and they're awarded. So there's a lot of relational work that happens outside of the program. Yeah, I've seen that a lot too. I'm going to go back to Juana's question. I said this may be time for a demo with her question, but she said, can you talk about the workflow for grant seeking and the grant writing process? How can AI and or automation help even just searching for grants as cumbersome? So using grantable. So don't use grantable to search for grants. We haven't tried to tackle that problem. It is a thorny one as anybody who's gone after grants knows. There are so many thousands of different websites and databases and places where grant opportunities are announced. There are several companies that focus on aggregating all of those opportunities and trying to make that searchable for you. Candid or Foundation Directory was the original player there. There are lots of folks grant station, instrumental, a bunch of folks have come up and more are on the way. I'm sure all of those companies are probably working on generative AI applications for their technology, although it's not ideally suited to what they're trying to do. They're trying to understand your organization and match you with funding you might qualify for. So it's not clear to me exactly how AI will impact that yet. As for putting the grants together, grantable is targeting that. We are trying to get your information in one place to make it really searchable and to store it as these blocks so that AI can use exactly the right pieces to help you draft. We're working on some of those process management or project management features where in this left side, we're working on some like a to-do list kind of experience where AI will help to extract all of the to-do items in a proposal to make sure you've gone systematically and done all of them and checked them off. So we can help organize and accelerate that actual assembly process, the writing process. Submission is kind of where it leaves our hands again and there's numerous grant management systems out there. Cyber grants, submitable, foundant, flux to name just a few that you may be familiar with or even Google forms. I mean, mail chimp, I think, also has a way to or survey monkey. Sorry, I knew it was an animal. So there's all kinds of ways that grantors take in their grants. So I think at some point, we'll try to also help streamline that with ways to make submission easier. But I'm also just wanting to draw a clear line there that that's where we sort of hand the ball off again to other softwares, which I don't love. It's a really clunky part of the grant seeking process, but the part at least in the middle where we can help you create a proposal that's really high quality and have that proposal stay in your knowledge base, that can at least make the everything after us more efficient. But that's what we're doing for now. Okay, so we're going to do some speed questions here. So I would love for everyone to go to the Q&A and if there's a question that you would like answered, you said, you know, I want to know the answers to that question. Once you go ahead and give it a thumbs up, that'll let me prioritize a question that needs to be asked first because we have 24 and counting in the Q&A field. So here we go. I'm going to go with Allison. She says, does Granable have a stronger security feature than chat GPT? Chat GPT isn't meant for sensitive and personal data. So that's not as much an issue of security as in, you know, the data within open AI systems is actually stored on Microsoft cloud servers. So really, really high security as is our system, which is stored on a platform called Superbase. So getting at your data there is not as much the issue. I think what you're referring to is the privacy policy and the data usage policy. So if you use the free version of chat GPT, I believe your data is within the agreement for open AI to study your usage and your behavior to improve their system. The way that Granable accesses open AI is actually through sort of like a different door, so to speak, called an API. And the policy of the API is not to use that data for training or improvement of their models. So because our data is going through that door, like software talking to software, it's not being used. It is stored for 30 days so that their teams can check for malicious activity after which it's deleted. Okay, Leon said when uploading documents, what protection are there in your platform so it won't be shared with an unauthorized user? Kind of same question you just answered. Yeah, I mean, it's really that's it's the way that our database service provider is structured. I mean, there's a fundamental layer of security that they create that is SOC2 compliant, which is one of the main standards used. And then there's a way that our engineering creates the separation between different organizations and users. So I mean, for the software to work, it really needs to be linked with you and your profile and your organization. So it really isn't an occurrence where you could stumble into another user's data or that it would get mixed together. Okay, Dr. Joe says I won't grant funders see that your grant application is quote unquote fake, the fact that it was done by AI. I mean, open AI, the company that created chat GPT just shut down their AI writing detector, because it couldn't detect AI written writing anymore. I've seen this across the board. A lot of people tried to create these sort of checkers to see if it was AI drafted or not. And none of them were effective, because there's this cat and mouse game where the model will get better, the detector will get better, the model will get better, the detector will get better. But we've actually passed the point where it's really working. A lot of professors have pretty much abandoned that use after failing a bunch of students incorrectly. So I don't think the writing will be discernible as AI written, especially if it's using writing that was originally human written, or if it's written in collaboration with a human being. That's really a red herring here. I think funders that I'm speaking with recognize that the needle is way in the direction of inefficiency for grant applicants. There's way too much burden on the applicant right