# How to Use AI at Work (without feeling overwhelmed and behind)

## Метаданные

- **Канал:** Culture Amp
- **YouTube:** https://www.youtube.com/watch?v=9ayUyxx2hQQ
- **Дата:** 24.04.2026
- **Длительность:** 1:01:28
- **Просмотры:** 89
- **Источник:** https://ekstraktznaniy.ru/video/50592

## Описание

This is a recording from a Culture First Chapter event hosted by the Global Chapter on Thursday, April 23rd, by Jessie Jacob, Culture First Community Manager at Culture Amp. 

If you'd like to learn more about the community to join upcoming events or our online discussion forum, check out community.culturefirst.com.

In this recorded session, we:
- Built our own AI-powered assistant (no code required)
- Turned real documents into insights in seconds 
- Created content like decks and internal resources using AI
- Explored tools beyond ChatGPT — including live demos

This session was facilitated by Akshay Ganesh, an AI and product leader who has spent the past 7+ years building systems that transform how organizations work, communicate, and operate.

His experience spans everything from AI-powered workforce systems to generative AI tools that process documents, analyze data, and support real business decisions — especially in HR and people-focused environments.

## Транскрипт

### Segment 1 (00:00 - 05:00) []

All right, welcome everyone to how to use AI at work without feeling overwhelmed and behind. I don't know about you all, but I am certainly feeling overwhelmed and behind, so I'm very grateful to be here and learning with you all with our amazing guest speaker. My name's Jessie Jacob. I am at the helm of guiding our Culture First community. I'm very grateful to be here with you all today, and I'm so excited to be here with Akshay Ganesh, who is a software engineer consultant, AI expert. He has done a lot of collaborating with our chapter in some of our chapters in India. So, shout out to Risti, who's in here as well. Thank you so much for making this all happen. I appreciate you. Um a little bit about today is about is it it's not about becoming some AI expert. I don't know about you all, but I do like learning about AI, but making the time to actually learn and tinker is the thing that it can be challenging. So, hey, shout out to you all for making the time for actually being here and to actually continue to learn. And so, today is about discovering a couple new tools that will help you give you some hours back during the week and also give you some confidence to try them later on your own. So, I always like to have a little session disclaimer in here that this is a community led event that is sponsored and powered by Culture Amp, but we recognize that things touch on sensitive topics and subjects. And things, you know, we're not going to get everything right, but we do aspire to create an inclusive environment. So, we want to be able to have hard conversations and discuss bold topics. So, we welcome your feedback if and how we can improve these sessions in the future. All right, I would love if you all would grab your phones, scan this QR code, and Um, draw a little doodle to describe how it feels to use AI. Now, I'm going to switch these screens real quick. Um. Come on. Here we go. I want to use this tab. Tabs on tabs. All right, let's see. Yeah, okay. The Okay, fire emoji. I'm not sure if that's a positive thing or it looks like maybe that's fire. We have lightning bolts. Hey. The eye the yellow big eyes, fantastic. That's and the spinny eyes, I hear you on that. Someone has a heart, that makes me feel helpful and sweet and endearing, cool. Okay, these are a who? A star, okay. Yeah, okay. Someone's like got this like brain Oh, yeah, shrug emoji. Yes. This is fantastic. Okay. Question marks. Yeah, okay. So, we yeah, you we get the gist of this. A lot of confusion, uh, where maybe some overwhelm, maybe there's like a little cloud here. Um. But also like we love it and we're learning and so we got some stars, some lightning bolts, and some hearts. So, I like I also help, yeah. Okay. All right. Um, great. Well, thank you all for checking in on that. I I'm going to keep us moving here. I don't even know how to remove these little things now, but all right, we're going to keep it moving. All right, so if this is your first Culture First Community event and you're like, who are these people? We are group of people around the world united in the shared belief that a better world of work is possible and I always like to say and by golly we're willing to do something about it. So that's who we are. Um we have five core principles that hold us all together. At this point now we have um about 100 global chapters who are hosting events um both in-person and virtually and this is the glue that holds us all together. These is this is these are the five core principles that we're going to ask you to embody during our time together. The first one is around fostering belonging and acceptance. Just letting the other people here and yourself feel like they can be heard, seen, and appreciated for just showing up and being themselves. We ask that you be willing to reflect and grow. So if we're on this mission to transform the future of work, that actually goes back to our organizations doing the work, but that ultimately starts with us being the change we want to see in the world. So you all are already doing this by willing to grow and learn and develop

