Creating Agents with Microsoft Copilot Studio | #CopilotChronicles

Summary

Highlights

• Introduce the Amplify series aimed at enhancing knowledge in Dynamics 365 and Power Platform AI 💡.
• Experience the shift towards an agentic AI model, which automates business processes with minimal human input 🤖.
• Dive into the practical use of Copilot Studio for creating agents with no-code or low-code methodologies 🎨.
• Discover how AI enhances productivity and innovation within businesses through intelligent applications 🌟.
• Engage with a series that caters to both beginners and advanced users, starting from basic AI introduction to complex scenarios 🚀.

Key Takeaways

• The Amplify series is designed to accelerate learning in Dynamics 365 Copilot and Power Platform AI 🚀.
• Discover how AI can drive autonomous goals with minimal human interaction through the agentic AI model 🤖.
• Copilot Studio enables the creation of agents using a no-code/low-code approach, making it accessible to more users 🎨.
• Series offers sessions from beginner to advanced levels, catering to various skill sets in AI and Dynamics 365 💡.
• Microsoft's Copilot Studio integrates seamlessly with tools like SharePoint and Teams, enhancing its usability 🛠️.

Overview

Chapters

• 00:30 - 03:00: Introduction and Welcome The chapter "Introduction and Welcome" begins with musical elements and repeated emphasis on the term 'Heat', possibly indicating a thematic or contextual setup for the discussions or topics to follow. The exact details or narrative flow are not provided in the transcript, suggesting that further information or context would be found in the full chapter content.
• 03:00 - 10:00: Introduction to Amplify Series In the introductory chapter of the Amplify Series titled 'Introduction to Amplify Series', Paruru, the events and program manager for Microsoft Reactor Bengaluru, welcomes everyone to another episode of Copel Chronicles. The chapter sets the stage for the discussions and interactions expected in the series, establishing an engaging and enthusiastic introduction to what is to come.
• 10:00 - 23:00: Concept of Agentic AI The chapter focuses on setting a respectful and inclusive tone for participation in the event. It emphasizes learning together, being respectful of diverse views, understanding differences, and engaging with kindness and consideration. Participants are encouraged to ask questions, which will be addressed in the event.
• 23:00 - 29:30: Amplify Series Details and Schedule This chapter introduces Kamal Shri, a developer advocate from the Microsoft 365 team, who shares excitement about an upcoming session. It hints at the positive reception of the Copilot Chronicle series, indicating a strong engagement from the audience with many attendees already backstage in anticipation of the event.
• 29:30 - 38:00: Questions and Answers - Amplify Series In this chapter titled 'Questions and Answers - Amplify Series', the narrator recalls the initiation and evolution of the Copilot Chronicles Park series. They share their excitement and anticipation for introducing a new series featuring a talented group of technology experts from Microsoft.
• 38:00 - 49:00: Introduction to Maker's Path in Copilot Studio The chapter introduces Maker's Path in Copilot Studio, which isn't just a tech series but a guide designed to accelerate learning in the Dynamics 365 Copilot and Power Platform AI ecosystem. The chapter hints at more detailed discussions from the entire team, meant to unfold further topics.
• 49:00 - 60:00: Hands-on Demo: Creating an Agent in Copilot Studio The chapter introduces a new series as part of Microsoft Reactor, focusing on Power Platform and Dynamics 365. The hosts welcome Hashid, Kundan, Manmad, Niharika, and Barat to the Microsoft Reactor collaboration with Copilot Studio. The session promises hands-on experiences with creating agents in Copilot Studio.
• 60:00 - 67:00: Publishing and Testing Agents The chapter begins with a warm welcome to the audience by Kamill, expressing gratitude for being present. Some core team members from the amplify program are introduced. The discussion is set to begin with a slide presentation, indicating that the group has already connected and are ready to proceed with the content of the amplify series.
• 67:00 - 70:00: Testing and Debugging Agents The chapter titled 'Testing and Debugging Agents' discusses the work done by the Global Delivery Center at Microsoft, located in Hyderabad, India. The focus is on deploying the latest dynamics and power platform AI solutions for customers. The team leverages their deep understanding of product features and practical use cases to deliver effective solutions, which is highlighted in the 'amplify series'.
• 70:00 - 75:00: Final Q&A and Conclusion The final chapter titled "Final Q&A and Conclusion" focuses on summarizing the benefits of attending the sessions discussed. The main highlight is on the real experience shared across different products, emphasizing the added value for participants. With the growing importance of AI, the chapter discusses how Amplify aims to empower teams with essential knowledge to not only utilize but also unlock the true value of co-pilot, emphasizing the understanding of what it is, why it is important, and how to benefit from it.

