Unlocking the Potential of AI with Copilot Studio: Power CAT AI Webinars

Unlocking the Potential of AI with Copilot Studio: Power CAT AI Webinars

Show Video

Welcome, welcome, welcome everyone. Let's see. We have a f a few folks rolling in. We're about 37 people. Um, welcome to this PowerCat AI webinar. Um, sponsored obviously by PowerCat. No plug there that we won't take, but good morning. Hopefully everyone's going to

continue to roll in. We're going to get started right on time um with a couple of introductions. So if you are happy and you know it a little bit of interactivity let me see you raise your hand some inter can I get emoji okay okay thank you Jamie thank you on thank you John I love to see it all right well let's continue who are we um we'll move along with some quick intros my name is Lenise Lansancy I'm a product manager here on the power skill team um but I'm just facilitating today this is all about um the AI webinar hosted by deer.

Deer will say hello. Quick intro there. Deer, want to say hello to everyone? Yeah. Hello. I'm I'm Derer. I'm uh I'm also a program manager in in the PowerCat team, but I'm not in the scale team. I'm more a solution architect uh in in the copilot studio team. So, I'm uh working day in day out with some uh strategic customers globally uh making them successful on on Copilot Studio.

And Wayne, thank you. Deer, say hello to the folks. Hey everyone, Dwayne Robinson, principal co-pilot architect and uh PowerCat organization uh kind of specializ in copilot studio. Uh look forward to you guys uh getting something out of this today. Thank you. Uh so in this uh in this webinar we will see basically uh how and when to use uh copilot studio. So um we have a very packed agenda. So

um first of all um we will see how AI is transforming uh your business processes. So how AI is is um has an impact on on how we look at processes and and how we interact with with with our processes. Um then we're going to look uh have a look at which tools do we have in our tool belt? Which tools do we have in in our Microsoft platform that can assist in uh in building these uh AI agents? uh and then uh we'll also have some proof from uh from our customers that um show the impact of uh of uh of those AI agents in uh in in their organizations. And then I guess the most important part the the thing that we're all looking forward to is is uh Dwayne uh that will uh demonstrate a couple of uh uh scenarios demos uh on on how to use uh Microsoft Copilot Studio. And then like Lanis already mentioned at the end uh we will have also uh hopefully some time left for some uh some quick uh Q&A. So with this set um if we if we

look at um uh our customers then um we already see that they have uh implemented some some amazing u AI agents. um AI agents uh basically um in in whatever department or team that you have within uh within your organization uh there are so many use cases so many opportunities to uh get those uh agents in in in your processes. For example, if if you're in uh customer service and and you want to have an external facing agent that supports uh your customers on uh customer inquiries, um uh you can also have an an internal facing agent uh in in customer service that is going to help your customer service engineers uh answering questions or maybe you can uh make that also an autonomous agent so that the agent is is uh able to to answer those uh those questions autonomously without having a service engineer. uh uh in in in between. But you can also have a sales agent uh a lead generation agent that is going to create uh um your uh your opportunities, your leads in Salesforce. Uh you can also have an an

HR agent for employee on boarding or uh having an agent that assists in in in creating learning plans and and development plans for for your employees or in in in IT context uh where you have an an IT agent explaining all the diff different procedures and processes that you have internally to uh for example order a new device or uh that just explains the the policy on uh getting a new environment and uh and and so on. So, so basically what you see is that there are so many use cases that are um that are available across cross industry cross department uh where you can u uh have an AI agent uh supporting your uh your processes. Um, I'm pretty sure that when you just sit down for for half an hour and think about think about your processes, I'm I'm pretty sure that you will come up with some excellent scenarios that you can uh that you can transform uh in uh in uh AI agents. So this this gives a yeah another issue because you have that many scenarios that you can that you can uh transform in in AI agents but how would you prioritize th those use cases where where where to start exactly? So to to help you we we have three questions that you can that you can ask yourself as first of all um we recommend that you clearly describe your your the problem or the inefficiency that that you want to uh that you want to solve that you also um look at which are the stakeholders which are the teams that will benefit uh from uh from your solution and then also clearly describe what are uh your uh success criteria for for this agent and suppose uh let's take that customer ser um your success criteria could be that I want to reduce the the the service tickets by by 50%. Something like that or uh in an HR context I I want to uh increase the uh customer satisfaction of my employees. Um, another uh question that you can ask yourself to prioritize those those use cases is is to identify which area that probably will have the highest potential uh for impact when when you when you're introducing AI agents. For