### Segment 2 (05:00 - 10:00) [5:00]

with all these new tools. We ask that you have the courage to be vulnerable. So this looks different for all of us. So whatever that edge of vulnerability is for you, learning about new AI tools, I don't know about you all, is incredibly vulnerable for me. So um definitely already playing with the edge of that, but this concept of like we're constantly wearing masks and layers and if we're able to just be more of ourselves, gives other people permission to do themsel- do that themselves so that we're also um doing this work together. We ask that you put your learning into action. So you don't just consume information in one ear and out the other. We ask that you actually put all these things into the way that you live, work. And the last core principle is connection inside business outside, which is a fancy way of saying that in the context of this container, we ask that you connect first to this human being and that if that leads to you all doing business outside of this gathering, that's cool, but this is not a place for like hardcore sales and networking. This is a place for us to actually connect and learn and grow and develop together. So, with that, can I have a thumbs up that you all thumbs up reaction emoji either in the chat or Zoom reaction that you can agree to these today? Um, amazing. All right. Now, I want us to give a little warm welcome and some love to our guest speaker Akshay Ganesh. So, can I have like some clap emojis, some confetti, some like welcome, what's up, like come off mute, like give a whoop and a holler uh, to our speaker Akshay. Yes, thank you, Emily. All right. Thank you. Over to you, Akshay. Awesome. Uh, thank you for being here. Thank you for this opportunity, Jessie. And I'm really surprised how many people have turned up so early in the morning. Uh, again, thank you from the bottom. Um, uh, I'll give you a quick intro about myself. I'm Akshay. I am a software engineer by profession, but I love to do connects with folks, especially uh, in the since the last 4 years and help them, uh, you know, demystify certain ideas around AI. Uh, also maybe help them also probably try to integrate some parts of AI with within their own, uh, you know, workflows that they do day-to-day. Uh, yeah, yeah. Awesome. Um, with that said, right, um, I would want to get a hunch of where all of you are in your AI journey to now. So, what I'll do is I'll pull I'll send a poll link. If you could actually it's an anonymous poll. Like click on that and then show your opinion on where you are in your you know AI journey in your within your HR discipline. And how does it help? Like do you use it? Is it still some jargon that you hear from time to time? Awesome. I'm getting some numbers. I'm sure you will not see the numbers, but I'm getting 60% use daily, which is really good. And we have like the next highest number is weekly. And we have few people who have never used it in that stream. So. All right. Yeah. Yeah, let it flow in. All right. Cool. What my next would be what is your biggest hesitation or maybe a concept you would have that's holding you back from using it in your day-to-day lives or maybe you know in your stream, right? As now you see a different question on your screen right now. Um And uh feel free to like even uh chime in the chat if you don't see an option that doesn't you know that you don't uh really fall into. Can it be all of the above, Akshay. Yeah, I get it. I completely get it. I mean, that's one of the few things that a lot of people will really try based on the height. But, somehow it does not really cater to what they are looking for and that sort of gives them a push back in trying something new again. And that I that

### Segment 3 (10:00 - 15:00) [10:00]

even I have gone through it, right? And there's like n number of topics that are coming out every other uh so Yeah, and I get it. That's fine. That's all right. All right, so we have a fairly balanced you know 20 30% is on privacy and ethics. There's the same number on accuracy. Uh few people have seen, you know, it's divided between I don't know where to start and not enough time to learn and maybe one or two people have said about quality. Which is great. I I uh Yeah, yeah. All right, awesome. So, that gives me kind of like a like a platform to get started on with and how do I drive this session that we have today? Uh with that said, right? Maybe we can we jump on to the session directly, Jesse? That sound good if Yeah, yeah. Awesome. All right. So, to get started, right? I'll quickly share my screen. Okay. All right, so hopefully you can see my screen. You can see the plain old good old ChatGPT in front of you, right? Uh everyone would have used ChatGPT at this point or at least used some of its competitor at this point, right? Uh we know we all know what it does. You know, you put in some prompts, you get some answers out. Uh you can create images. I think you can even, you know, play with audio as well nowadays. And it is fairly accurate nowadays. It's been growing in terms of, you know, how reliable the output has been. But I in my journey, I use it every day. It could be, you know, as simple as getting some technical questions answered, or it could be also how do I do certain dish even that I don't have a specific ingredient at my home. What is the replacement for that, right? So, I use it for everything. I kind of don't use Google as much as I can. I if it's in front of me, I'll use this cuz it gives me the entire thing without having me to go out and look at five different places, right? So, with that said, right? How do you take this ahead, right? Um you all might have heard of this this term known as agentic, right? How do you convert your ChatGPT or, you know, move from your ChatGPT textual kind of conversation to an agentic conversation, right? Where uh the primary difference between both of them is that it actually does stuff for you, not just returns you answers in your textual manner, but it also does stuff for you. It It has you can actually have a constant set of knowledge bases, meaning let's say you have your certain policies or certain documents that you would want it to answer based on of. And you would want that constant, right? Let's say I have a specific need for my use case. For instance, I in my organization, there's like a 50-pager uh document that contains my employee benefits. It's very hard for me to go through it whenever I have a query. So, what I did is I created a custom agent that actually feeds off the same document, the same knowledge base, and answers me with the right questions. If it doesn't know the answer, it doesn't cook up. So, that's where the key is, right? So, I'll quickly go on to show you how you can create that, right? So, that's known as custom GPTs. Some of you would have done this before, not have. So, for that to work, you really have to like basically log in and uh you can go to this explore GPT site, right? So, here you have an option to create a GPT. Now, this is where the entire agent things come in, right? You can obviously, if you have like coding background with, you know, uh development in your hand, you can do a lot of things with it. But, to get started with, right? Uh how do you build an HR agent, right? So, I'll quickly build an agent for uh in this case, right? So, uh let's in in my case, I'll for this demo, I'll create a employee onboarding FAQ assistant. So, uh let me quickly put in the name, that is employee onboarding assistant. And I'll give it a nice-looking image as well. That shows it's like all AI stuff. It's just make makes it easier for me when I come to the home screen. It like I'll see the logo and then click on it easier. Uh so, you can give a small description. Helps on um getting things about uh document. And here is where the interesting part comes in, right? Uh the instructions actually make or break your agent. Um if you don't specify it clearly enough

