RX 24 - AI In Cross Cultural Contexts - Luke Quinlivan

Transcribed Content

All right. Welcome, everybody, to this final session. Please give Luke Quinlivan a warm welcome. Good afternoon, everybody. Thanks for being here. It's the last session of the conference. And I'm very privileged and honored to be sharing with you all today. My name is Luke. And if you have any questions throughout this session, you are more than welcome to put them in that little Slido chat. You can put in an anonymous question if you're brave. You can put your name in there. And we'll try to make some space towards the end just to go through some questions if you've got some as well. But hopefully, we'll have a little bit of fun. That does say AI in cross cultural context. It's a capital I. That's not Al in cross cultural context. I don't know who Al is, but we're not talking about him today. So let's jump into it. The next few slides does include that Slido stuff, so don't worry. It is not going away. But what I want to just quickly share to give you an excuse to walk out of here if you feel like this isn't the right session for you is what we're going to cover a bit today. I want to introduce to you the Church Kingdom City, where I'm from, because that will lay a bit of the foundation of some of the AI tools that we use and give you a bit of context as to why we are using some of these tools and hopefully introduce some cool concepts and ideas to you. I'm going to talk a little bit about multiculturalism and what that has looked like to us and what that might look like for you, how we've used some AI to help us. And hopefully, we'll have some fun times as well. What we won't be talking about is a comprehensive, full breakdown of AI. Brian Davis has done a couple of great sessions. He's sitting out the back. Didn't know he was coming. Thanks for coming. He did a fantastic session this year. I encourage you to check out the recording if you can. And even last year as well, he's got a great session on there. I've linked that in the takeaways, which will be sent out later as well. But his session did inspire us to try some stuff and get a bit creative when it comes to using AI. But let's jump into it. This is a little bit about me. I'm Heaps Multicultural. I don't know why there's two pictures in the background. But that's my family. That's me on the right, in case you weren't too sure. That's my wife, Beck, on the left, short for Rebecca. Now, you may not have noticed this, but I do have a little bit of an accent. I'm from Australia. I'm from Perth, Australia. You've probably heard of Sydney or Melbourne. They're along the East Coast. I'm from Perth, which is on the West Coast, the whole other side of Australia. We're the most isolated city in the world, so it takes a minimum of four hours to fly anywhere outside of our city. That's my daughter, Miller, in the middle. She's a gorgeous little princess, and she knows it. That's my son, Ezra, down at the bottom there. He needs Jesus. Please keep praying for him. His eyes are completely fine. He just decided to pull a face for this photo, so I don't know why he did that. But he's five. He thinks he's 10, so please pray for us. We're navigating this as a family. And then our daughter, Mila, she's three, and she loves her older brother, Ezra. So we've got a three year old who's following a five year old who thinks is a 10 year old. So we've got a little bit of multiculturalism happening in our family as well. I've been a part of the team at Kingdom City for just over fourteen years now, ranging from a whole bunch of roles from youth pastor, Bioware College stuff, film stuff, and for a long time, lot of our database tech IT kind of stuff. I'm not a developer. I'm really good at googling stuff. That's my main role. I do a lot of googling, trying to figure out how to do stuff. And we migrated to Rock in September. So we've just been on the platform for nearly a year now. We're about to celebrate our one year anniversary, but I'll talk about that more in a moment. A little bit about Kingdom City, where Heap's multicultural. Our church started just over eighteen years ago. Our senior pastor, Mark Varaghes, he is he was born in Singapore, grew up in Malaysia for a little bit, then his family migrated to Perth in Australia when he was about 10. So he's a brown guy with an Australian accent. His wife, Jemima, she was born in New Zealand, but then grew up in The Philippines in a missionary school there. So she's white, but lived in Asia, and she has an American accent, but she's Australian. So our senior passes and their two boys, you don't even wanna know what their ethnicity is. It's crazy. It's all over the place. But our pastors have a very rich cultural history between them that's already quite mixed and really interesting. But Kingdom City started eighteen years ago in Malaysia. So pastor Mark, he was in Perth, felt to start a church in Malaysia, moved to Kuala Lumpur, sold everything he had started there. A few years on that journey, a church in Perth where I'm from, reached out to him, said, hey. Look. We need new senior leader. Our church isn't doing too well. Do you wanna start to lead the church? That was about sixteen years ago. So then he then started to lead both the church in Kuala Lumpur in Malaysia and Perth in Australia. So from very, very early on, we've had this interesting mix of two different cultures interacting, and he's tried really hard and, we believe, successfully built a culture that's