example, uh maybe your customer service desk is just uh overload with with all uh with with tickets. So then uh I I would advise to to have a look at the uh the processes in in in customer service desk or maybe you just have some u a lot of unhappy employees about all the administration uh that they need to do. Uh so then let's let's have a look at some uh some HR processes. Um and so that you you immediately get a a huge impact by introducing those uh those agents. And then the third question that you could ask yourself is do I have the necessary data, the necessary data, structured data, documents um that will uh fuel your agent um with with all the necessary knowledge that you will need to answer u yeah questions about um uh FAQ documents and and and so on. Um, so make sure that you have access to to to that data uh and that you can u uh use that data and that you're also allowed to use that data in uh in your agents and then also have uh have the buyin from uh from your leaders because uh AI is is still very new very early in for a lot of organizations. So, um the chance

that you will need to have a feature um that uh requires some additional improvement uh uh some additional um um u approvals uh is uh is high and so you need to have uh buy in from from your leadership as so that you can uh enable those those features and that you don't need to look for workarounds and and and so on. And so these are just a couple of questions that could help you prioritize uh those uh those use cases and once you have uh prioritized them then um yeah we can have a look at which tool are we going to use to implement those uh those use cases. So our vision for Microsoft is that we uh want to have a copilot for every employee and an agent for every business process and uh an employee a user will will interact with uh Microsoft 365 copilot and then Microsoft 365 copilot which is available as as as a standalone tool uh uh in in your Windows or with all within all the uh Microsoft applications in in your outlook in PowerPoint in Word you copilot available and then you can ask questions like summarize uh my my my last lost emails or um give me the latest updates on on this particular document. So and then as a as a user you also have the the the uh ability to extend your Microsoft copilot um uh agents. So uh for example when you uh

when you ground uh um your agent in uh um uh in specific data for example you have a policy uh document and you want to create an agent specifically for that for that policy document and then you can extend the Microsoft 365 copilot um and by uh creating what we call a declarative agent and that's then also available uh within uh your Microsoft 365 copilot and then you have the more where you have a more complex example or when you want to have more flexibility more also more more governance and then you can use copilot studio copilot studio then to uh create all of those uh custom agents and then you can also service those custom agents in your Microsoft 365 uh copilot so that you have that single pane where you can access all those uh all those agents and that's also something that we will see in one of the demos of uh of uh of the way later on so And then what what what kind of u u as so now we know which tool we want to use about what kind of agents can we create uh with with copilot studio. So um if we look at all of those uh use cases from uh from our customers then we can um identify them uh in in three main buckets. So we have those rather easy simple scenarios and and a lot of customers start with these uh particular use cases which we call uh retrieval agents. Uh retrieval agents are agents that you just ground in in uh specific enterprise data and can be uh a policy document can be um uh can be some uh structured data in a in a SQL or or in a in a data vers table. So about you you

ground your agent uh on uh on that data and then you can ask uh questions about uh about that particular data. So very easy, very straightforward. And then the next bucket is is our task agents. The task agents um um will um are agents where you allow the agent to perform an a task on your behalf. It can be for example um I want to uh allow the agent to create a Salesforce opportunity or the Salesforce customer in my in my Salesforce uh environment. You can uh because we have so many connectors available and for so many applications and you have a wide variety of uh activities that you can uh uh that you can add in your uh in your agents. And

then the third bucket is the where you have the more advanced agents that are more the autonomous agents. they are uh running uh completely in in in the background. um are uh listening to various events that happen for example uh an email that arrives or uh a Salesforce ticket that is being created and that will trigger your autonomous agent and then your agent will then uh come up with with a plan based on all the knowledge that it has all the actions that have been configured in that a agent and then we'll uh create a plan and then we'll execute also that plan accordingly and then uh so that uh all running uh in uh in the background and so that the the three main buckets that we that we see. And then if we um have a look at some some actual uh use actual um yeah uh use cases for for those different uh different buckets and then for example if we take the IT help desk agent so that's um retrieval agent where you just ground your agent on top of an FAQ document and then um you can ask questions like how do I connect to the corporate network but then you can also have a retrieval agent where you're going to um connect to a uh to an external system and like for example your project management system and then you can ask questions like for example what's the status of this project and what's the remaining budget and then it will get that information from from your target system and then you can also let that agent also do some more stuff on your behalf for example like when you have a device uh refresh agent then uh you can allow the agent to create an order in in SAP for example and then you you have you can build towards the more advanced where you have for example a customer service agent that is then going to run autonomously in the background resolving tickets based on the knowledge that we have configured based on the uh um connections with for example our service now system and uh and and so on. So many different options, different uh uh agents that we can uh that we can build. Um but also from a creator perspective. So uh you

also have uh different options. So uh it doesn't matter if if you have coding skills or not. So if you're just uh a business user and then you can start with the uh agent builder in uh in in copilot studio to uh just extend at those uh uh Microsoft 365 copilots and just ground that uh uh with with your data uh with some documents from SharePoint and then uh you're you're good to go. So very easy. But if you're a bit more advanced, you you're uh and you have some knowledge about Copilot Studio, then can you can easily also create without any coding skills and you can create your also your agents in uh Copilot Studio. And then if you're

really a prodev and and and you want to develop certain uh certain agents and then you can also use a combination of copilot studio together with uh with uh Azure AI foundry for example or as other AI uh um Azure functions like uh um Azure AI search and uh and uh and so on. So also on this aspect from a creative perspective you have the full range depend no matter where you are in uh in the journey. So now that we have our tool uh we have our use case we have our tool. So now how do we uh create uh an uh an agent in incopad studio. So typically

you start with uh with creating your your agent. um various options that that we have available as so either you can start from a blank canvas. You can also start from an as existing template.

