View Recording: Next-Gen AI Agents – Comparing Semantic Kernel and Copilot Studio

Brian Haydin Hi everybody. 0:0:6.869 –> 0:0:9.709 Brian Haydin Welcome to concurrency’s webinar. 0:0:9.709 –> 0:0:11.109 Brian Haydin We’re gonna talk a little bit about. 0:0:12.659 –> 0:0:16.739 Brian Haydin Some new the next generation of AI, which is really a generic patterns and agents. 0:0:18.539 –> 0:0:32.499 Brian Haydin And in this webinar, what we’re gonna be talking about is comparing kind of a high code like code code, you know, version of this and semantic kernel and what you can do with low code in Copilot Studio. I’m Brian Hayden. 0:0:32.779 –> 0:0:35.819 Brian Haydin I’m a solution architect here at concurrency and. 0:0:36.349 –> 0:0:38.349 Brian Haydin I am joined with Ajay. 0:0:38.469 –> 0:0:40.989 Brian Haydin Ajay, if you want to introduce yourself. 0:0:41.929 –> 0:0:43.89 Ajay Ravi Yep. Oh, thanks, Brian. 0:0:43.249 –> 0:0:48.409 Ajay Ravi My name is Ajay and I’m a senior engineer at concurrency working mainly on the low code. 0:0:48.409 –> 0:0:51.729 Ajay Ravi No code set of things and studio and power platforms. 0:0:51.729 –> 0:0:54.489 Ajay Ravi So yeah, looking forward to this event. 0:0:55.659 –> 0:0:58.579 Brian Haydin Fantastic. Let’s move the next slide. 0:1:0.139 –> 0:1:2.299 Brian Haydin So like what is semantic kernel? 0:1:2.299 –> 0:1:4.499 Brian Haydin What is, you know, copilot studio? 0:1:5.99 –> 0:1:25.19 Brian Haydin So I’ll just give you the quick definition of the two. When we’re talking about semantic kernel, what we’re talking we’re we’re specifically talking about coding framework that is open source provided by Microsoft and helps us build and compose agents in copilot studio. 0:1:25.819 –> 0:1:26.459 Brian Haydin This is basically. 0:1:26.989 –> 0:1:30.709 Brian Haydin Doing automation using the copilot Studio framework. 0:1:31.109 –> 0:1:32.269 Brian Haydin So very low code. 0:1:32.469 –> 0:1:41.589 Brian Haydin Maybe some code Ajay can talk a little bit more about that, but let’s you know move on and talk a little bit a bit more about what semantic kernel is. 0:1:43.659 –> 0:1:45.819 Brian Haydin Uh oh. Domains to domains to consider. 0:1:47.819 –> 0:2:5.859 Brian Haydin Before we jump into this journey, let’s talk a little bit about AI and and how we’ve framed up this to our customers. On the left side, we have commodity and when we use AI, we’re talking about tools or licenses that you buy. 0:2:6.99 –> 0:2:12.819 Brian Haydin So think of like M365, copilot or ChatGPT. As you know commodity base things. 0:2:13.779 –> 0:2:21.579 Brian Haydin Their their Sku’s that you can just purchase and leverage AI with our customers though we’re often talking about mission driven opportunities and how to. 0:2:22.119 –> 0:2:27.679 Brian Haydin Create direct ROI you know for their customers by developing things that don’t exist in the market. 0:2:28.439 –> 0:2:35.79 Brian Haydin And So what we’re gonna be doing is spending our time basically in the mission driven and how we approach this the next slide. 0:2:37.849 –> 0:2:46.49 Brian Haydin This is a great way for us to kind of articulate what you can actually do, both in the commodity and the mission driven space. 0:2:46.489 –> 0:2:54.49 Brian Haydin So if you read this slide from top down, essentially you’re gonna see accuracy improve as you move through the stack. 0:2:54.409 –> 0:3:3.289 Brian Haydin So when you go to ChatGPT and ask questions or copilot, essentially you’re just getting answers that are grounded in the information that it was trained on. 0:3:4.99 –> 0:3:8.459 Brian Haydin But then you start moving into the commodity AI like the M365 copilot. 0:3:9.99 –> 0:3:13.779 Brian Haydin And it’s able to ground against data that you’re that your organization has. 0:3:14.19 –> 0:3:19.499 Brian Haydin So we’ve increased the accuracy because it knows more about you, but it’s still kind of a commodity tool set. 0:3:21.59 –> 0:3:36.339 Brian Haydin And it doesn’t really do any specialized workflows, but as we move down to like you know the semi custom area we’re building agents that have specific actions that we want to take or request that we needed you know to respond to. 0:3:37.979 –> 0:3:38.19 Brian Haydin And. 0:3:38.629 –> 0:3:50.709 Brian Haydin And then we can get even further into the, you know, custom ML area where we’re doing very focused, you know tools don’t really solve the problem. We may have to build new models for the organization. 0:3:51.229 –> 0:4:2.189 Brian Haydin So this just kind of gives you an idea of where you know where these projects kind of land in the spectrum and what we’re looking at is something that’s going to bridge between the semi custom and the custom ML. 0:4:7.489 –> 0:4:23.129 Brian Haydin So another reason, another way you would look at this is starting with the you know, why are we doing this or what? What value is a particular AI opportunity going to bring to the organization. And this is a value analysis that we do with our customers we’ll do. 0:4:23.129 –> 0:4:31.209 Brian Haydin An ideation session and then help to stack rank the effort that’s involved and what the expected ROI might be, but. 0:4:32.99 –> 0:4:36.979 Brian Haydin You always want to ask yourself when you get into these custom code or these semi custom areas like. 0:4:37.739 –> 0:4:39.219 Brian Haydin What is the value proposition? 0:4:39.219 –> 0:4:50.499 Brian Haydin Why am I doing this? And it might be OK just to say we wanna try something out and we know that this technology’s gonna work to, you know, to give us a vision of how this might work in more complex scenarios. 0:4:50.779 –> 0:4:52.339 Brian Haydin But this is a good way to start. 0:4:56.739 –> 0:5:15.659 Brian Haydin So what are agents? Most of us on this on this webinar have used you know AI ChatGPT and so also most of us have probably played around with chat bots that can start to use documents or information to answer the questions better. 0:5:15.859 –> 0:5:23.779 Brian Haydin This is like that retrieval augmentation generation like framework that has become a lot more popular and that’s how copilot M365 works. 0:5:24.539 –> 0:5:25.699 Brian Haydin Is using that kind of pattern. 0:5:26.679 –> 0:5:31.239 Brian Haydin But you know, as this evolves, it starts to take actions. 0:5:31.239 –> 0:5:35.959 Brian Haydin So now I can ask copilot to do things inside of my document. 0:5:36.279 –> 0:5:40.79 Brian Haydin I can ask it to add a slide that that compares two different topics. 