cohesive between our two locations initially. Since then, we have then started to see more locations start to pop up all around the world. Our third location was in Phnom Penh in Cambodia and then Singapore. And it's just been growing since then from that time. So right now, Kingdom City is in 16 different countries. We've got 35 campuses across the globe. Some of these locations are quite large. They're in the thousands. Some of them are quite small. There may be a few hundred or so people. So we have needed a platform that would scale with us as a church but also allowed a high degree of development and customizability, which is what Rock was able to offer. So we migrated to Rock, like I said, just in September. It's been a great journey, but the Rock community has been a huge, huge factor in us jumping into this as well. At first, were a bit intimidated about what this would look like. We've got a great team of tech people, but we weren't too sure. Is this going to work for us? Is Rock going to be the best fit for us? And so far, it's been going really well. We've been learning a lot and meeting heaps of incredible people. Many of you who I've met as well have helped us out a lot. But that's a little bit about our story. So being branched out in all of these different cities around the world, we have all different languages that we interact, and we have all different histories and cultures. So it's always been a part of what we are doing as a church. And when we first started to talk to Triumph Tech, and we've worked with Nine members as well about what does globalization and multiculturalism look like in rock, their immediate answers are we're not too sure. We don't have many churches using Rock in our sort of use case. So we were super excited to jump on and try to figure out how do we do things like mobile phones with multiple country codes? How do we do things like translations? What exactly does that look like? So we've started to move into a space where we have utilized artificial intelligence to try and figure out how can we help with translating a lot of what we do as a church. Here's a couple of tools that I'll immediately plug because we have been using them for a little while, and I'll give you a demo in just a moment. But initially, as many people do, we started translating our services by getting someone into a booth of some sort watching the initial recording, doing a dub over that voice, and then we would impose that voice, the new translated version, onto the original recording. Problem is that's very, very resource heavy. First of all, you need to have people on hand who can do that sort of stuff. You then need to have a production team who can edit and work with that and then put it all together and then produce it and then put it back up. And then you need to make sure the person translating understood our culture as well and how we think and operate as a church and do all kinds of little stuff like that. And it is possible, but it is very costly. And it wasn't quite a resource that we were able to see working for us long term working that way. So that's where we started to look at some translation tools to help with that. Now I want to talk about Rask. Rask. Rask. I don't quite know how to pronounce it here. So r a s k dot a a r a I. What that will do is it will take a video file, you put it into Rask, and then you tell it what language you want the dub to come out from. But here's a really cool thing. It will keep the voice of the original speaker, and it would just translate them to the new language. Now let me give you an example. So here we have a sermon. This is our senior pastor Mark Vargu speaking quite recently. And so let me just play. This is the original version on our Kingdom City YouTube channel. This is our English one. Wanna issue you today and to everyone online. I've entitled today's message a forty day bold prayer challenge. Alright. So I'm gonna pause it there so you get the idea. So it's speaking in English. Now I'm gonna go to our Spanish web page, and let's have a listen. So you can sorta get the idea. It's actually keeping his voice. It's just changing the words that are coming out. And then lastly, Mandarin, which is very popular in lot of our countries as well. Okay. This one did not come out. Okay. So for whatever reason, this one didn't come out with his voice. But we do have to do dub translations every now and then. There are some languages where it doesn't quite work. We do a lot in Cambodia and put on pen. So Ras doesn't have Khmer, the local language that they use there. But that has helped us enforce an atmosphere that we have at Kingdom City called being relevant to our city. We have 10 atmospheres, and one of them is being relevant. Whenever we're in a new location or in a city, we want to try to be as relevant as possible. We want to try to break down that language barrier wall between people that might traditionally be there. But since AI tools have started to come out, it's become a bit more accessible to us. One other really cool one is Weglot. This is a little plug in that you can put on your website that the Nine Ebers team helped us out with. Now it's down here in the bottom left. Let me go find a page with a lot of story. Let's go to here. Let's go About Us. And then down on the bottom left here it's probably quite small, so let me zoom in. Oh, Okay. So English, and let's change that to Khmer. This is the Cambodian language. Page loads. And it's going to take the