now. So the effort really is to try to relieve that burden. And I think most people see generative AI as a tool and that the focus should be on authenticity rather than originality, so to speak. And I think what we're doing at Grantable is to really say in our design process, how can we create a synergy between the human writer and the AI assistant with the human writer really in control, which is why our interface looks like a word processor. It is a word processor. And it's very different from the chat interface that you may have seen in chat GPT. So I don't think that funders are going to be able to even use tools to detect. It would be speculative. And I think it would be very difficult to prove. I'll just say one more issue is it's essentially the same technology as spell check or Grammarly. It's just a degree of sophistication and power. So if funders really wanted to draw the line at no software assistance, I think it would actually be pretty difficult for them to make that case. And I don't think they would have very strong grounds to make it if the information is authentic. It's credible. It's accurate. It's not trying to be deceptive. If the applicant is just trying to be efficient. Okay, here's a lot of votes on this question from Megan. Does Grantable platform have a shared document feature similar to a shared Google doc where team members can collaborate on a grant together? What we have is that you can have multiple members in your account so you can invite people here just by email. So you can add their email in a paid plan and you can add people to your account and they can go in and work on the documents. We don't yet have the real-time collaborative thing where you can see somebody else's sort of cursor moving around and collaborate at the exact same time. That's also a pretty technically sophisticated feature that we would add in the future when we have more resources. But right now, we do just allow a collaboration in sort of an asynchronous way. Okay. Here's another question that's got a lot of votes from Belmont AME Zion. So it sounds like this will not work with the grant portal application. However, if you export it out of a PDF, you can use that to fill out the grant. I would guess you would then cut and paste the answers to the grant application in the portal. When I work on grants that are in a portal, especially the ones where you can't see the whole thing and you have to go page to page what I'm usually doing in the past is highlighting, copying and pasting into like a Google doc or a Word doc. And then I'm working there because a lot of portals we know don't save your work and if the internet cuts out or you hit refresh, then all your writing is gone. It's definitely a best practice to work not in the portal. And also, even if everything goes right, you submit and you don't have a copy of what you put in, you can't reuse that information. So all I would say to do is just copy from the portal, paste it into a new file like I did here. So you can just copy, I'll just move this bar again. So I just copied, this is a portal submitable where you can add stuff. But I copied this whole thing into grantable where I will work on it here. And then when I'm ready to submit, I'll put it in. You don't have to turn it into a PDF. And actually PDFs are a great document type. If nothing changes, they're good for keeping things visually the same. They're horrible once you start trying to work with them as probably everybody here knows. If you copy and paste a PDF into docs into Word, spacing often gets all wonky. And so we actually don't allow PDFs to be uploaded into grantable because it just messes everything up. So either convert them to DocX, which kind of takes away some of that weird spacing stuff and then upload it or just copying and pasting what you need is also just a fast way to do it. Okay. Someone asks about the chat. If you are unable to copy the chat, I would have to get hundreds of people's permission to send the chat transcript because we use an open source and not everybody wants their information. Not everybody wants their boss to know that they were in this webinar today. So I'm going to drop my email in the chat and you can email me if you'd like a copy of the chat transcript. So Marla says last week, TechSoup emailed an article about the risk of AI. How can these be addressed? Great question. Yeah. I think there are a lot of risks to AI and the ones that I think about for the nonprofit sector I think are when people don't have a clear idea of the outcome they're trying to achieve or if they try to eliminate human team work or team members by replacing them with AI or if they try to put AI between the organization and other stakeholders. That's where I see a lot of trouble creeping up. I really think that humans should still stand atop the AI tools and be responsible for their outputs. So I've written about this. This is a perfect timing. Actually, Aretha, you're sending out some links and stuff afterwards. There's a group that's organizing a responsible AI in fundraising framework that I've just signed on to and Grantable has endorsed. It lays out nine categories from privacy to collaboration to equity and impact inclusivity and is very specific about what organizations should do in order to use AI responsibly in fundraising. So a lot of work is being done to help guide us on this path to use these technologies correctly. Some of the risks that the ones that get talked about the most in the media, those are actually pretty large scale collective risks. And one of the reasons why I'm so passionate about getting the nonprofit sector to lean into using and learning about generative AI technology is because I think our sector has a huge role to play in AI governance at the big level, at a global level. I think our sector has been concerned with purpose driven work and human and planet centered work for decades. And now there's this technology that requires that kind of alignment and that kind of guidance like no technology before. So my theory here is we need to be very literate, very fluent in generative AI technology in our own work so that we can participate and shape the dialogue that's happening. There's a seat at the table, many seats at the table, for