0:5:40.119 –> 0:5:52.79 Brian Haydin I can ask it to go find information and bring it back for me, so those are some of the agentic things that we’re doing today, but we’re going to get into some more complex scenarios that get to closer to like a fully autonomous. 0:5:52.859 –> 0:5:56.459 Brian Haydin Where you can ask an agent, and I’ve actually seen demos of this over the last week. 0:5:57.109 –> 0:6:4.669 Brian Haydin Were you asking agent to go do something and it interacts with the external world in order to accomplish your goals? 0:6:7.659 –> 0:6:22.419 Brian Haydin Another way to look at this is how these agents have evolved, and if you go way back, you know when computing first started, you know, and the first things people wanted to do was save time and energy in the work that they did. 0:6:22.699 –> 0:6:24.419 Brian Haydin So they started doing basic automation. 0:6:24.419 –> 0:6:25.659 Brian Haydin They started writing scripts. 0:6:25.659 –> 0:6:32.419 Brian Haydin They started building, you know, small programs to be able to do some of the automation. And it became kind of cumbersome to have to redo. 0:6:33.299 –> 0:6:37.699 Brian Haydin To write code from the very beginning. So RPA tools you know started to take the place. 0:6:39.419 –> 0:6:41.779 Brian Haydin Power automate. It’s been a very successful tool. 0:6:41.779 –> 0:6:45.59 Brian Haydin There’s UI path, you know, bunch of other tools in the RPA world. 0:6:46.619 –> 0:6:54.19 Brian Haydin And recently, about two years ago, we started bringing in like human, you know, interactions like natural language processing. 0:6:54.19 –> 0:7:7.219 Brian Haydin And that’s where the chatbots you know became really effective because it’s not just about setting up these workflow automations, but it’s asking you know, a system on how to what I accomplished this or getting information. 0:7:7.699 –> 0:7:12.539 Brian Haydin And now we’re in this place of, like, doing, you know, Agentic work that’s, you know, where we’re at right now. 0:7:13.339 –> 0:7:18.299 Brian Haydin And there’s a couple of different concepts that like a couple different ways that that we’ve broken this up. 0:7:18.299 –> 0:7:19.619 Brian Haydin We started with just LLM. 0:7:21.499 –> 0:7:32.299 Brian Haydin And, you know, was very difficult to work with some of the documents because they’re, you know, very image oriented. You know, think of like your IKEA manuals on how to install or how to build a a desk. 0:7:33.299 –> 0:7:36.819 Brian Haydin There’s more pictures than there are words, but you know you and I can look at that. 0:7:36.819 –> 0:7:38.139 Brian Haydin How do we get this to work? 0:7:38.699 –> 0:7:44.219 Brian Haydin The multimodal models are starting to be able to interpret that information and be able to give you. 0:7:45.129 –> 0:7:46.409 Brian Haydin You know, answer the question. 0:7:46.409 –> 0:7:48.929 Brian Haydin What is the step by step on? How to you know, build this desk? 0:7:50.499 –> 0:7:56.139 Brian Haydin But those are like, you know, just singular agents, you know, reading a document, you know, trying to frame it up. 0:7:56.139 –> 0:7:58.179 Brian Haydin And now we’re, you know, starting to get to is. 0:7:58.179 –> 0:8:10.819 Brian Haydin How do I take you know that and break it into more complex tasks and say go read this document, go get this other piece of information. Go bring this all together. Do some planning activities and and give me a result. 0:8:16.199 –> 0:8:29.919 Brian Haydin So when you think about developing a multi agent, you know, architecture, you’re gonna probably evolve, you know, from the chatbot down into more complex actions using something like copilot studio. 0:8:30.159 –> 0:8:38.599 Brian Haydin Maybe you’re using semantic kernel, but it’s also kind of important to like. Think about how you’re going to scale this. 0:8:38.599 –> 0:8:44.359 Brian Haydin You know for your organization and that’s where when we get into custom ML that we start bringing in. 0:8:45.259 –> 0:8:48.939 Brian Haydin Conversations like ml OPS like we would do DevOps in any other engineering project. 0:8:50.199 –> 0:8:57.959 Brian Haydin So this is more of like a maturity of a way of framing up the maturity for your organization where you’re gonna need to like, think ahead. 0:8:59.859 –> 0:9:2.299 Brian Haydin And and we’ll talk a little bit more about that. 0:9:6.389 –> 0:9:8.69 Brian Haydin So what is Samantha currently gave you? 0:9:8.69 –> 0:9:17.509 Brian Haydin Like the high level definition semantic kernel at the end of the day it’s an open source framework provided by Microsoft. It works. 0:9:18.509 –> 0:9:19.349 Brian Haydin It works with. 0:9:20.899 –> 0:9:22.779 Brian Haydin C# developers. It works with Python. 0:9:22.779 –> 0:9:24.19 Brian Haydin It even works with Java. 0:9:25.739 –> 0:9:34.259 Brian Haydin And the framework itself is to be able to to create agents that understand how to pull the different pieces of information together. 0:9:34.549 –> 0:9:46.869 Brian Haydin How to chain agents together so that they can that they can perform more complex tasks and do that by using natural language in reasoning built into the framework? 0:9:53.959 –> 0:9:56.399 Brian Haydin So how this is actually used? 0:9:56.399 –> 0:10:12.419 Brian Haydin It’s it basically a framework that you import into a library. If you wanna think of it that way that you import into your application and then as it’s you know as part of its planning activities, it’s going to reach out to plug insurance. We’ll get into someone with. 0:10:12.419 –> 0:10:15.879 Brian Haydin These these definitions are in other AI applications. 0:10:16.79 –> 0:10:18.319 Brian Haydin I did mention that language, you know. 0:10:18.469 –> 0:10:20.829 Brian Haydin It supported were multiple languages. 0:10:20.989 –> 0:10:32.669 Brian Haydin It also can be used outside of the Microsoft ecosystem for other AI applications, so I would work with, you know, GROK or Google’s, you know, any of the other general models as well. 0:10:34.639 –> 0:10:39.959 Brian Haydin On the next slide, let’s let’s you know, talk about agentic patterns. 0:10:41.519 –> 0:10:43.599 Brian Haydin So what do we mean by agentic patterns? 0:10:43.999 –> 0:10:53.479 Brian Haydin So these are like really complex personal assistants that are going to accumulate the knowledge that it needs needs in order to answer the answer. 0:10:53.479 –> 0:11:3.559 Brian Haydin Your request so good way to think about this. We’ve all seen like the hallucination, you know, kind of aspect of llm’s they only have a limited amount of knowledge I can ask. 0:11:4.449 –> 0:11:6.369 Brian Haydin You know, I can ask opening. 