### Segment 4 (15:00 - 20:00) [15:00]

that uh you shouldn't really don't answer something that you don't know, right? That's not given to you explicitly. Uh these models, ChatGPTs, LLMs tends to say something on its own, because that's how it's trained for, right? It's not trained to say no, unless you say it to say no. So, what I'll do is I'll give an in uh instruction here to kind of like a simple instruction that actually tells that it will be a friendly onboarding assistant and it answers questions about company policies, benefits, etc. And I'll also define a tone which I would want it to answer. And if you want it to be in a separate way, let's say you want the answers to be always in a concise bullet bullets, that's fine. You can put that. You want the answers to be in a tabular structure, that's fine as well. Whatever suits your style, you can put it over here. Those would be read by these agents while answering your queries here, right? So, and one thing that's very important is that I explicitly say that if you do not know the answer, say no to them, right? This helps a lot, right? In one of our in a in the poll that we saw earlier that it tends to cook up, right? It tends to break down on the accuracy. So, this helps a lot. Uh now, with that said, right? I'll uh Conversation starters are just some quick questions that you would want to post in. It's not an important thing. I'll put something like uh When do my benefits start? Like this would be one typical question that I would want to ask it, right? It just makes it easier to start a conversation. Why I mean, if you're using it by yourself, you don't need this. But if you're sharing this agent with other people, other folks in your company who would want to use the same agent, they can give a you know, it gives a user a friendly experience. They can Okay, I can try you know, straight away use these questions and these are the forms of question that you can put it in and stuff like that. Now, here is where the key comes in, right? You would want to upload your knowledge to it. I'll go ahead and then a document that actually has it's like a small document that I created for employee uh handbook. Uh let me see if I can quickly open that file as well over here. All right. So, this is a very quick thing. It's a made-up name, Acme Corp. Uh so, it has some core values. It has some mandate some you know, basic stuff like what is a probation period, what is a policy, how many days times off are there, etc. So, this is just a sample. This is just to get an idea. This does not really you know, it might not really exist anywhere. So, I'll just put it over here just so that I can you know, kind of differentiate. It's from Acme Corp. So, I've uploaded that document. This is going to be the ground truth. what our agent is going to look up. If it doesn't know, if it doesn't have an answer within that, it's not going to respond. That's the idea here. Um and again, so this is where one interesting thing is, right? If you have you know, paid for a subscription or something, you might have access to even newer models. Which is not really necessary in this case. And if you're doing something very complicated, you can obviously use some of the higher models. But since I have not paid, I have the default one that is GPT 5 53. You don't have to select it. It It'll actually auto-select if you don't do anything. It's auto going to auto-select. I'm just going to select it. Uh and now you're going to give capabilities to this agent. Do you want this agent to go and do web search? Let's say my query is uh in the state of Texas, what are the policies pertaining to uh you know, uh laying off, right? For example, right? Uh if that is not included, it's going to say that it's not included. But if I still assert it to give me an answer, it's going to do a web search. Right? Do you want it to do that or do you do not want it to do that? That's where this capability would come in. I don't want it to do that, but you can obviously keep it enabled as well. And do you want it to have image generation capabilities? Something like, you know, uh a quick uh image generated for putting it on my slide that shows like a graph or things like that, right? You can use your imagination. So, I don't want any of that. I don't want uh it's not related to any coding, so I'm not giving any capabilities. Um Awesome, right? Uh Now is the question, right? Now, I've given everything. This is I've given that I've defined the behavior. I've give I've given it the knowledge base. How will it do anything for me? It's still going to answer me back through the chat, right? Where does the agent behavior comes in, right? So, that's