nonprofit sector leaders to weigh in on how AI should be used in society and around the world. So there are these big risks that you'll hear about. And then there are sort of more immediate to the individual and the organization, which I think frameworks like the one I'll share can help you to avoid and find community as well to help learn how to navigate well. Awesome. Deb asks, can Granable be helpful with budgets, timelines, and other quote-unquote structured information versus the common narrative information that you demonstrated? Not right now. We know budgets are a massive time issue. We do have some ideas for how we'll work on that in the future. We are looking at how to tag data in your narrative like facts and figures more exactly and to allow your KPIs or your metrics to flow through your reports more automatically. But for now, we don't handle spreadsheets. I think that's actually a probably we'll start to see either the big spreadsheet company like Google Sheets or Excel will start to add generative AI features. And then I think some independent startups will also start to create spreadsheets that you can talk to and sort of have calculations and formatting get done for you. Awesome. So just a friendly reminder, this is being recorded and the recording will be emailed to you. If you would like a copy of the chat, please email me at asymonds at TechSoup.org. Thank you so much. Sarah asks, is there an option to upload older proposals from programs so that Granable is only searching by program? For instance, a question about community served will have a different answer for youth versus adult education programs offered by the same organization. Great question. So what I would recommend doing is you have a number of options. You could create a folder that's called youth programs, assets or whatever, and you could drag your resources into that folder. You could also tag the resources. So you could say youth programs and tag your resources. And then when you go into work on a question, when you search, you can actually filter by folder and by tag. And then that can bring you down to the level of just the information that you want. If you let it go automatic, it may not pick the right stuff. So if you do have a bunch of different programs, I would recommend tagging them or creating them by folder and then going in and grabbing selections that you want the AI to use. Okay, this is a two-part question. Someone asked this earlier about the tokens. What are the tokens used for? And then Shayna asked, for $24 a month, I assume it pays for one user, for how many devices can you download the software on or is it web-based? Okay. So tokens come from actually from the computer science that's being done with these models. Words that you give are tokenized, which means they're sort of broken up into fragments of meaning. And when these language models are trained, basically they're mapping relationships between the different parts of words across hundreds of billions of connections. And then they come up with probabilities that help them to generate the outputs that they do. So tokens ended up being a unit of measurement because there's only so much that you can put in at one time and have the model put out. So tokens are now sort of proliferating across the software sector, where that's how we're measuring and making sure that you're staying within the limit of what GPT can handle and give back to you. So that's sort of the measure of how much activity that you're doing. So if you send huge amounts of context, like if you send a whole document to GPT and say, write me one paragraph from this, that's kind of an inefficient way to use because all of that will be tokenized. Whereas if you grab particular snippets and are more precise, it doesn't have to read all that other stuff. It also doesn't have to make a guess on what you're sort of wanting it to focus on and your outputs will be better as well. So that's where tokens come from. As far as users on a paid profile, you can have multiple users. So I think it's three for the first tier. And I think it's 10 at the second tier. And then it is a web-based app. So you don't need to download it. It's just any of those email addresses can log in from any web-enabled device as long as it's tied to the email. Then you're free to work away from there. Awesome. So just a few more questions. We're almost done. Becky asked, as someone who reviews a lot of grant applications, I'm curious how similar two applications for the same grant might turn out if they both use Granible. I think they'll be pretty different. I mean, this is anecdotal. They could end up being more similar. But because we're using, we're starting with the organization's content and just kind of repackaging it the way that human grant writers do, unless the original content is very similar, then the AI-assisted content should not be very exact copies of one another. I think if you do go to ChatGPT, or again, like the restaurant analogy, and you ask for different answers, especially if you aren't giving ChatGPT a lot of context from your organization, the answers you get will sound like the average grant answer. That's what it's doing. It's kind of finding in its mathematics, in its algorithm, okay? What is the combination of tokens? What is the combination of words that should come out here? And that will tend to be more generic in feel. Okay. Shanya, I hope I have your name right. Ask any data to support increasing in productivity or efficiency and or funding by using Granible? We're getting ready to study that, as I'm sure folks in this room know. Studies are pretty resource intensive and we're trying to apply for grants using Grantable to study the impacts from my own experience in the nonprofit sector, especially working at a small BIPOC-led organization that really struggled to get funding and then a large white-led organization that won grants left and right, couldn't avoid winning grants. I think what tends to separate those outcomes are relationships and resources. If you have strong relationships with grantors and if you have an organization that can throw one or two full-time staff members at pursuing grant funding, you're going to win a lot and then once you won, it's easier to apply