0:11:7.959 –> 0:11:16.39 Brian Haydin Eye model 01 for, you know, piece of information that was that was built after or that was knowledge that was created after it was built. 0:11:16.39 –> 0:11:16.679 Brian Haydin I don’t know. 0:11:17.119 –> 0:11:19.799 Brian Haydin Just tell me. I I stopped in June of 2023. 0:11:19.799 –> 0:11:25.239 Brian Haydin I don’t know what that is or I can’t read the weather because it’s not what I was trained on. 0:11:25.239 –> 0:11:32.839 Brian Haydin I don’t have that knowledge, so agentic patterns are ones that are going to go out and retrieve additional information, maybe process. 0:11:33.799 –> 0:11:36.959 Brian Haydin You know, using another LLM to do some other work in the bring it all together. 0:11:44.349 –> 0:11:44.829 Brian Haydin Next slide. 0:11:46.959 –> 0:11:47.279 Brian Haydin OK. 0:11:47.279 –> 0:11:57.319 Brian Haydin So actually I’m sorry. Go back to the so more specifically, what agents do they take actions they have memory, which is a very important concept. 0:11:57.799 –> 0:12:13.759 Brian Haydin They can remember the request that you asked so that when you give it additional feedback, it can refine its its responses. A very big, important part of what Samantha Colonel’s going to do is it plans. So as you register plug insurance and agents, you give it inst. 0:12:14.199 –> 0:12:15.39 Brian Haydin This is what this. 0:12:15.649 –> 0:12:23.249 Brian Haydin This portion is able to able to do and then it orchestrates all those different actions together to to drive towards your results. 0:12:25.829 –> 0:12:30.429 Brian Haydin And for those of you that are thinking that there’s gonna be code, this is the most code you’re gonna see today. 0:12:31.999 –> 0:12:33.519 Brian Haydin So and there isn’t much. 0:12:33.799 –> 0:12:35.959 Brian Haydin It’s, you know, one single line. 0:12:36.599 –> 0:12:45.159 Brian Haydin But what is semantic kernel like? If you’re a developer, that’s gonna be using this in Visual Studio. At the heart of it, it’s a dependency injection container. 0:12:45.479 –> 0:12:48.439 Brian Haydin So what you’re gonna do is register. 0:12:48.999 –> 0:12:51.599 Brian Haydin You’re gonna register your agents to different components. 0:12:51.599 –> 0:12:53.239 Brian Haydin Those might be APIs. 0:12:53.559 –> 0:12:55.439 Brian Haydin Those might be other LLM models. 0:12:55.439 –> 0:12:57.639 Brian Haydin Those might be other agents that you registering. 0:12:58.829 –> 0:13:4.589 Brian Haydin And then give it instructions around the planning so that it knows how to how to respond to it. 0:13:5.229 –> 0:13:6.269 Brian Haydin What is it gonna do? 0:13:6.269 –> 0:13:9.389 Brian Haydin It’s basically gonna select the optimal path that’s gonna take. 0:13:10.29 –> 0:13:12.909 Brian Haydin It’s gonna process the natural language of your request. 0:13:13.69 –> 0:13:16.789 Brian Haydin It’s gonna figure out which one of the agents probably have relevant information. 0:13:16.989 –> 0:13:21.229 Brian Haydin It’s gonna build kind of the plan of how to get you to your final destination. 0:13:22.799 –> 0:13:25.279 Brian Haydin And then return, you know return the results to you. 0:13:31.339 –> 0:13:39.579 Brian Haydin There’s several different components or bunch of different components that kinda listed them here, so I’ll describe them. The root of it. You’ve got this kernel right? 0:13:41.119 –> 0:13:48.679 Brian Haydin And the kernel is all that dependency injection you know planning, you know, aspect of it and what you’re going to register to it. 0:13:48.999 –> 0:13:51.279 Brian Haydin Are you going to register different personas to it? 0:13:51.279 –> 0:13:54.279 Brian Haydin You’re going to define who the users that are interacting with it. 0:13:54.439 –> 0:13:57.839 Brian Haydin Think of this is kind of like your prompt engineering aspect of it. 0:13:57.839 –> 0:13:59.959 Brian Haydin You’re going to add AI services. 0:14:0.899 –> 0:14:4.499 Brian Haydin Different models that you can bring in through a variety of different services. 0:14:6.39 –> 0:14:8.999 Brian Haydin Plugins we you know, I’ve used that word a couple of times. 0:14:9.319 –> 0:14:10.599 Brian Haydin Plugins could be. 0:14:11.679 –> 0:14:21.79 Brian Haydin You know any any way for you to retrieve data or information in order to return it to the planner in the kernel to figure it out? 0:14:21.79 –> 0:14:31.999 Brian Haydin So it’s typically APIs that you’re registering and then giving it some sort of declarative. This is what I do. So for example, if I wanted to, you know, grab the weather. 0:14:32.819 –> 0:14:40.339 Brian Haydin This is an API that you would specify that this is an API that returns weather information for a given longitude and latitude. 0:14:42.239 –> 0:14:47.39 Brian Haydin Then we have agents which also can kind of like be considered plug insurance. 0:14:47.39 –> 0:14:48.519 Brian Haydin They’re very tightly related. 0:14:50.199 –> 0:14:55.599 Brian Haydin But those are other agents that you built that can be consumed almost as if it was an API. 0:14:56.79 –> 0:14:58.759 Brian Haydin And then finally, we have to have a way to store information. 0:14:58.759 –> 0:15:2.679 Brian Haydin Sometimes this is memory. Sometimes this is retrieval augmentation generation. 0:15:3.589 –> 0:15:8.869 Brian Haydin But the end of the day we have vector stores that help us to find more detailed information. 0:15:13.919 –> 0:15:23.559 Brian Haydin So let’s talk about a real world example that some of us might experience. And how do we pull this all together to make a semantic kernel agent? 0:15:23.999 –> 0:15:34.759 Brian Haydin So if you are going to walk into a store and go camping and you need to get a new tent, you’re going to interact with an, you know, somebody at the store. 0:15:34.759 –> 0:15:39.319 Brian Haydin Maybe it’s Rei and say I’m looking to go, you know, camping out in Colorado. 0:15:39.319 –> 0:15:40.639 Brian Haydin Like what kind of tent should I get? 0:15:41.399 –> 0:15:45.119 Brian Haydin You’re going to interact with a human, and they’re going to probably ask you a bunch of questions. 0:15:45.119 –> 0:15:47.639 Brian Haydin They’re going to pull on the knowledge that they have in their head. 0:15:48.639 –> 0:15:58.639 Brian Haydin And Samantha? Colonel. You know, one way that you could look at this is like, what are the steps that you would have to go through in order to answer that question? If you were working in the store? 0:15:59.239 –> 0:16:4.719 Brian Haydin So you know, first thing you need to do is kind of who am I talking to define this persona? 