text I believe that is not a text, it might just be a picture and then change it straight away on the page, and then Spanish as well. And it works quite well. So we can then go on and fine tune this. And Weglot are using a bit of AI in the background to help make sure that all works. And we've also applied this to our internal site as well. So this is a rock website. But we have applied that same plug into our internal site. So we have team in Mexico. We have team in Indonesia. And so English isn't their first language. And so it can be a bit tricky for them to understand what's going on. So that's been a really, really helpful tool. But there is one cultural group that sorry, let me take a step back. You might be thinking to yourself, Okay, we don't really operate in all these different countries. So how exactly does this benefit us? But I would argue that there is a cross cultural connection that you may not be aware of, and I would like to use it as a bit of a use case today to kind of share some of the other tools that we have been using as well. This is meant to kind of represent rock admins and pastors in a local church. Sometimes it might seem like you guys are all talking different languages. When you start to come into the meetings and talk about connection requests and SQL and love and all this sort of stuff, I don't know about you. This probably just happens to us. But our pastors look at us with just a blank face. Was like, what are you talking about? I don't quite understand. And vice versa, the pastors will be coming to us with ideas or questions or they want to know how to do something, and they're just using language that may be quite foreign to us. Very often, I will argue that there are quite different cultures when it comes to the tech team, the rock admin, and the end users, our pastors or our incredible volunteers, and these two different groups of people. Now, our staff, we have nearly 300 staff across the 16 countries that we're in who are all amazing. They're phenomenal people. Many of them work full time jobs, and then they come and do church as well. So we are needing to simplify our communication between the two groups of people and make sure that our team, as ROC admins, we can make sure that we are getting information out to them in ways that are effective, in ways that will actually help them. And then vice versa, we also want to get their feedback. But they may not know how to articulate their frustrations with the system in a way that is helpful for us. They might just say, oh, it's not working, and they'll send us a message. No problem. What's not working? No response. Hey, can you let us know what's not working? The form's not working. Awesome. Which form isn't working? The one on the website. Oh, Okay, cool. Which one? Can you send me a screenshot? And it's like a not a screenshot, but it's a picture of the form, and it's too close in. It can get a little bit tricky. So we've started to use some AR tools, just like I was talking about how we are trying to be relevant to our city, to make sure we are staying relevant between our team as well and hopefully get to the point where we're all high fiving and happy. Everyone else was using these real cool AI generated images. I just got free stock images that I could because it seems to emphasize things really well. And I love stock images because they're so just corny and boring. I try to plant them across our website to see who notices. One person said, I've seen that. Okay, cool. Thank you. Okay, AI. Let's talk a little bit about AI tools and what they are. AI is something that has popped up everywhere, absolutely everywhere. Nearly everything, every product, every SaaS has AI as a part of it now. And it has become a bit of a buzzword. It has become a bit of a way to try and sell a product or to try and let you know, hey, we have AI, just to make sure that people are in the know with what's going on. So it can be really tricky to figure out what exactly is AI doing, or how is it benefiting this product, or what exactly are we supposed to do with this? So myself, I've been asking our real smart techie guys what exactly is going on? How can we use this sort of stuff? And I want to give a shout out to Nick, one of our incredible developers. He's in Malaysia. And we've got some great team members John, Melinda, Radiance, Thomas, Vanessa, Grace. They're all amazing people. They all work in ministry as well. So they're not all just doing rock stuff, but they're all also using this stuff as much as we can. Now, here's a very, very simple breakdown, layman, non techy idea of what's going on when it comes to AI. I like to think of AI as three parts. There's three parts with it. There's a lot more parts to it. If you go ask Brian, he'll break it all down for you as well. But the first part is the data resource. So what exactly is the information that the AI model is getting? For example, you've probably heard of ChatGPT. What that is doing is using all the information it can get on the internet, a lot of information. But there are tools where you can scale that right down to, say, just your business standards of practice documents, which I'll share how we do that in a moment. So there's the actual source of information. And the more comprehensive and detailed that source is, usually the better output you can get from the next part, the contextual language model. So that is really the phenomenal tool. When you're chatting with AI, it's that language model