Creating Agents with Microsoft Copilot Studio | #CopilotChronicles Transcription

• 00:00 - 00:30 [Music] [Music] Heat. Heat. [Music]
• 00:30 - 01:00 Hey everyone, hope you all are doing great. I am Paruru, events and program manager for Microsoft Reactor Bengaluru. And yes, I welcome you all to the another episode of Copel Chronicles. But before
• 01:00 - 01:30 we start our today's event, let's go through our code of conduct. We are all here to learn together, participate together. So please be respectful of other people views, understanding the differences, be kind and considerate in a way we all engage. We encourage you all to participate. Please drop all your questions in the comment section and we'll pick it up from there. I would now not like to delay further. I'd like to bring in our host
• 01:30 - 02:00 for Copa Electronicals Kamal Shri who is also a developer advocate from Microsoft 365 team. Hey. Hi Kamill. How are you? Hey Pat, I'm good. How are you? All well doing great. Uh sessions looks exciting. I have like too many people in the backstage. So what's what's coming up next? Oh firstly I I'm always excited. So firstly audience thank you so much for the overwhelming response again for the Copilot Chronicle series. But this
• 02:00 - 02:30 time we've got in a twist where I remember the first day when I really started Copilot Chronicles Park. So I did know the urge for the co-pilot. The audience wanted some series where they fell into co-pilot in and out with different flavors. So this time I am eagerly waiting to introduce another series where we have a amazing set of technology folks from Microsoft where we are excited to kick off a new series
• 02:30 - 03:00 called amplify. So it's just not not a just a tech series where you just talk about a particular concept. know it's designed to accelerate your journey into the world of Dynamics 365 copilot and power platform AI ecosystem. So I will not kind of reveal everything because I have the entire team here in the background waiting to jump in and talk more about
• 03:00 - 03:30 it. So I will hand it over to them. So folks, we are introducing another series who will also be uh a part of Microsoft Reactor. Every week you can get a flavor of Power Platform and Dynamics 365. So let's not wait. Pat, what do you say? Let's welcome them. Yep. All ready. Yeah. So welcome Hashid, Kundan, Manmad, Niharika, Barat. Welcome to the Microsoft Reactor in collab with Copilot
• 03:30 - 04:00 Chronicles. A very warm welcome to all of you folks. Thank you Kamill. Thank you for the warm welcome. Uh it's really nice to be here. Um I've got some of my core team members also for the amplify program and let's get started. Uh man, you want to share the slides? Yeah, please go ahead. So we have already connected. All right. So this is the amplify series
• 04:00 - 04:30 launch slide. Um we are part of global delivery center in uh Microsoft. We are based out of Hyderabad uh India. Um let's quickly is the amplify series. Um our work at global delivery center involves working with our customers very deeply to deploy the latest dynamics and power platform AI solutions. So we designed this series to bring our deep understanding of product features as well as real life use cases of specific implementations of core and AI features
• 04:30 - 05:00 across these two products. Uh we are talking about real experience uh in most of these sessions that we will have. So it'll add real good value for you to attend these sessions. Uh with AI becoming a core part of how we build, work, and engage, Amplify really aims to empower you and your teams with the knowledge that they need to not just use, but truly unlock the value of co-pilot. This includes helping you to understand the what, the why, and most importantly, the how. The series is
• 05:00 - 05:30 structured to span across four key themes. We begin with core introductions to set the foundation. Then we go deeper into specific feature areas to explore co-pilot capabilities and then we bring out the agentic experiences where participants get to see how AI agents can really drive autonomous goal oriented outcomes and then finally we focus on custom development using copilot studio how to extend the co-pilots that are already there in the product across the dynamics as well as
• 05:30 - 06:00 the power platform uh landscape right um again through amplify our goal is to help you unlock AIdriven productivity, foster innovation and prepare your teams to lead in this new area era of intelligent business applications. Next slide please. Man. All right. Uh quick introductions.
• 06:00 - 06:30 Um I'm Harsher. I'm a consultant manager at global delivery center in Microsoft. I've been here for 10 years. Um, and um, I really really like working with the dynamics and power platform uh, capabilities that our products bring. Um, moving on, Dwanch, if you're online, my camera is bad. Uh, hello. I am Dwanch. Uh, I am a technical consultant at Microsoft. I work with our
• 06:30 - 07:00 enterprise customers to adopt dynamics and uh power platform and I have a total experience of three years next to Kundan. Hey hi everyone uh I am Kundal and I am a Dynamics 365 and power platform consultant. So I have been with Microsoft uh for past 9 years. I'm wearing clothes also. See again because that closed because you have closed in
• 07:00 - 07:30 this print. Have you just pulled from the last print? Pass on to Barat. Thank you. Uh my name is Barat Mana. I've been working with Microsoft on Dynamics and you know Azure technologies for more than a decade. It's been a fantastic journey so far. you know transitioning from uh you know net uh to you know ashure and to you know dynamics
• 07:30 - 08:00 and then now copilot and the entire you know agentic era um apart from you know implementing large to very large implementation for the customers of Microsoft uh I'm also leading u AI and co-pilot initiatives and adoption drives for Microsoft at know global delivery center and love to have you all on this sessions yeah hi everyone I am Niharika I have been working with Microsoft for around 4 years and last two years I have been working deeply with customer focused on
• 08:00 - 08:30 copilot and AI. So great sharing that knowledge with all of you. Thank you. Hello everybody. Good morning, good afternoon, good evening wherever you are. Uh I'm Manmut. Uh I'm principal consultant working here with uh Dynamics platform. I have been with Microsoft from past 14 years and uh I have had great opportunity to work across uh various industries and organizations and