### Segment 5 (20:00 - 25:00) [20:00]

where uh you know, you connect your tools to it. There are tools that are there. Uh I for this demo show a tool that I have built. Uh and that's added in this section known as actions. So, I'm going to create a new action. Um all these are somewhat technical. Uh you know, if you're integrating with a third-party tool like your HRMS tool or your LMS tool, you name it, right? All of them does give some form of uh APIs, you know, application programming interface for you to connect to your agent. So, I'm going to post this here. This is the basically giving it like an address to it, right? It's going to tell you what all capabilities this agent could do. It can see what are the end points it can access. And uh Don't worry about it if it's going over here. That's fine. This is just an like uh a sparkling on top of the cake that I'm putting a cherry on top of the cake, right? This is not something that you need. If you have it, great. If you don't have it, that's fine as well. Uh all right. So, I'm to quickly uh save this. And I'm going to add that bit part of uh you know, I'm going to modify the instruction to also include that, right? I'm going to add a phrase known as use the provided actions/tools if required. Right? But then just going to give a pretext to it that you have this, use this if you want, right? So, that's it. And uh I'm going to create it. Uh it's going to only for me. Yeah, sure. If you want to share with this someone else, right? You would have to uh also follow some form of uh mhm this uh mandatory policy. You would have to at- attach to your uh actions. I'll show you that where you have to do that. Uh so, basically, if you see here, right? If you edit this, you would have to also attach your privacy policy URL. So, with that only it will let you share it with others, but if you want to use it with for your by yourself, it's fine. You don't have to put any of them. So, now that I've created it, let me quickly open that from the URL that it generated. Awesome. So, I it is open now. Like let's say if I go to chat GPT again, it will most likely show up over here. It should show up. Maybe it's going to have reached some limit. Yeah, if I go to my GPTs, it's going to show up over here. Awesome. It's only for me. Right? So, I'll ask some questions like uh maybe I can pick it up from the document itself, right? Uh if I request something like how do I request time off? Uh So, it's understands it's pertaining to that uh document which has this information that it's part of the Acme Corporation. It's going to tell you the answer. And maybe I'll This is the time that I'll show you what the tool does, right? my tool is. It's a very simple lead management software that I built just for this demo. It shows like these are the employees that are there and they have, you know, seven annual leave left. They have like X number of sick leaves left and zero are pending to be approved and things like that. Now, I'm going to try to see, right? This might not work in a live demo. And we always have that demo curse with us. So, I'll try to see if I can actually use this using this agent over here. Can you list me all the employees. Fingers crossed. Yeah, so it's saying You can see that it's talking to my app, right? In your could be your HRMS app. I've just created something that on here for the demo so that you can see it. Now, you see that this is actually retrieve data from our application from our back end and has fetched it, right? There's no coding here. Those templates that you see that I copy-pasted are something that the tech team would usually handle and give it to you. So, those are open API JSON files. They are very standard templates which you import in and then that kind of lets your ChatGPT know that this is how you should access those functions. access those information from that application. I can do other things as well, right? Like I'll quickly show you what all are the capabilities that it has. If I open this, right? Yeah, so it it kind of gives you a very technical name. It lets you list employees. It let gets you get an employee information. It lets you apply a leave. It can give you the leave balance. So

### Segment 6 (25:00 - 30:00) [25:00]

if I can if I go to this chat GPT and ask for things like how many sick leaves do Rahul Mera have? Since I can want to talk it to it because it knows that there is a function that it can call to get that exact information. So, you can see over here, right? This is the agent part. Now, now it retrieves information very well. Now, what if I wanted to do something, right? Not just read, I want you to give something back, right? do some action for it. So, I see that Ananya have called in sick. Apply a leave on her behalf. Yeah, it's going to just give a confirmation because it's going to not it's not reading anything, it's going to send that information, right? So, it's going to ask, "Are you sure, right? This is what I'm going to send. Are you sure you're going to do this, right? " I'm just going to give a confirmation. And yeah, it's has submitted. Now, if I if you see this, right? This is my dashboard of my application. You see there is no leaves that are pending, right? Ananya has no nothing to be approved. If I quickly refresh this page, right? Yeah, there it is. You can see that Ananya has uh you know, sick leave that was reported by the team that is yet to be approved. Is it like is it wow moment? Is it interesting? Is it going fine? How do How do Can we get a thumbs up in the chat? Is it interesting? Have we uh seen your uh seen this before? Awesome. So, uh that's the gist I wanted to share of how you can use this agentic behavior to actually do something that you do on your regular basis. Uh now I've shown something with an actions included. You don't have to do that. Let's say if you do something on a repetitive basis, if let's say people reach out to you for certain policy questions questions or queries around that. And if it's a very big policy, you might not know all the intricate details on your by heart, right? So, that's where one application where you could use it. Uh like this is you have then you know, maybe imagine analysis for your exit interviews. Like if you have that documented somewhere, right? And if you want to ask questions around it or maybe L& D recommendations, manager coaching, policy Q& A, maybe navigating benefits, right? Everything that is repetitive for an HR can have its own agent. Now, I've built a onboarding agent. It's very important that you segregate your information. Don't give everything to it at once. Like don't give the entire like 50 PDF to it, right? It's going to get confused. It's not It'll not know where to really fetch the information from. That's where you would see the accuracy drop. So, it's the best thing to do is to have a give it a persona and ask it to do just one thing. Now, I've asked onboarding FAQ. So, it's going to just do onboarding FAQ. If you if it if I ask something around, you know, a different maybe L& D or maybe growth, right? If I ask something around that, it might not give you the answer and that's how it should be designed. That's what will let you give you the let you let this agent perform to its maximum capability. Awesome. Um All right. I think that was like a lot of information, a lot of things that you saw, right? Don't worry about if you missed anything. I'll share a quick handbook with after the session, where you can I'll we have step-by-step information on how to do this, what all are the alternatives that you can do, things like that. All the demos that I'm going to show today are kind of free, right? You don't have to pay anything, at least for the first few usages. Uh ChatGPT, of course, is free for now. Touchwood, it will be free for like a long time. All right. So, with that said, right? I'll uh quickly go on to uh the second part of the demo. That is uh let's say you have a document uh that is rich in information. Let's say something around uh employee engagement survey, uh and you have a document of it, right? Uh This can help you answer questions. But, what if you want to do analytics on top of it? What if you want to plot charts? What if you want to summarize it, and you want to make it like a podcast, right? 50-second audio that uh you don't want to really read through it, skim it and get the information out. So, that's where uh Google's NotebookLM would come in. Uh NotebookLM is free. Um you can um use