that's interpreting what you're saying in a contextual way for the data resource and then outputting it in a way that makes sense to you. If you've ever used Google, which I'm sure everyone has, usually if you want a successful Google result, you have to put your query in a specific way. If you just put in restaurant, you're just going to get a list of restaurants. If you put restaurant near me, you might get a bit more of a refined answer. If you put a Mexican restaurant near me in this area, you're going to get a lot more of a clearer answer. The cool thing with AI chat models is you don't have to start off very specific. It's designed to try and get more information out of you and try to figure things out. The last part is that sense of self learning, that the AI seems to be learning what you're saying. Now, my friend Nick, he's been teaching me a little bit about what's going on behind the scenes. Most AI models are using one of two kind of self learning models. The first and most common is called Rags, Retrieval Augmented Generation. And that's simply taking your data resource, the stuff at the top there, retrieving it, and then generating an answer for you. But the other more advanced model that I believe ChatGPT is using as well as Meta's Lama is using is called fine tuning. What that is going to do is try to learn from your content without you having to teach it and then try to refine its incorrect answers, which I'll talk about again in a moment. But this is a bit more CPU heavy. It's going to use a bit more resource. And it's wanting to improve the answer it gives you over time. So for example, in theory, we all have documentation manuals of some sort, for example, like the rock documentation. But there's always going to be use cases where we sort of deviate from the manual and we do things a different way. Ideally, in a perfect scenario, the fine tuning model would actually figure out what those deviations might be for your use case and then provide helpful answers when you ask the data resource or something like that. So that's a very basic idea. But these three aspects, again, there are a lot more aspects to it we're going to talk about a little bit now. So one thing to be aware of is what's called hallucinations. AI tools are a people pleaser. They want to give you an answer. Sometimes it makes up an answer. Sometimes it's just going to generate something that might sound right. But then if you were to ask it, is this the correct answer, it would just say, no, this is not the correct answer. So this has been an interesting ethical topic that's come up a lot when it talks about AI. There's been a lot of cases where lawyers have been using AI to try and generate answers on cases, and it just gives them a wrong answer. But then they'll go ahead and use what the AI has generated. Then people have come back and said, that's not even a real case, what you're quoting. There's all kinds of stuff. So AI, it's called hallucinations. It's one thing to be aware of. And that's one aspect of how we are using AI when it comes to Rock admins and our passes. So let's get a bit practical. When it comes to building some AI stuff, it's almost a bit like a Rube Goldberg machine where you have all these crazy bits all over the place just to kind of do one action. I'm not sure if many of you have seen Mr. Bean, but there's a great example of that when he's got an alarm clock going off and he has this elaborate contraption just to turn it off. He could just turn it off himself. But that's what a Rube Goldberg machine is. It's just an elaborate contraption to do one simple thing. Sometimes building out AI tools can be a little bit like that, especially when you want to be careful of hallucinations and all this sort of stuff. I'll come back to this in a moment, but let's move on. So one thing that we do is we document everything that we do. Like I said, we migrated in September. And one of the first things we made sure that all our Rock admins did is every time we had a question from the team, we made sure we documented it. We made sure we made an SOP standard practice manual of some sort. And this was for the tiniest thing. Like, how do I get to the search bar? We made a document for that. How do I add a small group to my campus? We made a document for that. Every little thing, we made sure we made an SOP. We did this using this tool called Scribe. You can see them on scribehow.com. I've heard this come up a few times in some previous sessions today and in previous years. This is a fantastic tool. What this will do is it's almost like a screen recording. So you can tell it to start. There's an extension you can put on your browser. And it will follow along everything that you do. And every time you click, it will then take a virtual snapshot. But then it will also summarize that action that you took. So it's really hard to see in there. But for example, step number four, it says click this text field. And then you can edit that as well if you need to. But that has saved us hours, a significant amount of time. To make something like this would take a huge amount of time every single time we had to make a manual. We would have to take screenshots. We'd have to put it in maybe a Word document. We'd then have to annotate the image. We'd then have to put a description for every single image. We then have to proofread it. And that was if I was doing it. If the next person did it, it would look completely different. The annotations would be a