• 08:30 - 09:00 geographies helping organizations implement Dynamics 365. Uh so in this journey I have seen platform evolve uh solving real world problem and being part of digital transformation across the globe. Uh what excites me right now uh is this is the most incred incredible time uh of innovation of where we are seeing lot of art artificial
• 09:00 - 09:30 intelligence use cases that are coming up and uh the the space of change is rapid uh I truly believe that we are just scratching the surface of what is possible and I'm here to share my experiences uh learn from all of you uh so that we can grow together as a community. Uh all right. So uh uh with
• 09:30 - 10:00 with respect to this amplify series uh uh just who should attend this series is uh all the developers uh or technical consultants, functional consultants uh technical managers, solution architects or a business analyst uh where we will delve deeper uh into some of the use cases that we are seeing or uh we will discuss us what is possible. Uh so from
• 10:00 - 10:30 that perspective uh this series will be helpful for uh everybody who uh has curiosity to learn about artificial intelligence and what is possible with Dynamics 365. Uh moving on uh why why we have introduced this series? Uh the primary reason is uh the AI which is changing the way we work and uh how it is
• 10:30 - 11:00 changing it is changing uh because the the way we see it is we are looking at the agentic future and when I say agentic future uh it refers to artificial intelligence system that operates as autonomous agent uh which is capable of planning tasks or making decisions or taking actions to
• 11:00 - 11:30 achieve outcomes with minimal human interaction. uh unlike traditional AI uh and when I say traditional AI which often acts as a passive tool uh that responds to prompts or uh we request for some information and we get uh some sort of response. But uh in case of agentic AI, it behaves more like a proactive digital assistant
• 11:30 - 12:00 uh that can carry out multiple steps that can carry out multiple tasks. It can learn from interactions uh it can dynamic dynamically adapt its behavior based on the context. So with agentic future we can automate end toend business processes and how these business processes can be automated using agents uh with Dynamics 365 platform. We will delve deeper as we uh
• 12:00 - 12:30 uh go through this uh series. So we have uh scheduled multiple sessions and each session focuses on uh specific use cases or specific tools that Dynamics 365 provides. Uh we will look into that one. Uh now going to uh just to understand uh the evolution of this traditional
• 12:30 - 13:00 uh SAS model to agentic AI. uh so uh I just wanted to touch base on uh how this technology or how agentic AI has evolved so far. uh so so far what we have been doing is it's a tra traditional SAS model and when I say traditional SAS model any business applications that we see uh we see uh multi-layer approach
• 13:00 - 13:30 that we follow to implement any kind of business application wherein we have a UI layer uh which is centralized and standard user interface uh which has limited flex flexibility uh with respect to user experience customizations. Uh again end user interaction it is tightly coupled with application workflows that are configured. Uh then we will have uh a
• 13:30 - 14:00 business logic layer or what we call process and logic layer uh which is a monolithic or tiered middleware we where we implement multiple APIs multiple services uh which is again could be a subscription based uh where based on subscription you can access features or there are certain access controls that are in place Uh again this business
• 14:00 - 14:30 logic layer is specific to a specific business uh which is rigid where we implement rules uh based on certain processes based on certain industries based on organizational needs. uh it has limited adaptability uh to user context or a behavior because over the period of time it is possible that the user context may change or behavior may change and if those context
• 14:30 - 15:00 or behaviors are changing we will have to go back and make the changes to this process and logic. And uh last but not least the data layer uh wherein uh we see the data store which is a centralized data store uh where uh we store the data in a structured format uh which is a schema schema driven uh storage and which is being used for reports or insight based on the static data model. So this is the traditional
• 15:00 - 15:30 SAS model that we have been using uh so far. and uh how uh it is uh how how it has evolved. So uh in last few months we have seen uh there is a evolutionary step uh in artificial intelligence world uh where again uh the structure of the application primarily it may not have
• 15:30 - 16:00 changed. We still see the UI layer which is a front end which is traditional UI with maybe selective AI enhancements where we use smart forms or we use chat suggestions. So that kind of uh uh enhancements that we have been doing in last few months or year uh uh from AI UI perspective there are slight improvements in uh user personalization
• 16:00 - 16:30 also. Uh similarly when we talk about uh process and uh business log business logic layer uh where we have been using a hybrid approach where uh the traditional middleware continues to handle your core logic and at the same time we have been uh configuring uh certain AI agents which augments the decision-m process uh like uh in
• 16:30 - 17:00 Dynamics 365 world uh if we think of features like uh routing uh so which is augmenting the decision-m process and which is automating automating certain business processes uh at the same time we are using machine learning models for let's say anomaly detection or certain prioritizations or recommendations so AI
• 17:00 - 17:30 I is infused in business logic layer. Uh going to the data layer again it is traditional uh storage uh which is enriched by uh your artificial intelligence or machine learning infused analytics. Uh we are getting realtime data signals which are leveraged for predictions and insights. Uh so at each and every uh layer there is artificial intelligence which is infused
• 17:30 - 18:00 uh and then moving on to the north star uh which is agentic AI model. So uh when we uh when we look at the agentic AI model which uh which is coming uh really fast uh where again the application structure looks similar. We still have UI layer, business logic layer and data layer.