### Segment 7 (30:00 - 35:00) [30:00]

your regular old Google uh login. It's also by Google, the way you I'll share the link later. But, once you go to NotebookLM, um you go ahead and click a notebook. Uh the concept around here is not really like ChatGPT. It's not like you have one chat, and then you go to a next chat, and you continue that chat. It's like a notebook, you'll have a journal. You put in your information inside that, you ask it to give out certain information. So, it'll you can have multiple sources within it. Uh uh you know, websites, you can link it within it. It is not really uh you know, a one single chat, and then you open a new notebook or something, right? So, if you have a, uh, continuously evolving, uh, you know, thing that you do, let's say, if it's around a month long or a 6-month long employee engagement survey, and you want to iterate on top of it with various information that you plug in from time to time, that's what the, you know, notebook LM would really work over here. So, for this demo, what I'll do is I'll open another PDF file that I've kept in handy. And that is All right. So, just want to post it here. Yeah, it's a very simple, uh, employee It's just a 5-6 pager, uh, employee engagement survey report. It has good amount of numbers within it, not just text. It has, uh, like, some metrics that you would want to make sure that you really get it right. Uh, there are chart charts that are there, tabular information that are there. Um, Yep. Yep. Yep. All right. Um, right. So, I'll quickly upload these questions here. All right, I have Let me do that. Perfect. It takes a while to, like, understand what it is and things around that. While it does that, right, I'll quickly open the questionnaire. I see there are some questions. Uh, Okay. Where do I start? Yep. Yep. I see that, uh, thanks, Amelia, for your response to few of them. Uh, do do do do Oh. Yeah, I get it. I completely am I'm in your favor. It's just easier to, uh, show this demo. Um, and you will get to do a lot of things in the free period with ChatGPT, right? As in compared to Cloud. Really good, but if you want to get new, you need to like pay the $20 premium subscription and things like that. Uh yeah. Anything that is there. Awesome. All right. Uh I'll quickly right now that this has processed, right? It's actually has given me a small summary of what it concludes. Obviously, it is not containing everything. It also gives me those conversation starter questions that I can start with. Uh over here, maybe I can, you know, things ask things like, you know, what are the top three themes that employees are, you know, concerned about if I can really put it. It's going to The good thing that I found about this one as in compared to other models like other like websites like ChatGPT or Cloud is that even the answers that you get right within a document, it actually gives you where it is. Like it gives you the snippet of uh what information it is. It tells me that it's from this PDF. It also gives me a snippet of what actually where it actually found it. So, then it's kind of, you know, bases your the responses that it gives on top of the actual truth that is the document that we have given to it rather than, you know, this gives you confidence to actually trust the data that it actually sees rather than it cooking up something on its own. Uh so, this is one. And let's say if I want to, you know, there's a meeting coming up in the next 2 minutes and I want to quickly uh cook up some five questions that I would want to discuss with the leadership. Uh and that's what I'm going to do. So, it also understands that, you know, certain data points are not really really required to be discussed at a leadership level. Again, this is something that is evolving, but it has worked a lot for me. Um and the best part is that it is not just you know, pertain to just text textual answers. You can also go ahead and you know, you can convert and create a small podcast, create mind maps, generate a report on top of it. You can actually say uh let's say if you wanted

### Segment 8 (35:00 - 40:00) [35:00]

it in English, really short, and generate um audio uh very brief. I wanted it to be brief. Awesome. It's going to take like a minute or so to generate, but it actually gives you that thing. And it's really sounds really convincing as well. Uh so Awesome. Let it go through. All right. Have any of you again quick quiz time. actually used these tools before? Like obviously chat GPT I know, but if have you any of you created any custom GPTs, created an agentic behavior on top of it, used notebook LM for your research? Awesome. Yeah. Perfect. That's a really good use case to create onboarding podcast. Because if you have your onboarding the all the high-level information that you would want to introduce to a uh onboarding employee. to download it. They can actually give you that kind of you know podcast kind of you make it fun, make it useful for them. That's a really creative. You know, creating flashcards for your intro email meetings and things like that you can use in your presentations and awesome. All right, this should have been done right now. All right, looks like it's taking couple more minutes. Maybe I can open something that I've already generated, right? I can see that you can see the time here. Okay. Like people are making more money and they are mathematically — able to hear? It sounds like Can I get the thumbs up? counterintuitive I know. Almost like a um a glitch in the data or something. But it's actually one of the most revealing paradoxes in modern organizational psychology because it proves that the old-school concept of what a paycheck means is just well, it's fundamentally broken. — Which is exactly why we are opening up a corporate x-ray today and why you listening to us right now are the crucial third member of this conversation because whether you Awesome, right. So, the persona, the number of people you would want the the audio track to have, the kind of conversation you would the style of conversation, all of them can be modified in that specific instructions that you give. You want it to have a certain voice, certain pitch. You know, the text box is enough. All right. That's a perfect use case that to use it to get scholar scholarly articles. That's really good. It also has an It also You can also do web searches over here, right? If you want to do some form of Let's say again it's going to be you'd have to also do your own thing, but if you still want to do some high-level information around what are the certain metrics that you would want to fetch up from your competitor, you can it can actually do that web search for you. It can collect the sources for you. It will show you these are the sources. And then you can obviously you know select those sources or unselect them for it to answer based on If you think if you see that there's a website that you don't really find it reliable, you can obviously unselect them and it will not use it for curating its answers or any of the other studio applications that you ask it to do. Awesome. All right. Now we'll go on on to our final you know demo that we have over here that's for today that is for content generation. All of us or probably most of us would have created you know a slides for their living at one point of time, right? Or maybe once a year. I do it like every week, right? So I use these tools and the life senior like I used to spend a lot of time in you know selecting the right fonts, selecting the right colors. Well, I'm not really that creative in terms of you know getting these things up and running. I'll take time to get that right. Uh so this these tools have actually helped me in you know freeing up my time. At least get me the I can then maybe pick and choose from them as well. So the tool that I'm