and it sort of snowballs from there. Whereas if you're under resourced, underrepresented, or connected, it can be really tough to get a start. So that's one of the main reasons I went to about founding Granible was to give those organizations the ability to either save time or apply for more funding, whatever suits your needs. So if you write in five times faster, you can save 80% of your time or you can apply for five grants instead of one. Generally, what I've found is that organizations that can take more shots tend to win more funding and then the ball sort of starts. You've solved that cold start problem. So we're going to study that. We're really interested in what the numbers will be, but that's our theory of change, so to speak. Awesome. And Dottie says, can Granible be connected to Airtable? It doesn't have any integrations yet, but I'd be really curious to know what integration, what that integration would serve. I'm guessing maybe Airtable is being used as like a master boilerplate database. Okay, Carrie got a lot of votes on this one. Does Granible only help with narrative portions of RFPs or can it help with budgets too? Like if budgetary info is uploaded? Yeah, I think one of the things that's really important here is that as powerful as the large language models are, they are not calculators. They're actually based on statistics. So when you ask ChatGPT what's four plus four, it will probably say eight, but the reason is not because it's calculated four plus four, like a calculator does. It's because four plus four equals eight has probably appeared millions of times in its training data. So what it's saying is when I see four plus four, what likely follows is equals eight. If you did a random five-digit number plus a random six-digit number, it would probably give you the wrong answer because all it knows is number plus number equals probably number. But that specific sequence probably hasn't appeared very much in its data. So that's where a lot of people are being misled and it's hard to see when these models are hallucinating because they're so confident. So a lot of people are asking for calculations to be done and very common ones will be commonly correct, but uncommon or specific calculations for your organization or your scenario will likely be incorrect. So knowing how the models work, you can start to see how you really need a different set of tools, a different set of software to do budgeting and to do spreadsheets. Folks are working on trying to combine those two to have this large language model like GPT type software to be able to talk with and use other kinds of software like spreadsheets, but that work is ongoing. So just be really careful when you're asking for things that are not specifically about writing narrative. Sorry, I was on mute. Deb asked, can you export your content to a Word or Google Doc file? Yep, just DocX here. Awesome, awesome. So I'm going to do two more questions. I'm going to find the ones that has the most. You've been an amazing, I could not, this would not have gone well if it was just me. Becky says, can Granville pull an impact statement from a collection of different grant reports? Ooh, that's a good one. Pull or make an impact statement, Becky, because you make the impact statements. You know, like when you do your annual reports and things like that, can Granville pull an impact statement from a collection of different grant reports? Yeah, definitely. So if you have the statement, like the prompt, so something that would say like write, what is your impact statement or whatever it is. And then when you go into this process, if you grab all of the different snippets from those other reports and click compose, and you give it any sort of additional instructions, it will draft that impact statement for you with those faxing figures. Thank you for your comment in the chat. This is available, very well organized. Thank you, Aretha. Thank you, Phillip, for sharing your deep knowledge. I'm learning a lot about Granville, but also a whole lot more. Yes, a lot, a lot more. Last question from Modesta. Does Granville have templates that we can work with to get started? We're working on those. We really want to, we have a really cool pilot program coming up with a really big environmental justice organization that is going to try to help thousands of environmental justice organizations nationwide to prepare to apply for Inflation Reduction Act funding. And what my dream is, is these very small groups that have maybe never applied for a grant. What we want to do is exactly that, create a template for that specific funding opportunity and create an experience where not only is it able to help you draft your answers, but it's really walking you through the application to make sure that you're thinking about it strategically, that you're checking off all the right requirements and going about it in a systematic way, especially if you are a small team or feel on your own. The feeling that we're aiming for here is that you don't feel so alone going after these big opportunities. So that's forthcoming. I would guess by the end of the year is when we want to have that launched. And that's a pilot. And then the next challenge after that is how we can adapt that to any RFP. So when we know the grant, we can go in and manually make sure that that experience is really good. The next level of difficulty for us would be anybody that puts anything into grantable. How can we make it so that these guides are not leading you astray, that they're actually being helpful in helping you diagnose the RFP that you're working on. So that's forthcoming. And it's a challenge that we're excited to work on. Thank you, everybody. Thank you, Carol. Thank you, everybody who said thank you, Phillip. This is amazing. The questions that were coming and your answers. I mean, and you gave detailed, thoughtful, honest, transparent answers. So I want to thank you for being a guest speaker here at TechSoup. Please come back again because there's so much more to talk about when coming to AI. Thanks for having me, Rita. Anytime you ask, I'm here. Awesome. Awesome. Bye, everybody. Make sure you're taking care of yourself as you're taking care of everybody else. Bye-bye. Thanks, y'all.