• 18:00 - 18:30 Only the thing is uh the way we see that UI layer or which is a front end is being replaced by copilot experiences where we are using natural language as a UI or chat uh as a UI. Uh this UI layer is task oriented uh where interfaces are driven by user goals. uh we do not have same rigid UI
• 18:30 - 19:00 which is little difficult for us to change uh then we have we are uh not really relying on form based UIs so let's say if if it is a web chat we are using adaptive cards or we are depending on natural languages so UI is going to completely change uh when I uh Think about the business logic layer. Again, agentic AI is taking over uh where uh
• 19:00 - 19:30 agents are going to autonomously plan, reason and act based on goals uh which learns and adapts from the context and feedback that is being provided. Again, it is loosely coupled uh like uh microservices or APIs that we use for execution. And coming to the data layer again uh there is no foundational change as such
• 19:30 - 20:00 from data architecture perspective but it serves a context provider to the agentic model. Uh AI uses structured and unstructured data. So probably uh in next few weeks or months the kind of applications that we may see is uh we may not depend on one single data structure or one single database. All
• 20:00 - 20:30 the applications probably they will talk to heterogeneous data stores. It will pull data from different different systems and it will surface it to users. uh so that's the evolution that we see that is going to happen. So uh this is this is how this agentic future which is transforming our traditional SAS model to the agentic AI model. I think we are in between where
• 20:30 - 21:00 we are uh uh leaving that transformation where artificial artificial intelligence is infused at each and every level. Uh but our northstar is uh to see agentic AI model uh which is already there. uh it's just from uh the traditional AI and agentic AI perspective uh the kind of differences that I see
• 21:00 - 21:30 personally is uh most likely the mode of mode of interactions with the applications may change and when I say mode of interaction with respect to traditional AI it's uh it is reactive you ask a question and you we get a uh response back while in agent authentic AI world uh you will see uh proactive initiatives where all these agents proactively they will be taking some actions uh be it your approval processes
• 21:30 - 22:00 or be it taking certain actions on your behalf agents will be doing it uh from task complexity perspective if I look at traditional AI uh even today we could use traditional AI uh to automate uh certain task where uh those tasks are uh getting executed uh one after another uh while agentic AI it
• 22:00 - 22:30 will be a multi-step process uh which can perform all open-ended tasks and then uh from learning approach perspective like for traditional AIS uh we have to pre-train the AI model uh which is a rule-based model while if we look at agentic AI approach uh which is adaptive which is contextsensitive it learns by itself. So
• 22:30 - 23:00 these are some of the differences that we may see in uh next uh few days or weeks or months that are coming. So this is the agentic future uh that uh we look at it and uh based on this what we want to cover as part of this uh series is uh with respect to dynamics 365 uh what AI features that Dynamics 365
• 23:00 - 23:30 provides. So we will start with introduction uh to the co-pilot how we use copilot in context of dynamics 365. Uh then we will uh see uh detailed features that copilot and copilot studio provides again in the context of Dynamics 365. uh what all agentic experience that Dynamics 365 provides and how do we extend this co-pilot so
• 23:30 - 24:00 that we can automate business processes or we can uh create agents which can automate end to-end business processes. So uh we will be covering this part. Uh with respect to schedule uh uh I think uh uh we are looking at every Thursday of the month at 5:00 p.m. uh Indian standard time. uh we will uh conduct session uh as part of this
• 24:00 - 24:30 series where uh we will look at use cases different different use cases uh with respect to dynamics 365 sales or customer service or uh field service and customer insights. Uh so uh this is the schedule uh that uh we are looking at. All right. I think so that's all that we wanted to share with respect to
• 24:30 - 25:00 amplify amplify series that we are planning to get there. All right Mmat I think you are officially uh joining the club of launching the series Amplify. So congratulation and kudos to the entire team here. Um yes um folks we are getting this launch the schedule is all up available on the screen. So based on the topic uh or the field of interest you can hop in and there will be a
• 25:00 - 25:30 registration link as well. So as a general audience man and the entire team I have like we already have so many Microsoft learns right or even events happening up the tech events. Now I want to understand how does amplify stands unique here. Why should I go with amplify in the tender series? Is it for more of a beginner friendly or is it more for an 400 500 level? How does it make more unique for me?
• 25:30 - 26:00 Yeah. Uh so uh our plan is to start with uh level 100 level 200 uh uh learning and uh we will uh we are we are planning to uh talk about specific use cases based on specific industries uh so that we can understand how we can implement Dynamics 365 for specific use case or specific industry.
• 26:00 - 26:30 So you will see uh real life problems or real life use cases that we can uh implement with. All right. Bar, you have a take on it. You're on mute. Sorry, first time. Yeah. So you had a question around amplifier, right? So really the thought around giving the name amplify is that if you see a i is you know kind of highlighted in blue. So it's you know
• 26:30 - 27:00 amplify people's you know learning and you know experience through AI right so on AI basically that's what it is oh that's an awesome AI amplifying it oh that's really that that's hashed's creativity there so great all right uh so what do I need a prior experience to attend this can I be a very begirl friendly or a foundational session or move gradually to a more advanced content content or can both
• 27:00 - 27:30 beginner and experienced professional come join the series? How how would your series take off? Yeah. Anybody? Yeah. I I I think both beginners and uh experienced folks they can join this. Uh so as I said that we will start with uh level 100 level 200 skill sets and then we will uh go deep deeper. Okay, great. And will this be more on
• 27:30 - 28:00 hands-on practice or it will be more of session based? Uh, it is going to be a session based. Okay. So, would you come up with some use cases with demoing or discussion or brainstorming some kind of business use case? Yes, we will have some scenarios that we can demonstrate and then we can h get into further discussions for those scenarios. So for another 3 four months we will not have something on hand on
• 28:00 - 28:30 practice right it will be more on getting your sec uh theory home something like that that's right that's right and um is there a certificate or a badge something that you would give to the attendees who are kind of joining this as a series or something you're planning to do a digital one. Yeah. Yeah, I think that's a really great idea. Honestly, we
• 28:30 - 29:00 haven't thought about it, but we can definitely plan. Uh we do have questions pouring in. I think uh Barat is from your team answering. Yeah, Barat is answering it. Um yeah, so we have students can join. Definitely a good question. How how how is going to benefit students? Say for example, I'm from a graduate who's going to be passed out in the next two years. Is it just
• 29:00 - 29:30 more of an awareness or is it more about what's trending happening in three uh D35 or AI? What from a student's perspective? What would they how would it benefit them? I I'll take that. Um if you look at the you know technology landscape right know especially uh for the uh those who are you know coming out of school and you know graduating you know in near future you know uh if you look at the power platform and dynamics 365 is really niche skills still in the
• 29:30 - 30:00 market right and so this is a great opportunity for them to jump onto it you know learn and then you know uh get a you know direct ticketing to the AI part of it as well right because you know the product that we had before the platform that we had before really really evolved into you know like manmouth has mentioned into an agentic AI experience for the future. So definitely it's a good uh learning for the students as well, right? This would be like any other product that you know they would be investing in for their career.