### Segment 9 (40:00 - 45:00) [40:00]

introducing is gamma. app, right? Uh I'll share the links, don't worry. And this is also having a free version. I think you can create a generous five to 10 slides uh in your free tier as well. Uh, so if you once you log in, uh, it'll actually can create new. Uh, it'll ask it'll let you create from the templates that is already there. Uh, or there is some one-line prompt that you have and then you want to generate on top of it. That's, uh, you can also do that. Uh, it'll also show you our recent ones, recent prompts where these this is what I did earlier. Or if you already have a document, uh, let's say in this case uh, the uh, you know, engagement survey. If you want to create something on top of it, it you can also upload that file over there. It'll use that as the baseline information on which it's going to create the uh, you know, uh, deck. Which is what is you know, practically useful because you cannot uh, put in all the information that you would want uh, in one single prompt, right? That's not really practical on your day-to-day life. So for this demo, I'm going to just uh, show this, right? So I'll ask you to create a new hire orientation deck. And I ask you to include certain aspects of that part as well, including the company culture, the first week expectations, key contacts that you want to, uh, you know, work with, some tools and systems that we use within the company, and also uh, a small welcome message from the leadership, right? So I'll put that in. And and I'll ask you to I can control how many cards do I want. I'll let's say put five. Uh, this can be default. Uh, and it'll actually generate an outline. So this is where you can this is where your input would come in, right? This is where the automation breaks. Here is where you would want to come in, customize it, change it, because it's going to give you a boilerplate for you. And it's based on what it thinks. It'll show you that this is what your culture means. It might not necessarily mean anything over here, but you would want to put it something on your own. You can like happy welcoming blah blah blah etc. right. So, you would want to update it or modify it based on what really makes sense for you because this whatever is there, it's something that you would not want to take it as such, right? You would want to curate it and then put it. Once you do that, it it's going to remove the you know creative thinking from your end. So, if you have information, you know plug it in, it'll actually create it something for you. Now, this is where you can give the you know text context. Do you want to be a minimal one, very detailed, extensive. You have a certain theme in your You can even customize them, right? Beautiful ones that are like some upload something on your own. You can even upload a theme as well. And you can if you want some images in the slide, right? artifacts and things like that. Do you want it to be you know searched from web, which means I'm going to be more or less going to be a Google search based images or you want it to be you know AI generated images. So, I'm going to select AI images. Google slide. And once I click on generate, this is one fastest. I'm not sure how they do it, but it's really quick. They like you can see it in front of you. Generates it's a real time thing, right? You'll get to see the overall structure of how the deck is going to look like. You're going you'll get to see how it's bifurcating the statements across multiple decks. And obviously, you can also customize it at this point as well, right? You can add a new deck if you want to split something over here. You have the options all the tools to update it here, add a new deck and things like that. You want to you know go back, something doesn't look right. You want to go back to the prompt and update it. All of the things you can do it from here, right? So, this is what I said, right? This is a doodle that can create it on its own. Right? It's an AI generated one. Now, I use it to get inspiration. I don't really use it as a copy-paste because my company doesn't really allow me to upload my assets anywhere else. And it really doesn't I don't have something that is very, you know, very close to what my company uses as well. So, the fonts would be different, the background colors paging styles would be different, etc. Right? So, I usually copy stuff from here. I take inspiration. Okay, yeah, this is the right way to, you know, illustrate something. And then I take it back. But if you are the lucky few who you can, like, you know, directly do a plug-and-play from here, you take you, you know, you share it, click on share, you can export it. You can export it all these kind of, you know, free ones that are there. And it's