There are several templates that are already available that you just can uh tweak to your needs or uh you just can describe uh what your agent uh should do and that will create your agent. And then uh what you're typically going to do in uh in your agent is um especially when you have an autonomous agent and you're going to add uh one or more triggers. Trigger can be uh an email, can be uh an event in Salesforce, can be an event in uh in in data versse. So your agent is then going to listen to those events and is then going to trigger um uh whenever that h uh event uh happens. Then the next thing that we're going to add in our uh agent is is knowledge.

Knowledge can be websites, internal websites, public websites, can be um uh files in SharePoint. You can also uh for example upload uh PDF documents in uh in the agent itself. Uh you can also connect to data versse. You can connect to fabric. uh so many uh connectors are

also available at to connect your data as knowledge in your agent and then your agent will be able to reason over uh this uh this data that you that you configure. The next thing that we also have in in cop studio is what we call topics. topics will allow you to uh have a a very prescribed u path that you want to use to navigate your uh end user uh through a conversation. And so while you explicitly want to have full control on what questions will be asked, what should be the the the the response like the the options for the responses that will be uh given to the to the user and and how to uh interact on those on those responses. So that's a very uh defined script um where there's um and that that's something that you can uh use uh with uh with topics. And then the last thing that you

that you can add in uh in uh in your cop studio agent is um is actions. So um actions can be um connecting to uh one of the uh thousands of of uh connectors that we have available that connects to um yeah huge amount of of uh external applications that you have available but can also be connecting to for example like a cloudflow where you're going to use an uh uh a custom prompt uh where we're just going to inject uh uh a specific prompt uh in uh in uh in your agent. So once once you have configured all of those different different aspects um your agent is ready uh you're going to publish that agent and then you're going to make that avail uh that agent available in one of the one of the channels that we have uh available in uh in copilot studio. So you can deploy your agent for example to your public website if it's an external facing uh agent but you can also um publish that to teams or Microsoft 365 copilot as when it's an internal agent but you can also use that in uh in um um connected uh your agent at to uh to a customer engagement hub like like Genesis or Dynamics 365 customer service or Salesforce so that you can have a a handoff from your agent to a to a to a physical uh physical agent. So many options that that are

available at to make your your agents available in uh um in those different channels. Uh once that uh agent is uh is available and you also want to monitor uh what is happening with with your agent to make sure that uh you have addressed all the different uh scenarios. if there's maybe some uh some some topic that is not being hit correctly or whatever. So that you're going to monitor your agent and then uh improve the agent where uh where needed and you have that all uh those built-in analytics in uh also in uh in the tool. And then as the last step you can

uh also extend with a lot of prodev uh capabilities. So you can also connect to uh Azure AI foundry or uh use Azure AI search or use some custom Azure AI models or uh connect with an Azure uh Azure function or or or whatever. So many options also exist to extend towards uh prodev uh capabilities. So that's typically at the cycle to to build uh to build your agent. So um well also let's let's take a moment to see what customers are actually saying about Copilot Studio. What do they specifically love about Copart Studio? So especially at the uh the time to market. So uh it's a SAS

platform. So uh there's no need to install anything on on your machine. uh it it also goes ve like very quickly and you can have an agent uh um in in in a couple of weeks uh deployed uh to your production environment. Um what we also see is that with copout studio it's uh you can have uh huge ROI and and cost savings. Um even uh for uh some very simple uh use cases where you're just uh listening to a mailbox and then based on that the content of that email doing some some actions is a very simple agent but can already have a huge impact and and a huge value uh within uh within your organization. It's also a very easy tool

to work with. So it's it's also very easy to add knowledge uh different knowledge sources in your incopad studio. uh you don't need to configure anything additionally or whatsoever just add knowledge in your agent and um copad studio will take care of it um it's also very easy to start building with it so it's very uh straightforward u like I said it just can describe your agent and then uh you will be guided through a process to to create that agent so without any coding skills or whatsoever um and then also uh you also have the ability to to to use uh the let's say the old and the new world, the the the new world with all the AI features and and the generative answers and uh and and so on. But you also have those uh those those topics for the more traditional approach where you have uh really a script that you want to follow uh and that you want to guide your users through that conversation. And since Copad Studio is also a part of the power platform and there are also a lot of governance controls that are out of the box available if you you you can um configure who should have access to a certain agent and who should be able to make changes to that agent uh or who should um uh what channels should be available in that agent or what connections connectors uh should be available. And so a lot of uh governance