• 30:00 - 30:30 All right. Great. All right. Um so I think you've got you guys have made the entire thing clear like who is Amplify for, what Amplify is all about, who it is intended, what you're going to cover for the next 3 to four months with the first Thursday happening. So the schedule is up. So folks please do join this series. You will definitely get to learn a lot from the team of experts and thank you so much the entire amplified team for joining today. Uh good luck.
• 30:30 - 31:00 Looking forward for more sessions with you folks. Thank you. Thank you. Thank you. And yeah you can answer the questions in the chat. And uh without further ado um let's move on to our next speaker which I'm really really excited again uh because I have my colleague Paulo joining here who is going to be talking about the makers path. See until now we've been talking about copilot
• 31:00 - 31:30 chronicles where I do the copilot developer camp. I keep talking about the build path, the extend path and uh we we do a lot with declarative engine agents and custom engine agents but today for the first time we are officially announcing it in reactor the maker path from the maker Paulo Paulo welcome to the show and glad to have you on copilot chronicles looking forward for your session what we could actually do with
• 31:30 - 32:00 copilot studio with a flavor of platform form. So, how are we going to create? So, looking forward to it, Paulo. Yeah, thank you Kamal for the warm welcome and yes, I'm really happy to share that uh we introduced a new pathways in the uh copilot developer camp and it's a pathway dedicated to the uh no code or low code mode. So you can learn how to create or to make agents to extend
• 32:00 - 32:30 Microsoft 365 copilot using Microsoft Copilo Studio. The uh overall idea of the new make path is to help you figure out the options that you have available. So for example nowadays when you want to extend Copilot or when you want to create just a quick question are you sharing your screen? Uh, no I'm not. You want me to share something? I I'm in the
• 32:30 - 33:00 starting screen would be great. Okay, I can definitely do that indeed. Okay, let me know when you see it. Can you see my screen now? Okay, I think it's Yes. Great. Okay, this is the homepage of the uh copilot dev camp and in there you can also find a quite useful uh flowchart which helps you to figure out what you can learn by going through the content
• 33:00 - 33:30 that we provide in the copilot dev camp and right now I was talking about the left side so the make side and in there you can learn how to create or to extend Microsoft by copilot using the copilot Studio agent builder which means that you are in copilot uh and in copilot chat I mean Microsoft copilot chat and you can create an agent from there just using the web UI of your browser. Under
• 33:30 - 34:00 the cover there is uh the engine of Microsoft 365 copilot which allows you to create a declarative agent on the fly. Then you can learn how to extend both Microsoft Teams and Microsoft 365 Copilot chat with Copilo Studio and that's the main section of the maker path. And then pretty soon there will also be a couple of labs about how to create SharePoint agents and about how
• 34:00 - 34:30 to share those agents in teams. So this is what we are offering. If you want to learn how to extend and how to create agents using the maker approach as you already said Kamal we already have coverage for those who want to use the proc code approach. So uh working with the uh basic uh architecture and orchestrator and language model large language model under the cover of
• 34:30 - 35:00 copilot and so uh relying on declarative agents or the custom engine agents if you want to have full control of the orchestrator of the uh foundational model and everything else in the actual implementation of your uh agent. So this is where uh what you have and you can start from here to learn bunch of stuff. An interesting information that I want to share is that since couple of weeks ago you can also go through a a kind of
• 35:00 - 35:30 a certification uh path. So if you go to the copilot dev camp which is available at akmscopilot dev camp you will also find this new award section. In fact, we introduced a new challenge for the whole community. You can go through all of the pathways. So, the make, the extend, and the build one. And if you will complete any of them, for each of them, you will get a badge. If you will complete all of them, you will get an additional badge,
• 35:30 - 36:00 which will be the B pat. We took inspiration from the judo award. I used to be a judoka when I was young and so I influenced the choice. Uh it's my fault if you don't like it and if you will also uh uh complete all of the pathways and you will become a black belt and so a pathmaster and you publish a sample in the uh community sample gallery you will become a sample sensei. So it is up to you to invest your time to learn new
• 36:00 - 36:30 stuff to uh get new knowledge about uh developing agents for Microsoft 65 and to share your knowledge with the community and with the sample and as such to achieve and to claim your badges and show to the community uh what you learn. So uh I don't know if you want to dig into some of these specific topics Kamal. So I'm open to your questions or to the community questions in the chat.
• 36:30 - 37:00 Uh we don't have as of now anything specific. Uh if we have some demo or something like that. Definitely I I can show you how you can do something with the options I was talking about. So for example, let me go to copilot and I can show you that you can uh create agents I was telling you with the agent builder. This is the uh out of the box UI of Microsoft 35y compiler chat. You can always select the create an agent option
• 37:00 - 37:30 on the right panel. And by clicking on this command, you will be able to create on the fly an agent using the agent builder. And with this one, you simply can rely on a descriptive model. So you simply have a conversation with the copilo studio gym builder and you describe what you want to have or you can switch to the configuration panel and you can provide the basic settings of your agent. So for
• 37:30 - 38:00 example if you want to create a new agent like I don't know let's make an example an HR agent and this will be an agent we can provide a description this will help employees to get information oh sorry for the typo information about HR policies and documents. Okay you can provide instructions. This will be the actual set of instructions for your agent to give the specific behavior you want to
• 38:00 - 38:30 have for your agent when you create uh your your custom agent. So it could be that you are an assistant helping oops helping uh employees to find information about HR policies and rules. Just to keep it short, actually a real prompt, a real set of instructions should be richer and
• 38:30 - 39:00 we will soon release a set of hints and instruction about how you should properly create and define your instructions for the agent and then you can browse for some content that for example you could have on your uh tenant in your SharePoint online environment. So let's say that for example we have site like the following one where we have a set of documents an hypothetical HR website where in the documents
• 39:00 - 39:30 library we have a set of policies and rules and that's what we want to use as the knowledge base. So we can browse for the content that we have in our tenant and we can choose that specific uh document library and here it is. We can select it and still just and simply using the web browser I can provide all of these settings for my agent. I can for example make the agent capable of generating images or generating code excert and stuff like that. I can
• 39:30 - 40:00 provide some some starter prompts like for example uh I can say uh um career uh path let me know how to improve my career. This could be a nice starter prompt and by doing that we are instructing the agent about what to suggest to the user or uh
• 40:00 - 40:30 hiring hiring policies could be the title and could be tell me how to hire new employees and so on so forth. So you configure simply using the browser the behavior of your agent. Okay. Once you're done, you hit create and you will get in a matter of two seconds a new agent that you can use uh with a really no code approach because I just use the web UI of the Copilot agent builder.