0:16:4.719 –> 0:16:5.879 Brian Haydin How am I interacting with them? 0:16:7.439 –> 0:16:13.679 Brian Haydin And then what is the information that I need to get from that individual in order to give them an informed response? 0:16:14.159 –> 0:16:17.679 Brian Haydin So we’re going to, we’re going to talk to you know. 0:16:18.469 –> 0:16:30.69 Brian Haydin Our computer and say what’s available in the system today because I’m they’re here in the store. If I can’t sell it to them, or if I can’t bring it to them, then it’s probably not something I wanna like have a conversation about. 0:16:30.789 –> 0:16:33.309 Brian Haydin You might pull up, you know, pull into the conversation. 0:16:33.309 –> 0:16:38.509 Brian Haydin Social media reviews this so you’ll have a social media agent that you would need to define. 0:16:40.79 –> 0:16:49.279 Brian Haydin You probably have sold some of these products before, so you’re gonna give them other customers feedback and reviews. And then another thing you might wanna bring to the table. 0:16:49.649 –> 0:16:50.889 Brian Haydin What have you used before? 0:16:50.889 –> 0:16:52.689 Brian Haydin What tents have you used? 0:16:52.689 –> 0:16:58.289 Brian Haydin You know, you know on other expeditions or have you been camping and been in other people’s tents? 0:16:58.289 –> 0:16:59.449 Brian Haydin Like what do you like? 0:16:59.649 –> 0:17:0.569 Brian Haydin Are you in, you know? 0:17:0.569 –> 0:17:12.89 Brian Haydin And so historical context is something that you would bring back as well and you pull all that knowledge together as an individual and now you can give them an informed, an informed response. So. 0:17:13.639 –> 0:17:16.79 Brian Haydin Essentially, this is how you would build this in Samantha kernel. 0:17:21.309 –> 0:17:25.669 Brian Haydin So now let’s talk a little bit about copilot studio and here I’ll turn it over to Ajay. 0:17:27.509 –> 0:17:28.949 Ajay Ravi Thanks Brian for that. 0:17:30.499 –> 0:17:35.979 Ajay Ravi So we’re gonna do like switch gears a little bit and talk a little bit about all the low code. No code side of things. 0:17:36.339 –> 0:17:49.619 Ajay Ravi So Microsoft introduced copilot studio back into 2023, Ignite as a local platform where you can build intelligent chat bots or even exchange to Microsoft 365 Copilot as well. 0:17:50.59 –> 0:17:54.739 Ajay Ravi So it is a conversational chatbot where you can kind of like. 0:17:55.699 –> 0:18:9.339 Ajay Ravi Provide multiple knowledge sources and then allow the copilot studio to kind of like interact with that shared knowledge source and then provide results. And in addition to that you can also automate a lot of your process and make sure that. 0:18:9.869 –> 0:18:12.189 Ajay Ravi Copilot Studio is doing a bunch of things. 0:18:12.189 –> 0:18:16.109 Ajay Ravi Oh, instead of like you having to manually do a lot of things so. 0:18:17.659 –> 0:18:35.499 Ajay Ravi And generative AI feature is being used in the copilot studio. So before last year there was power virtual agent which was previously available under power platform. So it did not have like the generative AI capabilities. So you had to like manually create a lot of different topics and. 0:18:35.499 –> 0:18:44.779 Ajay Ravi Can conversational pattern, so that was kind of a pain point in that scenario. But with the inclusion of the whole generative AI and the Copil studio. 0:18:45.859 –> 0:18:46.659 Ajay Ravi AI functionality. 0:18:46.659 –> 0:18:48.659 Ajay Ravi So now or you can easily. 0:18:49.429 –> 0:18:55.269 Ajay Ravi Connect your different knowledge sources like your business, know data and everything to your copilot studio. 0:18:55.269 –> 0:18:56.949 Ajay Ravi Analyt the copilot studio. 0:18:56.949 –> 0:19:0.29 Ajay Ravi Interact with your data and then get the appropriate result. 0:19:0.509 –> 0:19:16.29 Ajay Ravi So it is a low code platform and you can easily use the graphical interface in order to build a conversational pattern or in addition to that, if you can connect to your knowledge sources, it will provide those results as well. 0:19:16.819 –> 0:19:17.179 Ajay Ravi And. 0:19:19.59 –> 0:19:27.739 Ajay Ravi So you can connect your copilot studio with different data sources like in CRM application or a work day or like different platforms. 0:19:27.739 –> 0:19:33.139 Ajay Ravi So you can definitely connect Copilage studio and automate a bunch of app process as well. 0:19:34.49 –> 0:19:48.569 Ajay Ravi And since there is bell on top of our platform and then some, our connectors are available. There are currently over 1100 plus prebuilt connectors available and if you want to like, you know do additional HTTP calls and things like that. 0:19:48.659 –> 0:20:4.379 Ajay Ravi You can easily do that as well from studio, so it’s a low code platform and it allows you to, you know, do simple simple chat board using few clicks and then also if you want to like automate your process you can do that as well we’ll be taking. 0:20:4.499 –> 0:20:8.979 Ajay Ravi Look at some of the use cases and we’ll do a quick demo as well. Little bit later today. 0:20:9.339 –> 0:20:13.299 Ajay Ravi So yeah, that’s a little bit about the copilot studio. 0:20:14.859 –> 0:20:17.979 Ajay Ravi So let’s take a quick look at the architecture. 0:20:18.179 –> 0:20:26.619 Ajay Ravi I won’t deep dive into the the whole details, but I just want to like give you guys a little bit overview of what are some of the different things that are happen in the background. 0:20:27.99 –> 0:20:31.579 Ajay Ravi So Microsoft enter ID security is definitely the most important thing, right? 0:20:31.579 –> 0:20:40.19 Ajay Ravi So you can use copilot studio and then deploy it for an internal employees or even for external customers as well. 0:20:40.219 –> 0:20:44.819 Ajay Ravi So you can do it in an authenticated way or a non authenticated way depending on where. 0:20:45.309 –> 0:20:48.829 Ajay Ravi All story or data sources and then your knowledge sources. 0:20:48.829 –> 0:21:3.29 Ajay Ravi So depending on that you can definitely. You definitely have bunch of different options for the authentication side of things and then if you want to add additional security there is definitely options for doing that. 0:21:3.229 –> 0:21:7.549 Ajay Ravi So like I mentioned, since it is Bill on top of power platform environment. 0:21:7.549 –> 0:21:14.149 Ajay Ravi So there are definitely data loss prevention policies and other additional security that are in built in power platform. 0:21:14.149 –> 0:21:15.109 Ajay Ravi So you do have the. 0:21:15.419 –> 0:21:20.179 Ajay Ravi To restrict certain HTTP calls or like you know, certain sections. 