### Segment 10 (45:00 - 50:00) [45:00]

going to work out really well for you. So, I see a couple of questions. You need to see if Canva does something similar to what Yeah, yep, yep, it does. It does do that. But I'm not sure how much of a AI freemium that do they offer? Like, right now in Gamma, they do give you somewhere around uh Yeah, somewhere around 400 odd credits for free. And you can use it for a number of things, right? If you have a very big slide that's like 100 pages 100 sheets long, it's going to probably exhaust it very soon. But if you have like one, two pager, you can usually use it for a year. I usually don't run out of it because I don't really use it that for that big of a deck at all. But Canva, I think they only have a like a paid version to start with. Yeah, yep, you can import it. You can import certain like if you create something new, right? If you have a file, if you have exi- existing presentations, you can upload it. It will actually look at those as well, uh, while creating it for you. So, like for Akshay, you could upload like your If you're giving a keynote or a presentation or whatever, you could upload your talk track, like your script, and then be like, go and create however many slides based on this. Yeah, it's right, exactly. So, you can upload that. You can either upload it like your scripting thing, or if you have certain, um, color theme that you would want to want it to stick to, it actually picks that as well. In fact, it does both, right? Let me upload the engagement survey over here. Uh, so if you remember, it has like this bluish theme with gray in the middle, and there's white, and there's like green hints of green and there. Uh, so I'm going to assume that's going That's what it's going to take in. So, uh, I want to create a presentation again. And, uh, All right, it has pre-selected some of them for me. Okay, I'm asking it to generate. It's taken the content from my document. There's a like good amount of numbers that are there as well. So, Oh, yeah, I think you would have to change the color from here. Yeah, that's something I would have to probably see what How do you exactly change it? Uh, but I think before you generate it, you will have an option to select the themes from there itself. Uh, if there is, uh, if there is a set number of colors that you would have, you know already the the codes of the colors, HTML codes, you can plug those in. And if not, you can I think you can even ask the prompt using a prompt to use this color as your, you know, background and things like that. Uh, but yeah, long story short, it's also, you know, if you have a script for yourself that you would want to, you know, write your story on top of. This is like a great tool to start with. Awesome. Um All right. So, enough of demo, enough of, you know, quickly running through 50 pages and like multiple tools that you would want to, I know, if I hopefully find it interesting. Maybe I'll I'll share the resource links. You can all try it out on your own time and you know, if you still have some queries and all things like that, feel free to ping me Instagram. No, sorry, in LinkedIn or via email. So, I can help you out with that. All right. So, with that said, right? I would want to take a step back and ask you a question, right? Um I would want you to take one public commitment, right? This week you would use AI to do dash. And one risk I'll watch for is dash, right? I would want to understand where what do you feel after this session in terms of getting the accuracy right, more things done, in terms of understanding where it will break, where it will not break, what it can do, what it cannot do. Um maybe I'll put a quick poll as well. In the same link, the Slido link, you should see a question popping up right now. Um those who have missed it, I'll post it again in the chat. You can open it again over here. Perfect.

### Segment 11 (50:00 - 55:00) [50:00]

And we have four people, six people typing. Awesome. Yep. Create your create a brand and message guide Q& A. Uh accuracy and visual elements. Yeah. So, I have that. I mean, it's also that's where something, you know, it is difficult as a human also to kind of communicate the the visual element of it's like something you would have to feel have to figure out a way to communicate exactly what's in my brain to someone. I mean, I would probably that's the next uh breakthrough you would going to see because the information translation is the come sometimes can be the bottle neck. Create onboarding helper. Uh You'll uh you'll have a watch for accuracy. Uh AI discover ways you can improve getting content out to employees more quickly. Yeah, I see accuracy like three, four times. Work on them. The job search, perfect. The notebook LM probably does do that as well. You can search across multiple websites or hiring platforms. Uh create the brand page. Yeah. Onboarding helper, organize projects. Perfect. Uh That's all those are great uh like use cases to, you know, start tackling with it, right? Uh Yeah. I see. Yeah. I mean, uh accuracy and the bias, right? Those are the risks to watch for sure. Uh that's the responsible adoption side that you want to, you know, keep in mind. Yeah. Com- communications, policy drafting. Yeah, those are the quickest wins, right? Those are the things that like more or less clerical. Perfect. Um All right, I think that kind of I'm like uh I'm kind of I kind of doing a bit early really I'm in the time now. Uh Jesse, you have something to take or should I do uh something to close it right now? No. — we're going to take longer to do stuff over here. Oh, it's so good. Can we uh can we give Akshay like a little love, some like a little emoji reaction in the chat or on Zoom? Uh hoping a holler. Thank you so much. I think we open up the floor for just like some questions. There was a lot of really awesome stuff happening in the chat. Um or maybe if someone doesn't have a question, they want to come and share like something else maybe that you want to yes and like, "Hey, maybe folks should be thinking about this. " I know even some folks shared some amazing things in the chat like um Joy was talking about the podcast feature for scholarly articles. So, yeah, just opening up the floor in case folks want to ask a question or um share a way that uh something that they're doing. Yeah, Nia, what's up? Hello, hello. Hi, Akshay. Thank you for the session. Um my question is regarding the GPT. When you create one for your own, can you limit the access to GPT because it should not be available like on the web, right? For everyone to use. So, how do you manage that? Yes, you can access the you can So, you can either make it like this right okay, this way. Anyone with the link can open. People can't search for it and find it, right? That's only possible if you actually put it in their GPT store, right? If you go to explore GPTs, you'll find like hundreds of GPTs created. Those are the ones they have explicitly shared you know, designed to be used by others. But if you don't want it to be you know, open, you can only you know, keep it on that anyone with the link. That link is pretty complex. It's not some people can't really guess that link. It's like you know, in the YouTube you have the unlisted feature. Unless you have it is public, but you can only view it unless you have the URL. Same concept here. Yeah, got it. Thanks. Awesome, Risti. So, yeah, I just wanted to share and I had a question here. Share is that you know, to have the best prompts I actually asked ChatGPT to create a prompt for me and then push that prompt in ChatGPT or Claude. At the end of the day then I am not to be blamed. I can put okay, you know, they did something wrong. Other thing was you know