• 40:30 - 41:00 That's one option you have. And of course, if I go to the agent, I will be able to interact with this agent and I can provide this prompt or any other prompt of my choice. That's one approach. But you can also work with and of course here you can see I get proper feedback from the agent. But you can also do more. You can use the actual compiler studio uh user interface and you can uh create agents using this uh no code approach uh
• 41:00 - 41:30 actually no code low code here because there might be some codebased sentences here and there when you configure the agent. Let me switch to a proper environment like this one for example. And here you can uh rely on the Microsoft Compilo Studio user experience to create your own agent. And again by selecting the create button you can start creating an agent from a set of
• 41:30 - 42:00 templates provided by Microsoft or you can create a new agent completely from scratch. That's what I'm going to do right now. And again with an experience which is almost like the previous one, you can simply have a conversation with compiler studio and you can define and describe what you want to get or you can skip to configure and you can provide the actual instructions to configure your agent. So let's say for example that we want to create an agent as like
• 42:00 - 42:30 as before to support the uh employees. So it could be employes agent. Okay. And again, we need to provide a description. You can provide a custom icon if you like. We don't really need to do that right now. And the description could be these agent helps employees and HR uh folks to manage content and documents about HR. Okay, this will be my brief
• 42:30 - 43:00 description. We again need to provide instructions. So this is almost the same experience as before but you will see that then here in the copilo studio experience we can do definitely more than what you can do with just the agent builder. Again let's just provide a very simple set of instruction for the sake of simplicity. So you are an assistant helping employees to find information about HR policies and
• 43:00 - 43:30 content. Okay, we can provide starter prompts as like as before to uh give a better uh uh behavioral to the agent and help users to find the right sentences, the right prompt to interact with the agent and we can add knowledge base. But I'm going to create the agent just right now and then we are going to improve it using the whole user experience of copilot studio. So
• 43:30 - 44:00 here as soon as it will be ready we can play with the whole set of capabilities that we have in copilo studio and again we can provide a knowledge base and it can be now a public website a sharpen site a table in data versse or we can even consume external services or a micro graph connector. So plenty of options here to extend these kind of agents. Let's say for example that we don't want still to rely on the SharePoint site that we used before. So you can recognize the same kind of user
• 44:00 - 44:30 experience. I'm going to use the same set of documents as the knowledge base so far. And once I've done that, what is interesting is that for example, we can add a behavior which is the uh generative AI based orchestration to this agent which means that based on the prompt that the user will provide to compile studio, the uh uh genai orchestrator will process the uh
• 44:30 - 45:00 sentences of the prompt provided by the user using natural language processing and will determine the intent of the user. And based on that will uh consume target data sources in the knowledge base or will eventually trigger topics or actions which we are going to see what they are in a matter of few seconds. So let me enable this capability and as soon as it will be enabled it is not mandatory but it is a nice one to have in uh new agents with
• 45:00 - 45:30 the new fully AI based uh approach. If we go to the topics and actions in the topics we can define conversational path. So we can define uh the flow of interaction with the user. So for example, just to give you an idea, we can add a topic and a topic could be uh something that we use to collect information from a user or based on the
• 45:30 - 46:00 information collected by the user. We use to provide uh output content responses to the user. we usually need to provide a description for the uh topic and the description will explain to the genai orchestrator when this topic should be triggered. Now just for the sake of making an example, let's say that this employees agent will allow me to manage a list of candidates for
• 46:00 - 46:30 specific roles in my company. So I can use a description like for example the following one just to give you an idea. I can say this is a topic that allows to add new candidates and for an hypothetical uh role and to trigger this one we can use sentences like register a new candidate add a new candidate and so on so forth. So by instructing the trigger about how this topic should be triggered then we tell the to the genai
• 46:30 - 47:00 orchestrator when uh this topic should be activated. Once the topic gets activated we can do actions and this is a UI which looks almost like what we have for example in power automate in the power platform and here we can choose to use actions available through the platform. For example, we can send a message to the user. And the message can be an image, a video, a basic card, a
• 47:00 - 47:30 plain text message or stuff like that. But we can also say for example that we want to ask something to the user. We want to collect information and then we want to use that collected information to provide a response to the user. And eventually we want to collect the information using an adaptive card because why not? So let's say that we want to ask with an adapted card information about uh the employee that we want to uh to register in our system and by doing that we can simply provide
• 47:30 - 48:00 the uh JSON of an adapted card right here. For example, let me open up this one. We can see that this is a JSON based adaptive card. That's why I said low code because you do something with your mouse and web browser but every now and then you do something which is more uh like code based but just JSON settings or stuff like that. So here I'm defining the structure of my card to collect the first name, the last name,
• 48:00 - 48:30 the role and the skills of an hypothetical candidate. And as you can see, we already see a preview of this card and we already have on the fly the creation of a set of variables which will hold the values provided by the user. So that for example we can rely on an additional step in this topic and this step could be for example triggering a flow in power automate and by triggering a flow in power automate we can create a real uh candidate in the