0:21:20.179 –> 0:21:25.699 Ajay Ravi So you do have the ability to do that at the copilot level or even at the tenant level as well. 0:21:26.139 –> 0:21:36.419 Ajay Ravi So there are in build securities available there and the next thing is Gopala Studio uses BOT framework skill in the background. 0:21:37.19 –> 0:21:39.859 Ajay Ravi So when you are spin up in your copilot studio. 0:21:40.659 –> 0:21:49.659 Ajay Ravi It automatically comes with few entities, so entities are basically how copilot get to know what the user is talking about. Like you know, the place things, that kind of thing. 0:21:49.659 –> 0:21:58.979 Ajay Ravi So when you are spin up a new copilot studio, it automatically comes with some bunch of predefined entities and if you would like to build additional custom. 0:21:59.899 –> 0:22:0.859 Ajay Ravi You know entities. 0:22:0.859 –> 0:22:4.779 Ajay Ravi So you do have the ability to do that in a low code, no code base. 0:22:4.779 –> 0:22:15.299 Ajay Ravi So there are bunch of options there. So for example if you want to build ECHO Pilot Studio for an IT help desk and you know that the ServiceNow ticket number is pretty much the same. 0:22:15.299 –> 0:22:19.19 Ajay Ravi So you can kind of like design those custom entities and things like that. 0:22:19.19 –> 0:22:25.979 Ajay Ravi So that way whenever you start types in a pattern of or number, so it gets to know that the usage talking about. 0:22:26.779 –> 0:22:27.699 Ajay Ravi ServiceNow ticket numbers. 0:22:27.699 –> 0:22:29.539 Ajay Ravi So this is an example. 0:22:29.539 –> 0:22:31.379 Ajay Ravi So you do have the ability to create. 0:22:32.209 –> 0:22:53.89 Ajay Ravi Additional custom you know entities as well and in addition to that also I’ll like I mentioned so generative AI functionality was interested last year and before that there was no generative answer but you had to like build custom logic for all the conversation pattern. But with generative A. 0:22:53.89 –> 0:22:56.369 Ajay Ravi So you still have the ability to disable that functionality. 0:22:57.179 –> 0:23:8.619 Ajay Ravi But when that is being enabled so it pretty much uses the Azure open AI in the back end to look at the different knowledge sources that you have provided and then accordingly provide the appropriate response so. 0:23:9.319 –> 0:23:19.959 Ajay Ravi Azure Open AI is being used at the back end for some of those and in addition to that, if you want to build custom logic specific to your business without are using some of the Gen. AI functionality. 0:23:19.959 –> 0:23:29.39 Ajay Ravi So you do have the ability to do that as well. So you can design the conversation pattern like you know how the question should be structured, what kind of questions. 0:23:29.39 –> 0:23:32.839 Ajay Ravi The bot should ask before providing the service, right? 0:23:32.839 –> 0:23:35.279 Ajay Ravi So you do have the ability to do that as well. 0:23:36.59 –> 0:23:44.139 Ajay Ravi And you can connect your copilot studio with power automate, which is a pretty powerful tool with over 1100 plus out-of-the-box connectors. 0:23:44.139 –> 0:23:47.499 Ajay Ravi So you can use power automate. 0:23:47.499 –> 0:23:51.859 Ajay Ravi You can call a power automate from Copilot studio to get some of the details and then. 0:23:52.849 –> 0:23:57.649 Ajay Ravi Provide the result back to the user, so there are some powerful capabilities that as well. 0:23:58.169 –> 0:24:17.49 Ajay Ravi And then since it is built on top of power platform, so you do have the ability to build and then safely deploy it to test and production environment relatively quickly, all within all the power platform, our infrastructure as well and at any point of time if you want. 0:24:17.49 –> 0:24:18.329 Ajay Ravi To escalate to. 0:24:19.139 –> 0:24:22.219 Ajay Ravi A different agent or like an actual a real agents to. 0:24:22.799 –> 0:24:30.519 Ajay Ravi There are functional advanced functionalities where you can connect with Microsoft Dynamics 365 or other call center options. 0:24:30.599 –> 0:24:38.879 Ajay Ravi So that way you know it gets escalated and the user gets all the different conversation, you know, questions and everything. 0:24:38.879 –> 0:24:43.999 Ajay Ravi So there are definitely capabilities to do that as well from copilot studio. 0:24:46.519 –> 0:24:50.519 Ajay Ravi And there are analytics as well available in copilot studio. 0:24:50.519 –> 0:24:53.999 Ajay Ravi So you get to know how the conversation is going. 0:24:53.999 –> 0:24:55.679 Ajay Ravi What is the customer satisfaction? 0:24:55.679 –> 0:25:0.319 Ajay Ravi How the questions and navigation is going so there are some additional capabilities there as well. 0:25:0.319 –> 0:25:5.679 Ajay Ravi So based on some of those analytics, you can kind of like iterate and improve your copilot as well. 0:25:5.679 –> 0:25:10.639 Ajay Ravi So it’s a pretty and everything is in low code, no code off fashion. 0:25:10.639 –> 0:25:12.199 Ajay Ravi So if you would like to do. 0:25:12.979 –> 0:25:29.499 Ajay Ravi A code programming set of things you can definitely do that, but with this particular Gopal STD you can pretty much build intelligent copilots that allows you to do certain things for you and also get response from you know FAQ document or like different websites as well. 0:25:33.719 –> 0:25:38.999 Ajay Ravi So now let’s take a quick look at how the generative AI is changing the whole building. 0:25:38.999 –> 0:25:42.879 Ajay Ravi So traditionally, if you look at some examples like an Azure. 0:25:45.19 –> 0:25:52.419 Ajay Ravi Bot service so you had to like spend a lot of time manually creating the topics and navigation patterns. 0:25:52.419 –> 0:26:3.619 Ajay Ravi So that was one of the drawback and because of that there were a lot of missed opportunities where the copilot or like, you know, the chatbot was not able to answer some of the details and. 0:26:4.819 –> 0:26:9.219 Ajay Ravi And missed conversation or getting escalated to an actual agent, right? 0:26:9.579 –> 0:26:21.619 Ajay Ravi So that was one of the drawback before generative AI. And then every time you change the knowledge content you have to like manually spend a lot of ours and then update the copilot as well in code. 0:26:21.619 –> 0:26:27.699 Ajay Ravi So those were some of the drawbacks previously, before the whole generative AI was interested, right? 0:26:27.979 –> 0:26:34.99 Ajay Ravi So you had to like manually author a lot of topics resulting in mesh opportunities as well in some scenarios. 0:26:35.209 –> 0:26:43.409 Ajay Ravi But with the inclusion of generative AI and our functionalities and copilot studio, there are a bunch of different benefits. 