### Segment 12 (55:00 - 60:00) [55:00]

while I use myself and I see a lot others again, not talking about tech, but any people apart from the coding part. While they are using this ChatGPT, Claude or many things else, we are mostly using it for operational purposes. However, what I've seen is you know, tech guys at the senior level are also able to use it for strategy purposes. So, you know, what how can we differentiate the prompts or the usage you know, from operations to strategy irrespective of the fields we are you know, HR or admin, procurement, consulting, anything. Oh, okay. Awesome. So, a lot of thing comes down to the context that you give to the tool, right? And be be, you know, say be ChatGPT or Claude or Notebook LM. You would the choice to take whatever it gives out is something that you still want to, you know, take it with a pinch of salt, especially at the ELT level, right? You would not want to trust things and blindly. Uh and if the if the met if the data is there, uh let's say from the strategy standpoint, right? You at a strategy decision level, you would not be giving the entire set of data and then ask you to take a decision. You would have it broken down. You would have a pipeline that translate certain data to certain actionable insights. And that pipeline helps you to take the decisions from a strategy standpoint. Now, this comes This is slightly different, right? We are not generating content here. We are not transferring content from a PDF to a PPT. This is deducing something out of information that we already have. That's the strategic side of things, right? That's where you would have to have that right context set. multi-agent setup, you know, one agent breaks it down for you. One agent actually searches uh other resources and then takes out data. One agent actually summarizes everything for you. So, maybe that's the direction that you would want to head to if you want to use it use these models or these tools in that front. Hopefully that didn't, you know, confuse you even more. But long story short, it's going to be difficult to do it with just one tool because you need to do some form of translation as well, you understanding the data, deducing the key metrics out of it, and then taking actions on top of it. Uh Yeah. Awesome. So, the best strategy in HR isn't a tech strategy, right? It It's a human one. So, my final words would be is to start small, you know, start this week. And remember, AI gets you, you know, 70%. You are the irreplaceable 30%. Uh and uh you know, that's thank you for showing up today. Yes. Oh my gosh, thank you all so much. Um a few like closing things that I wanted to share with you all. One, um someone in our community here was shared an amazing idea with me, and basically was talking about doing a series where we could basically part of it we yes, we can learn new tools, but then we could also continue to do like show and tell sessions with like how we are actually using these AI tools in our workflows, especially as people and culture professionals. Um so, if you are like, yes, I'm I would really enjoy that. Um I'd love your feedback, and I'd love just like a little yeah, like please let me know if that would be interesting to do an ongoing series. I'm thinking about how we might keep this conversation going. I'm actually looking to start um an AI and HR topic-based chapter that like has this going on an ongoing series. So, if you are interested in either like co-leading that chapter, or if you were like, no, this is interesting, I want to be involved in that. Um I'd love if you would um maybe the call to action's like just send me a DM on LinkedIn. That would be the easiest way to just be like, yeah, that sounds interesting. Um so, I'll drop my LinkedIn in the chat for us to stay connected and you to kind of give me like a yes, I'm interested in either joining that conversation on an ongoing basis or actually leading that conversation. Um and then let's see um other stuff um we have an amazing session coming up in May around let's see if my slides will advance. Uh I'll call leading through uncertainty with author Simone Stolzoff. If some of you know him, this is a free session on Wednesday, May 13th. Oh my gosh, my

### Segment 13 (60:00 - 61:00) [1:00:00]

computer is running so slow so I don't know what you all What's happening? It's a free for all right now. Anyway, Simone has a new book out called how to not know. Um I had the chance to see him speak in September. He was amazing and so I'm so excited for him to collaborate with our community and to have an exclusive conversation with him the week that this new book is coming out. So if you're interested and like to join us for that, um you'll see more information about that. So All right. With that, it wouldn't be a Culture First gathering if we didn't do a clap out. So if you all would entertain me and um coming on camera for just a moment if you can and you're willing and you're able to uh and if you can come off mute, uh we're going to do a clap out. So what's going to happen is like Akshay is going to go 1 2 3, we're all going to clap and then this Zoom meeting's just going to shut off. That's what's going to happen here. So um thank you all for being here. We'll keep the conversation going. Cheers to co-creating a better world of work. Um Akshay, thank you so much. Will you count us to three and then we're going to we're all going to clap get out of here. On three. One, two, three.