• 48:30 - 49:00 target environment. This is just to give you an idea of what is the potential of this technology. Now just for the sake of simplicity I will simply uh send a message to say that okay I have created the employee that you are looking for and I for the sake of saving some time I will not go through all of the process of creating the flow and all the other stuff but we can say we have got your data and the candidate candidate got
• 49:00 - 49:30 saved into the system. Okay, just to give you an idea of what's going on and then of course we can say that we are done with this topic. So we can end the current topic. If I will save this topic and I will call it register a new candidate quite easily in a matter of few seconds we'll be we will be ready to trigger this topic using some of the
• 49:30 - 50:00 triggering sentences like for example I want to register a new candidate and here in the right side as you can see I have a test section where I can test my agent we can see on the left side the activity map which let us know that it it identified based on the input provided by the user my user intent and so it triggered the topic that I just created and I see the adaptive card to uh register the new candidate which
• 50:00 - 50:30 could be whoever you want can be myself uh developer advocate and I'm good on computer programming okay so that's just to give you an idea of what you can do now the interesting part comes when you publish your agents and by waiting few more seconds you will be able to make those agents available through multiple channels and when I say channels and I use the plural word it means that we can create an
• 50:30 - 51:00 agent once and we can reuse the same agent across multiple and different channels like for example Microsoft teams Microsoft 35y copilot chat uh a website uh and stuff like that. I will show you soon and actually now what are uh the available channels here you can see this is all of the uh list of channels that we support right now. Let's say that you want to make this agent available in teams and eventually
• 51:00 - 51:30 Microsoft 65. You select the channel you choose to make this uh agent available through the channel that you selected and by selecting also this option we will make this agent available also in the Microsoft 365 copilot chat. So we select add channel takes a while. Right now, whenever you add a an agent to teams and to Microsoftify by selecting this checkbox, you need to save and
• 51:30 - 52:00 publish the agent one more time. This is a requirement so far. And once you will be done with that, let me do it actually now. Uh you will be able to experience to use your agent in teams with a botlike user experience or in the Micros copilot chat. And I'm going to show you what the experience will be. Just a matter of waiting few more seconds for the publishing to complete.
• 52:00 - 52:30 So should be while it's getting published. There's an interesting question by one of the audience subri he says can an agent built using co-pilot studio be submitted for the ongoing AI agent hackathon which we are a part of it. So what's your take on it? I I would say why not? I mean it's an agent like as any other agent. So uh have fun with it and yes please. Okay. So Sabri if you have an idea go ahead register for it. I
• 52:30 - 53:00 think we will leave the link for you to um register for the hackathon. Um I think the host will kind of uh add the link in the chat. So even if someone is not aware about this AI agent hackathon please register. It is a complete full-fledged a month where you have 20 plus 20 plus expert le sessions which is trained live on YouTube which are already done with one of the couple of sessions already. So today tomorrow we
• 53:00 - 53:30 do have few more sessions and the next week. So if you have missed it go ahead watch the stream you should be good to go. Yeah Paulo. Yeah definitely. So now the publishing is completed. So we can go back to the channel. We can activate the teams and Microsoft 365 channel so that we can go to availability options here and we can copy the link which will allow us to use this agent. So let me copy this uh link and paste it in the
• 53:30 - 54:00 web browser and I will keep on using the web- based UI of teams. And now I can uh open add and open the agent that I just created. So let me select add. You see now I'm adding the agent and as soon as it will be ready I will be able to open and interact with it. So again mother waiting few more seconds and then we will be able to see it in action here in teams right now. Okay. So this is the employees agent and as soon as we will
• 54:00 - 54:30 get the welcome message from the agent we can say like before uh I want to register register a new candidate. Let's wait for the welcome message because the agent is uh right now writing to me and sending me a message. But as soon as it will be ready, we can trigger this uh initial prompt or maybe let me try already to submit this prompt and wait for the actual welcome and here it is.
• 54:30 - 55:00 Okay. So let me now trigger the actual uh action. So now this is the topic that I triggered and I can say that again I want to register a new employee developer and whatever since I'm not really saving the data in a real system I'm just going through the process to show you how it behaves but we can also use it in Microsoft copilot chat. So if I go here to the chat we will see that in the right panel we also have the same agent available
• 55:00 - 55:30 here. So if we want to give to our end users the experience integrated with Micros we can use this immersive experience and still interact with the agent. So uh I can say let me register a new candidates and this will still trigger the same topic as before. So we will have again even in the user experience of copilot chat the uh adaptive card and we can interact with our agent. That's
• 55:30 - 56:00 one part of the story. The second and uh final part of the story and I'm going to cover it very quickly is the section based on the actions. With compiler studio we can not only create topics but also actions and through actions we can do uh actual stuff interacting with external system. So uh let's say that we have the uh site the SharePoint online site of the HR department with an hypothetical list of candidates stored
• 56:00 - 56:30 in an Excel spreadsheet like this one. Let's say that we want to create an action to read or maybe we can also create an action to write to create new items in that Excel spreadsheet. Now I selected to uh create a new action which will simply list rows present in a table using the Excel online connector for power platform. And here you can use whatever connector for power platform you like even a custom one. This will be
• 56:30 - 57:00 a connector that I'm going to use to just get the list of all of the employees or the candidates actually in the list of candidates that I have in Excel. So this will be the uh name and this will be the description of my custom action. And once I have created the action, I can go and configure the additional settings to target the Excel file that I want to read the data from. So here we are. We can edit this action.