0:26:43.929 –> 0:26:50.529 Ajay Ravi So now you don’t have to manually create all the different logics and topics in Copilace studio. 0:26:52.99 –> 0:27:5.139 Ajay Ravi Instead, what you can do is you can easily connect to different data sources or knowledge sources that are available or even API calls, right? So and the generative AI will be able to answer all based on the user questions. 0:27:5.379 –> 0:27:16.339 Ajay Ravi So you can easily connect to your public websites or if it’s for an internal scenario so you can easily connect your SharePoint site where you have added a bunch of different. 0:27:17.359 –> 0:27:20.79 Ajay Ravi Documents related to HR or bunch of different things, right? 0:27:20.79 –> 0:27:31.359 Ajay Ravi So you can easily connect your knowledge to assist your copilot studio and then allow the copilot and the back end LLM model to kind of like look at those documents and provide the results. 0:27:31.359 –> 0:27:47.239 Ajay Ravi So with the inclusion of the whole generative AI integr, it easily allows you to use those knowledge sources to provide answers which will result in less escalation and not less. You know, questions being unans. 0:27:47.469 –> 0:27:48.669 Ajay Ravi Right, so there are. 0:27:48.669 –> 0:28:4.139 Ajay Ravi So the generative AI is really changing that aspect, and then you also don’t need all like in advance data science person or like in a developer to kind of like build this because all these can be done in a low code, no code fashion. So it is easy. 0:28:4.139 –> 0:28:6.229 Ajay Ravi To build, maintain and enhance as well. 0:28:10.529 –> 0:28:16.209 Ajay Ravi So just say let’s take a quick look at the generative answers process and some of the consideration. 0:28:16.489 –> 0:28:29.729 Ajay Ravi So whenever user starts typing something to the copilot studio so the query or rewriting happens where it will kind of like or rephrase the question or let you know item which user have mentioned in the chatbot. 0:28:29.729 –> 0:28:35.249 Ajay Ravi So it will pretty much look at use query rewriting oppressors in the back end and then. 0:28:36.59 –> 0:28:52.19 Ajay Ravi What it will do is it will look at to the content retrieval where it will check the different knowledge users that you have added to your Copal list studio so it can be a public data where you have added bunch of different public websites to your copal studio. 0:28:52.19 –> 0:28:52.459 Ajay Ravi It will. 0:28:53.99 –> 0:28:53.779 Ajay Ravi You can do that. 0:28:53.779 –> 0:28:58.619 Ajay Ravi And then the copilot will look at the data sources and try to generate a response. 0:28:59.219 –> 0:29:4.459 Ajay Ravi It can be like a SharePoint SharePoint for internal scenarios where all you have connected your SharePoint. 0:29:5.459 –> 0:29:5.979 Ajay Ravi Site to your. 0:29:7.89 –> 0:29:13.969 Ajay Ravi Copilot studio and then it will look at the different documents that you have added in that SharePoint to generate response. 0:29:14.529 –> 0:29:32.249 Ajay Ravi You can also all use some advanced functionalities where you can connect your Azure open AI as well. So you can use your personalized model as well in there. So there are functionalities for that and in addition to that if you want to upload documents directly to your cop. 0:29:32.249 –> 0:29:34.9 Ajay Ravi Studio so you can do that as well so. 0:29:34.819 –> 0:29:44.979 Ajay Ravi Previously there was a limit of just three MB per document, so that was recently increased to 512 MB. So that gives a lot of different capabilities as well, right? So. 0:29:46.819 –> 0:29:53.859 Ajay Ravi Microsoft recently increased that and the file will be securely stored in DATAVERSE, which is in power platform as well. 0:29:54.789 –> 0:30:6.309 Ajay Ravi And if you want to use some custom data, you can definitely use some HTTP calls or like a power automate to get and fetch some of the information from different knowledge sources as well. 0:30:6.799 –> 0:30:25.429 Ajay Ravi So pretty much after the query rewriting happens, it will look at the different knowledge losses that are available and then accordingly do the summarization or make sure that you know some of the has been removed and some of the harmful and the malicious content has been removed and. 0:30:25.429 –> 0:30:27.119 Ajay Ravi Then accordingly provide the result. 0:30:27.439 –> 0:30:32.799 Ajay Ravi So this is how it happens in the back end and everything is done by Azure open AI. 0:30:32.999 –> 0:30:35.999 Ajay Ravi So you don’t have to manually make any changes there so. 0:30:37.9 –> 0:30:38.289 Ajay Ravi Hello code no code pattern. 0:30:42.199 –> 0:30:54.79 Ajay Ravi So just going over some of that top use cases, so you can, you know, use the scope LS studio for your internal organization as a for AB to B or B to C as well. 0:30:54.79 –> 0:31:9.399 Ajay Ravi So there are a bunch of our advantages of using these copilot studio. So for internal HR chatbot is one of the most common use cases which we have seen where you can do a bunch of things like employee can ask for end questions about their live policy or. 0:31:9.399 –> 0:31:11.319 Ajay Ravi Like you know about all the company policy, right? 0:31:11.869 –> 0:31:20.269 Ajay Ravi And in addition to that, if an HR person wants to onboard new user or automate a bunch of different things so you can do that as well. 0:31:20.269 –> 0:31:24.149 Ajay Ravi So we’ll be taking a quick look at some examples shortly, but. 0:31:25.709 –> 0:31:35.749 Ajay Ravi You can pretty much use your copilot studio for your internal employees or for B2BB to C like offer as a external facing customer or chatbot. 0:31:35.749 –> 0:31:37.789 Ajay Ravi So you can definitely build that as well. 0:31:38.229 –> 0:31:45.749 Ajay Ravi And Microsoft is investing heavily on Copa, listed with lot of new features coming up pretty every now and then. 0:31:45.749 –> 0:31:48.909 Ajay Ravi So all new functionalities are definitely getting added as well. 0:31:53.899 –> 0:31:59.259 Ajay Ravi So we’ll take a quick look at an example of HR chatbot. 0:31:59.259 –> 0:32:3.899 Ajay Ravi So what we have done is we have connected copilot studio with few different places. 0:32:3.899 –> 0:32:19.459 Ajay Ravi One is like Microsoft to do where you can kind of like automate and create some personal tasks and in addition to that we have connected to dynamics 365 where you can all start onboarding process for an employee or pretty much you know send off a letter. 