different color. The language they use will be very different. If we needed to change it, only the person who had access to the original draft could go back. So there was all this stuff that was really hard. But by using Scribe, we had a shared team account. There were a few of us that would use it. We would go in. We would make the SOP. We'll then export a PDF and put it into Microsoft Teams for our team to then go and reference. Or we would just send them a link to that PDF as well. If we need to change it, not a problem. We would just go in, make the relevant changes, change the annotation, and within seconds, we had a new version, and we're ready to go. So this was a fantastic resource, not super expensive. I think maybe a few hundred US dollars per year for five users. But in terms of the amount of time we've saved and being able to onboard all of our pastors with this sort of stuff, a huge benefit. Now, the next thing we did is we then provided regular training. Quite often, we'll do a series of master classes, where we'll spend a week doing all kinds of training for our team. Like many of you, we are trying to introduce new features, or we've made updates to existing features. So usually about every four to six months, we have been doing online training for our team, so just a bunch of Zoom meetings. So it'll be myself. I have my screen up. And we'll walk through some of the changes and just show our team, here's what you need to do. If there's any new people that have joined our staff, we can direct them to these videos. But what's really cool is Zoom has also started to build in some AI functionality into Zoom. They call it Zoom Companion. Now the two features that most excites me are twofold. There is the transliteration. Yeah, that's right. It will take everything I'm saying and then put it down onto a document. The second thing is the meeting summaries as well. So it will take everything that's said and then create a very simple summary. So what we've been able to do is use that exact same documentation method and then share that with our team. So if they want to then go and find a specific piece of information, we don't have to go through and transcribe the entire recording. We don't have to come up with our own summary. Zoom does that itself. And usually, depending on the length of recording, might only take a few minutes to process that out. And then we can then share that with our team. So we have two lots of documentation. We have the Scribe SOP screenshot stuff. We then have the much longer documents that we're getting from Zoom AI. But we have started to encounter a new problem. We have all this documentation. We have all these resources. But that can be very overwhelming for a lot of our pastors. To our tech team, it makes absolute sense. It's organized into categories. You go to this folder on Teams. You just find the relevant area. And bam, there you go. There's your SOP. But for our pastors who don't necessarily think that way or they work a full time job, they can't necessarily sit down for hours a day to watch this content, it's a bit harder for them to go in and find specifically what they need. So we have been looking at other solutions to try and make this content a bit more accessible to them. So we've been trying to make it a bit easier. We want to get our content to our pastors, and there are some really interesting AI tools that we've been using to do this. The primary one is called Kodi. Now, you may not have heard of Kodi, but there's a good chance you have used it. Kodi is what Rockchip is built on. It's the exact same tool. We just asked Rockchip, hey, what AI model are you? And it just told us, hey, I'm Kodi. Well, Okay, cool. Let's go have a look at that. Now, the way Kodi works I mentioned before the data resource, the first part of most AI chat models. What you can do with Kodi is put in all your business documentation, and it will learn it. It uses the rags method. So it's more just a one to one, here's the information that you've got. It's not quite at that fine tuning level where it's going to learn from the data that you've got. But what we've now got is a tool where we can put in all of our business documentation, and then people can start to chat with Cody to then try and get some answers. I might try to do a demo, but internet may not be the best. But what that means is we now have a way where we can get all that documentation from our Zoom meetings, those massive transcribes that takes pages that no one reads, put it into Kodi and say, hey, just learned this, and then people can start to ask questions about it. People can start to ask specific things that we've done from our training in a conversational way as well. So let me try did I get oh, I didn't get it ready. Okay. This is a little bit sneaky of our back end. We call our Rock instance home base. And let me go to SOP final. We put beta on there just to make it sound like it's not really going to work, but this is what we do. Okay. So let's pretend I'm a pastor, and I want to say we trained our team to try and use this as an SOP fighter. And I'll explain why in a moment. Okay, so it's going. It's looking at our resources. How do I add a connect group? So connect group, that's our small group. Let me zoom in a bit there. There we go. So what it's done is it's now given us a step by step list from our SOP documentation. And it's also resourcing our training content as well. But let's see if this works. Because the majority of