• 57:00 - 57:30 We can say that from an input point of view, we want to configure the location of the file as a static value and it will be the actual URL of the SharePoint online site that I have and that I use to store the file. For the document library static value, I want to target the library of candidates that I have in
• 57:30 - 58:00 my site. Then the name of the file is the following one. So I'm a lazy developer. So let me copy and paste the value. Sorry about that. So the file will be a static value again and will be xlsx. And last but not least, we want to read a specific table in the Excel file which will be the candidate table. Let me save it. And again we added a new capability to our agent. And by
• 58:00 - 58:30 publishing the agent again we will be able to interact with this action and send a prompt to the agent to get for example the candidates in a specific department. You see we have specific roles for the candidate. We can say give me all of the candidates with role HR director or HR manager just for the sake of making an example. So as soon as it will be published, I can show you the behavior
• 58:30 - 59:00 in the test UI of copilot studio as well as in uh teams or in Microsoft 365 copilot chat. So now let me test it right here. Let me start a new interaction and let me provide this prompt. As you can see now I'm triggering the action that relies on the Excel test sheet. However, in order to be able to being able to access the Excel spreadsheet, I need to connect to the Excel connector and as such, let me
• 59:00 - 59:30 do the actual connection and once I will be connected, I will be ready to consume my external Excel spreadsheet. Okay, so let me retry. Every now and then when you are too fast retrying, which is usually the case in a demo, it will fail the very first time. And in fact, you see I was just waiting for it. Let me refresh the session. And now we should be ready. So let me list the candidate with that
• 59:30 - 60:00 specific role. And you see we will get back the candidate from my Excel spreadsheet in the HR director and HR manager. You see we have got the whole list of candidates from the Excel file in JSON format. But then the AI engine was able to filter out only those with that specific role that I'm requesting for. And again you can have the same experience in teams or in Microsoft compiler chart. Uh we are hitting the end of the session. So back to you Kamal if there are any questions or any
• 60:00 - 60:30 burning topics that you want to cover before we close up. Oh nothing much Paulo but definitely this is something uh people are going to love and we are looking forward for more sessions on the makeup path. So this was one general awareness as to what creating agents in the st studio and it's like creating once and using it multiple times wherever you want whether on teams or co M365 copilot chat and how to integrate
• 60:30 - 61:00 an excel getting data from it the action part really an awesome thing just wanted to understand we always are so busy building one publishing one getting it in on other M365 suites thing. How about testing of an agent? Have you ever thought about it? Like testing of agent that you built? How do you do that? Well, testing an agent built with Compilo Studio can be easily done with the test interface that you have right here in the Compilo Studio user
• 61:00 - 61:30 experience. So on the right side, you can interact with the agent. You can see in the activity map what's going on. You can dig into the input and output arguments as you can see right here. So you have a really nice testing experience fully integrated in compiler studio. Then it can also be that you want to test it also from the uh for example from the Microsoft 65 user experience. If that is the case you can enable the developer mode. Now let's see if I remember correctly the syntax you
• 61:30 - 62:00 should be able to do developer mode on if I'm not mistaken. And by doing that you can enable the uh developer mode in Microsoft 365 copilot chat. Oops. Of course that was not the right one. I might be wrong on the name of the command. Uh sorry about that. But with a command like that you can enable a kind of debugger mode in um Microsoft copilot chat so that you can see what's going on under the cover. So for example for a
• 62:00 - 62:30 declarative agent you can see uh what kind of uh external plugins will be triggered by the agent with uh what kind of input and output arguments and stuff like that so that you can do additional debugging in the final user experience of mic compiler chart but definitely if you're building an agent with compiler studio this test UI is the uh most appropriate one and you can also choose if you want to use the full debug mode or If you want to switch into the
• 62:30 - 63:00 regular mode, so you will get the same behavior here in this test environment that you can have in teams or in Microsoft compiler chat. Perfect. Uh we have another question saying can an onrem database be used as a data source? Well, if on prem is not directly accessible from a cloud-based uh platform like this one, but of course you can rely on technologies like the data gateway or you can rely on a custom
• 63:00 - 63:30 REST API of your own that you can configure as a custom power platform connector. So, uh the short answer is out of the box no. But if you create a custom API or a data gateway then you can even reach an onrem uh data source. Yeah. Okay. And yes Naven we do have the session recorded. I think the host will post the link of the YouTube. This is also getting live streamed on YouTube.
• 63:30 - 64:00 So you do have the complete sessions getting recorded. And uh one question from Barat he asked is the copilot studio license sufficient for the deployment or do you need the entire package? Well I'm I'm not a licensing expert honestly but in order to use copilo studio you can you can rely on multiple options. You can have a copilot studio license at tenant level and then you need to license the makers who will make the agents and of course if you plan to use the agents in Microsoft
• 64:00 - 64:30 copilot chat you will also need the micriy copilot license for the end users or uh you can also rely on a consumption model. Don't ask me about the details of the prices and stuff like that because I'm not the good one for that kind of questions. But with a consumption model, you simply use it and you will pay it uh as you go uh on a consumptionbased model. Perfect. I think again watch out for the on the official Microsoft website the pricing part since we are
• 64:30 - 65:00 not up to date with the pricing it'll be more detailed. Um so yeah that's it with the questions. Thank you so much for the makers part and audience again uh we had few questions and how to register for the amplify series. Please do follow up the reactor uh link that's been provided. You will find more sessions what's the upcoming sessions and you will have a link to register. So we do have sessions published in the comment section. So thank you Paulo so much for joining in today. My pleasure. Thank you
• 65:00 - 65:30 for having me. Yeah and thank you audience for joining for another season of uh Copilot Chronicles. Looking forward for more such surprising sessions going forward. Stay tuned. See you. Bye.