0:32:19.539 –> 0:32:21.219 Ajay Ravi Training materials, all those things. So. 0:32:22.109 –> 0:32:23.309 Ajay Ravi Within few clicks and then few. 0:32:23.809 –> 0:32:32.569 Ajay Ravi Messages you can hard to that directly from Copalis studio that will allow you to, you know, not go to multiple places for doing all these things and then for. 0:32:34.109 –> 0:32:41.709 Ajay Ravi An employees type, so you can ask questions about HR queries, PTO balance, leave application, that kind of thing as well from Gopal studio. 0:32:43.389 –> 0:32:46.709 Ajay Ravi So I am going to bring. 0:32:49.489 –> 0:32:51.129 Ajay Ravi No go palette. 0:32:51.689 –> 0:32:56.249 Ajay Ravi So you can see that this is pretty much in teams. 0:32:56.249 –> 0:33:1.889 Ajay Ravi So what I’ve done here is I have deployed the HR chat board in my teams. 0:33:2.169 –> 0:33:15.529 Ajay Ravi So, similar to how you’re kind of like chatting with different users, so you can kind of like, you know, use this particular chat to interact with the copilot, you know, ask questions, automate certain things as well. So. 0:33:16.309 –> 0:33:19.589 Ajay Ravi You do have the ability to deploy safely to your team’s channel. 0:33:20.399 –> 0:33:21.359 Ajay Ravi So what? 0:33:21.359 –> 0:33:25.239 Ajay Ravi I’m going to do and teams is not the only place where you can deploy. 0:33:25.239 –> 0:33:32.239 Ajay Ravi You can deploy to your public websites or like you know customer service. All those scenarios can definitely be handled as well. 0:33:33.79 –> 0:33:40.759 Ajay Ravi So what I’m going to do is I’m going to start a conversation with the copilot and ask few things. 0:33:41.79 –> 0:33:46.479 Ajay Ravi So the first thing which which I’m going to do is do a little bit around the automation side of things, so. 0:33:47.269 –> 0:33:50.269 Ajay Ravi I want to like to something like. 0:33:55.709 –> 0:34:3.509 Ajay Ravi So you can see that pretty much provided a couple of details and it’s asking for the employee name. So I’m going to provide the employee detail. 0:34:3.509 –> 0:34:9.389 Ajay Ravi I’m just going to onboard myself in this scenario, so now you can see that it pretty much mentioned that you know. 0:34:9.389 –> 0:34:22.669 Ajay Ravi Thank you for your confirmation and onboarding has been started and alert has been sent for manager approval so you can see that within just few clicks you have actually started and done an automation and then also created. 0:34:23.629 –> 0:34:30.269 Ajay Ravi And how your dynamics 365 instance, So what this does is let me just. 0:34:43.689 –> 0:34:43.849 Ajay Ravi OK. 0:34:43.849 –> 0:34:53.169 Ajay Ravi So if I so this is an example of a development so you can see that all the record which I just onboarded so you can see the new record got created and then. 0:34:54.709 –> 0:34:58.69 Ajay Ravi It provided automatically created an application ID as well. 0:34:58.349 –> 0:35:4.829 Ajay Ravi So you can see that the copilot was able to create and then return all the details with. You know the ID as well. 0:35:4.829 –> 0:35:10.789 Ajay Ravi So within few clicks you can start some of these onboarding process from copilot studio. 0:35:11.69 –> 0:35:14.389 Ajay Ravi So it also like you know, do you need help with anything else? 0:35:14.389 –> 0:35:18.349 Ajay Ravi So I’m just going to click yes and start asking few other things. 0:35:18.349 –> 0:35:20.829 Ajay Ravi So what I’m going to do is. 0:35:21.709 –> 0:35:25.589 Ajay Ravi I am going to ask a question. Get onboarding status. 0:35:28.279 –> 0:35:33.79 Ajay Ravi So you can see that it pretty much replied. I can help you with the application detail. 0:35:33.79 –> 0:35:35.359 Ajay Ravi Could you help me with the employee ID? 0:35:35.359 –> 0:35:41.239 Ajay Ravi So what I’m going to do is I’m going to type in number which is in our system. 0:35:46.199 –> 0:35:59.279 Ajay Ravi So you can see that all the status is exchange rate, global offers. So and if you look at the system here you can see that all the application sheet is, you know exchange rate offer. 0:35:59.519 –> 0:36:9.879 Ajay Ravi So this way you can interact with, you know your business data in various systems to get some of these information without having to navigate to different locations as well. 0:36:10.439 –> 0:36:12.79 Ajay Ravi So I’m going to continue. 0:36:12.869 –> 0:36:15.469 Ajay Ravi Chatting so the next thing which I’m going to do is get. 0:36:17.89 –> 0:36:18.529 Ajay Ravi They’re on boarding. 0:36:21.439 –> 0:36:22.639 Ajay Ravi I’m just going to. 0:36:25.699 –> 0:36:26.19 Ajay Ravi OK. 0:36:26.99 –> 0:36:30.699 Ajay Ravi So now or what I want is the list of different records that have been assigned to me. 0:36:31.19 –> 0:36:37.579 Ajay Ravi So or you can see that from the system there are bunch of records that are assigned to me. 0:36:37.579 –> 0:36:46.659 Ajay Ravi So pretty much for the copilot studio’s able to do is look at these different records and then I’ll provide me even the link to the records. 0:36:46.659 –> 0:36:53.619 Ajay Ravi So if I click you can see that it will directly take me to the application or as well so. 0:36:54.429 –> 0:37:0.349 Ajay Ravi You know, you do have the ability to directly get some of those information and all the details as well, directly from copilot studio. 0:37:2.409 –> 0:37:4.849 Ajay Ravi So I want to continue my conversation. 0:37:4.849 –> 0:37:7.649 Ajay Ravi So what I want to do is I want to create. 0:37:10.679 –> 0:37:11.239 Ajay Ravi Filter. 0:37:13.669 –> 0:37:17.909 Ajay Ravi So you can directly do some of these set of automation as well. 0:37:17.909 –> 0:37:20.149 Ajay Ravi So I’m just going to mention my name. 0:37:28.249 –> 0:37:28.449 Ajay Ravi Thank. 0:37:29.669 –> 0:37:39.629 Ajay Ravi OK, so you can see that by providing few information or it pretty much created an awful letter and it’ll be sent to the e-mail address that is available there. 0:37:39.629 –> 0:37:43.109 Ajay Ravi So if I bring that one up. 0:37:50.549 –> 0:37:58.149 Ajay Ravi So you can see that I got an automated e-mail with all the details which I just provided and with all the details as well. 0:37:58.149 –> 0:38:8.589 Ajay Ravi So this way you can even automate and dynamically add some of these continuousing some of the automation side of things as well directly from copilot studio and power automate. 0:38:9.589 –> 0:38:13.749 Ajay Ravi So the next thing which I want to like do is I want to like continue my conversations. 0:38:13.749 –> 0:38:16.509 Ajay Ravi So what I want to do is I want to create. 0:38:17.309 –> 0:38:17.989 Ajay Ravi A task for myself. 0:38:17.989 –> 0:38:21.29 Ajay Ravi So that way you know I don’t forget so. 0:38:22.59 –> 0:38:25.419 Ajay Ravi I’m just going to mention all create tasks so. 0:38:26.989 –> 0:38:29.69 Ajay Ravi Send welcome e-mail to. 0:38:33.389 –> 0:38:36.69 Ajay Ravi So it pretty much As for the priority. 