our documentation is screenshots, it doesn't quite know what to do with that. So the actual PDF it's reading may not have that much information. That's not the answer that I was looking for. These are not the droids that you're looking for. What we have started to train Cody to do is to look for the URL to the SOP, which is on Microsoft Teams. If you remember before, I was talking about hallucinations, how AI can tend to just make up an answer to keep you happy, to give you an answer of some sort. We did notice that start to happen quite a bit with our documentation. It's not the fault of the AI tool. It's because of our data resource. Because so much of it is screenshots, at this stage you can upgrade to high models of ChatGPT where it could look at screenshots. But it did start to make up answers. So that's why we've called it an SOP Finder. So what we are trying to train our team to do and there's a bit of an explanation on here is to help them figure out a way how to get to the document they need to see to do what they need to do. If you've ever used the search bar in Microsoft Teams, it's terrible. It doesn't quite work. You'll put in something and it'll give you something else. Copilot is meant to help with that. That's Microsoft's AI tool. I've not really used it extensively enough. But this is at least a star for us to use AI tools to try and bridge that gap between our rock admins and between our passes as well. Again, relatively inexpensive, not free, but way to get us up and running. It takes me all the way to the start. Let's bam, bam, bam. Okay, Kodi. Okay. The way we make this work, there is a little bit of a workaround we have to do. By default, Kodi won't allow you to put external links. Sorry, it won't allow the user to click an external link. What we do every time we create a new SOP, we upload the PDF to Kodi. We then get the original URL to that document in Teams. And then we put it at the bottom of the item in Kodi. Say this is the original file with the URL. And then we've allowed our Kodi instance to let the user click an external link. So what that means is when they are searching through that chatbot, most of the time it didn't do it then, awesome. But most of the time, it will then provide them with a link to the SOP on Teams. So it means we can keep stuff in Teams and keep our documentation in there because we want it behind that login wall. But it means we can still use Kodi in a conversational way as well. So it's a bit of a clunky Rube Goldberg workaround. And we do see this getting better over time as well. Here is oh, yeah. So here's how we have to do it. So here's just an example of how to access a report. So it's automatically taken out the images, which isn't that helpful. But we just put link, and that's the link to the original SOP. And when someone will ask it I should have just done this in my demo. Next time. When someone asks for it, how do I find the Faith Explained report? It will then give them that link to the SOP as well. So one day, we would love to make sure all the answers are clear in Cody, but we will need to enhance our documentation over time. Now one thing did I not oh, cool. Yeah, Okay. One other thing we've then started to do is Kodi will then record all the chat logs as well. What we do as a way to get feedback from the team without them knowing is we'll take all those chat logs, We'll then put it into an AI source of some sort, say, hey, here's a bunch of questions that are coming in. What are some common themes that are coming up? Or what are some repetitive questions? And we don't know who the questions are coming from. They're anonymous. But we have a list of all these questions that have started coming in. So we have a big conference coming up in Kuala Lumpur, Malaysia, and in Perth. We've started to see more questions coming in about our conference and passers asking, how do I see this sort of information? So what that's meant, as a development team, we can see, Okay, the passers need more information when it comes to our conference. So as a result, based on their answers and based on the summaries that other AI tools have been giving us, we've been able to develop reports that have helped our passes see what they need to see, help them see the stuff that they may not know how to ask us. Usually, will get a question, hey, I need a better conference report. No problem. What would you like to see? How people have registered. Okay. Can you please help define what that means? And it's just rarely enough information. It can take days sometimes with the back and forth. And you add in a language barrier and a time zone barrier, that makes it really hard. Whereas something like this, we can start to almost extract stuff out of the passes without them knowing. It is on there. You may have seen it on screenshot. It does say that we record their answers. But this has been our way to try and use AI to bridge that gap between the tech team, between the passes, and try to get stuff out of them without them knowing. So what's next? Where would we love to go with this? The first thing is some deeper rock integration. And at the keynote for Rx24, John shared some of the exciting stuff they are looking to do when it comes to AI. And one idea that we've got and again, this is inspired from Brian as well. But somehow on the page, put hidden text that explains what that page does. And then ideally have a button on the website sorry, on the internal page where a pastor who might be new