0:38:36.109 –> 0:38:44.309 Ajay Ravi So I’m just going to set as medium priority. So on doing that so you can see that a new task got created in to do so. 0:38:44.309 –> 0:38:45.909 Ajay Ravi Now if I look at. 0:38:53.129 –> 0:39:1.529 Ajay Ravi So you can see that in new task got created in to do all and it has been assigned to me and it has also set up a due date as well. 0:39:1.529 –> 0:39:6.129 Ajay Ravi So it depends on the priority. Like you know we can automate some of those things as well. 0:39:6.129 –> 0:39:14.609 Ajay Ravi So you can see that this way you can directly create all records in different places so that way you know you don’t forget some of those things as well. 0:39:15.529 –> 0:39:18.889 Ajay Ravi So I’m just going to continue my conversation. So now. 0:39:19.669 –> 0:39:21.149 Ajay Ravi I want to ask few things related to. 0:39:21.729 –> 0:39:29.609 Ajay Ravi Time off, so I’m just going to mention time off so it will give me the prompt with you know, what exactly do I want to like? Confirm, right. 0:39:29.609 –> 0:39:32.689 Ajay Ravi So I want to ask something regarding the national holidays. 0:39:32.689 –> 0:39:42.929 Ajay Ravi So if I can just select from this option and you can see that you know it pretty much, you know provided all the details for the current year holiday and also it provided me the link. 0:39:43.249 –> 0:39:47.489 Ajay Ravi So if I just click on this link, it will directly take me to. 0:39:48.269 –> 0:39:55.29 Ajay Ravi The holiday list document as well with all the details. So this way you can get and fetch some of this information as well. 0:39:56.39 –> 0:39:59.799 Ajay Ravi And I’m just going to ask one more thing regarding time off. 0:39:59.939 –> 0:40:5.59 Ajay Ravi So if I want to like ask you know, how much balance do I have, right? 0:40:5.59 –> 0:40:20.79 Ajay Ravi So I can pretty much or click on this video balance and what this will do is it will look at the you know the PTO balance system and then it will provide me the results. So you can see that it looked at the system and then provided like. 0:40:20.79 –> 0:40:26.939 Ajay Ravi I have 88 hours of application balance, so this way you can interact with you know different system and get some of those details as well. 0:40:28.729 –> 0:40:33.449 Ajay Ravi So I’m just going to continue here and then ask couple of additional things as well. 0:40:33.449 –> 0:40:36.249 Ajay Ravi So what I’m going to do is how. 0:40:37.839 –> 0:40:38.959 Ajay Ravi Early should I? 0:40:44.589 –> 0:40:59.669 Ajay Ravi So I’m going to ask a question which is specific to a document that is available in SharePoint site so you can see that our pretty much all second how early should I apply for leave so you can see that it pretty much provided me the details that you. 0:40:59.909 –> 0:41:3.629 Ajay Ravi Know I need to, you know, request at least 10 days in advance. 0:41:3.629 –> 0:41:6.349 Ajay Ravi So the place where it’s coming from is. 0:41:6.349 –> 0:41:9.909 Ajay Ravi So I have this document that is added in the SharePoint. 0:41:10.719 –> 0:41:15.279 Ajay Ravi So you can see that it pretty much looked at some of these documentation and then. 0:41:16.19 –> 0:41:26.99 Ajay Ravi You can also see that there is citation there directly, so I can just click the citation and it will pretty much take me to the document as well with all the different lay policy details. So. 0:41:27.499 –> 0:41:39.859 Ajay Ravi You can easily get some of these information hard directly, and then finally I’m just going to, you know, at any point of time, if I want to, like, stop, I can just, you know, mention thank you and. 0:41:41.979 –> 0:41:45.819 Ajay Ravi So you can see that. Oh, you’re welcome. And we truly appreciate your time and opportunity. 0:41:45.819 –> 0:41:57.459 Ajay Ravi So this way you can create intelligent copilots and chatbot with low code, no code platform and then you know allows or you know assign certain tasks rather than having to do it manually. 0:42:0.929 –> 0:42:1.449 Ajay Ravi So. 0:42:2.999 –> 0:42:9.319 Ajay Ravi Coming back to demo, we just briefly looked at like you know how Copal STD was able to interact with. 0:42:10.999 –> 0:42:14.759 Ajay Ravi Microsoft to do Dynamics 365 SharePoint and bunch of different documents as well. 0:42:19.69 –> 0:42:37.229 Ajay Ravi So just to kind of like all let you guys know, so the copilot studio and power automate automation is really powerful where you can kind of like integrate both these and then allows you to bring your business data and get results to all your copilot. So it’s really. 0:42:37.509 –> 0:42:42.909 Ajay Ravi Powerful and Microsoft is investing heavily and then lot of new features are coming day by day as well. 0:42:44.719 –> 0:42:57.879 Ajay Ravi And just to let you guys know, so Microsoft is heavily investing and there are a lot of new features coming almost every now and then recently earlier this week, Microsoft introduced a bunch of new templates as well. 0:42:57.879 –> 0:43:2.399 Ajay Ravi So you can start with some of those template for ECHO pilot development and then. 0:43:3.379 –> 0:43:15.819 Ajay Ravi There was a major UX refresh of few months back and all these are like some of the features that are currently available and then also some of the features that are expected to come over the next few months. 0:43:20.249 –> 0:43:26.169 Ajay Ravi So where to I will pass it over back to Brian now so. 0:43:28.879 –> 0:43:30.559 Brian Haydin So what can you do next? 0:43:31.79 –> 0:43:32.199 Brian Haydin We’d love to hear from you. 0:43:32.199 –> 0:43:33.519 Brian Haydin Thanks for joining us today. 0:43:33.519 –> 0:43:37.239 Brian Haydin I if there’s any questions, you know, feel free to drop those in the chat. 0:43:37.839 –> 0:43:46.39 Brian Haydin But what we would love to do is we’d like to meet with you if you’re interested in pursuing or learning a little bit more about this one-on-one. 0:43:46.39 –> 0:43:53.319 Brian Haydin So please fill out the survey link is in the chat and we can help you evaluate a scenario. 0:43:54.109 –> 0:43:59.69 Brian Haydin You know, and determine whether this is something that’s gonna fit into the semi custom or the custom ML space. 0:44:0.199 –> 0:44:4.79 Brian Haydin Or, you know, we can do an executive envisioning session with you. 0:44:4.79 –> 0:44:6.719 Brian Haydin Or an executive briefing for your organization as well. 0:44:9.109 –> 0:44:12.989 Brian Haydin So I’ll stick around for another minute or two, but otherwise, thanks for joining us today.