to staff or new to their role clicked that button saying, I need help with this report. We have a lot of reports that can be quite detailed, and it can be overwhelming for the pastors to figure out how to use that report. So on the page, if we have a bunch of text somewhere that explains what it does with a button that can just refer back to that text on the page so ideally, all they'll need to do is press that button and then maybe using a Lava short code, it could send that text on that page so the chatbot knows what the question is, send an answer back to the person saying, this is how you use this page, or this is what the point of this page is, or something like that. So building that more deeply into Rock so they don't have to go to another window. They could all do it within the same page itself. Eventually, we'd love to start using fine tuning conversation models and make the chatbot a bit more reliable and less prone to hallucinations or just giving false answers as well. And eventually, one day, we'd like to open this up to our wider church membership base as well to give them ways to access our content or to ask questions about Kingdom City in a conversational way. We have an amazing online campus as well. We have all people from different parts of the world asking us some really interesting questions. And a lot of that time, our pastors are just under resourced to get back to people. So we would love to try and find a way to help them. Great. So look, that is the end of the presentation. But let's go to see if there are any questions. Okay. Let's jump into here. Does AI know that we're talking about him? Yes. I believe AI's nickname is Skynet, so just be careful with what you say about them. But yeah, does AI know he's talking about? Probably. Very likely. Do you know of any software that does live translations? Okay. Yes, this is a bit trickier with live translations. In terms of voice to text, there are a few that we've started to play around with. YouTube has some stuff built into it as well. But we've started working with Sardius. Sardius do a lot of live streaming solutions. They have been working with IHOP for a long time. They do all the live streaming for the Global Leadership Summit as well. But they've been starting to utilize AI tools to help with translations, to try and help bring information and translate it alive as you go along. The live dubbed translations is a lot harder, a lot trickier. I do believe more people are trying that out, but we don't currently have a solution for that. But it is something that we are looking into. I'm on Zoom with people. They'll be hesitant that AI is being used. Does Zoom AI alert everyone? I believe it does. There's another tool called Otter.ai that does something similar. We've turned that off. You have to talk to Zoom support because we don't want other third party AIs being used in a lot of our stuff. But I do believe it will just like it lets you know it's being recorded, it will let you know that an AI summary will be generated as well. But I do believe as much as possible, these companies are trying to be transparent to let people know AI is being used here. And that's something that we are figuring out as a church, that we want to try and let our church members know when AI is being used. One thing to be very clear is we don't want AI to touch our person records. And I believe John shared the same sentiment in the keynote as well. As much as possible, we don't want that in that space. What we would be Okay with is AI looking at metrics or looking at information about the person records, but we don't really want there to be too much cross contamination. We would love to see AI helping our pastors ask questions about reports and things of that nature as well. Is there a reason for not putting documentation directly into your Rock instance versus Microsoft? Fantastic question. So we did actually look at that. But Cody can't look at documentation that's behind a login. So what we've done, what we tried to do is put our documentation onto Rock and then say to Cody, hey, have a look at everything in all the folders in all these pages under here. But Cody can't log in and see that information. I contacted their support to see if there's some way we could make it to work. And they said it's not something we have right now. It is something they're looking at doing. Once that is up and running, then it is definitely something we would look to do, is to somehow get the content from Scribe. You can actually embed that directly onto a web page, so you wouldn't even need to export it to PDF. But you can embed content from Scribe onto a web page, and then Kodi would then read that as it gets updated or as stuff gets added. That's the dream scenario. Or just have something built into Rock that does all that as well. But that is something we would like to do. But that issue of Kody not necessarily being able to read past a login is what's preventing us from putting it all on there as well. Does Kodi work with your internal translation tool? No, there's no translation on Kodi. It's all just actually, that's a good question. We haven't even tried it with another language. I suspect it would work if someone tried to type in a different language. It might give them an answer in their relevant language as well because that's a big part of these chat tools is being able to immediately recognize the language that's being used. So it's possible. There is a possibility there as well.