controls are just out of the box already available and that allow you to uh to do uh a lot of uh configurations. So and then if we map that to uh some some actual u value that we that we see uh from our customers and these are just a a couple of examples. Um, for example, if you look at uh HP, uh, HP has um has an agent that solves over 1.7 million customer inquiries per per month. So that that a huge uh huge

amount of messages that are flown through that uh through the Copart studio agent. So uh really shows that we can uh that we can take that uh um that huge amount of of of messages. um PayPal uh reduced their IT and HR help desk uh with with 40%. And this with um less than 6 weeks of of development. So that really shows that it can go really fast and and uh you can really show value uh very early in uh in in the journey. Uh another example, Rabo Bank. Rabo Bank uh

has um 80,000 uh uh calls per month and they were able to reduce 50% of those calls uh to uh uh towards the those uh to the uh contact center. So that uh a huge amount of of uh of calls that don't come to the actual service desk engineers anymore. uh pets of home um created the fraud uh fraud detection agent and and saved1 million of of dollars in uh in cost. Uh so that's also uh a huge amount and then like I say there are many many other uh customers uh that also have uh a huge amount uh of uh of value by just uh uh using those uh agents. With this said, um, I would transfer to Dwayne for some demos.

Cool. Thanks, Peter. So, what we're going to do today is we're going to go first into taking a look at uh the retrieval scenario and also the taskbased scenario. So, I'm going to share my screen real quick and deer, I'll let you know when to take it back. Um, and everybody give me thumbs up if you can see the screens. Screen on your side. Yeah. Yep. I'll see it. Yep.

Excellent. So, so where we're going to start today is we're going to start with a scenario. And the scenario that I have right here is that I wanted to build something fun to kind of play with for you guys. So, what I did is I have, you know, maybe an unhealthy obsession with Porsches. So, what I decided to do is use it as an example here of being able to go and answer questions about the Porsche 911 models and things of that nature. And we decided to actually make this a little bit more intriguing in the fact that we wanted to make it where we could use a couple of different things um using topics, actions, and knowledge.

But let's first take a look at a simple example of retrieval. And in this case, what we're going to focus in on is this knowledge where we've got uh a brochure and the model uh information that's been uploaded into data versse for it. And so when I come in and I want to ask certain questions like I can come in here and we can ask a question about you know what colors does it come in? And you're going to see that it understands what it is that I'm talking about because it understands the context of what it specifically is. Um, but then it's going to go and it's going to pull in some knowledge that's available to it and try to find an answer to this particular question. So, what you're going to see

is it comes back and it actually answers this question with uh things about exterior colors and interior colors and things of that nature. But I can even ask follow-up questions like, can you get a 4S and red interior? Now understand that it doesn't necessarily have to be exact words. So you'll see here it's actually understanding that I'm talking about the 4S and it's coming back and telling me, oh actually, yeah, you can get the Bordeaux red uh leather um and things of this nature. So it understood that what I'm trying to ask and actually get this information out of these brochures. Um so the other part of this and this is an example of just retrieval and retrieval is pretty simplistic in the way it works but when we look at something more uh more advanced we can have things like topics where topics would allow me to be able to come in and be able to say oh I want to collect some information to get from the person to be able to get um their information to request an info package.

But we also have this which is what we call an action. And in this one, what I actually went out and did is I built a API that will let you check the status of a of a order for your Porsche. And if you'll give it the information about the order number and you can see here's the description of that. And you'll see on the outputs, it's going to give me all kinds of information about the thing that I've ordered. Now, just to show you this in action, what I'm going to do is I'm going to I'm going to actually ask a couple of questions that are more in the taskoriented side. So, now we're moving to that middle tier that deer talked about and I'm going to say something to it such as I would like to get an info packet. Now, when I say that I want to

get an info packet from it, the what's happening on the back end is it's trying to figure out, well, what would be the example of being able to do that or what is the tool that I have to do it? And you'll see that it called the request info packet and it's asking me for my name. Now, with the beauty of slot filling, I'm actually going to go ahead and answer all the questions it was going to ask me. And you'll see it populate everything in and we've got the answer here. Now, that's great. But when we want to start interacting with APIs and not just doing custom topics, I can also say something like what is the status of my order. Now when I ask what the status of my order is, you're going to notice that the generative AI model sees that it needs an input of an order number. And you'll see that it asks me for the order number. So because I

didn't provide that, it automatically understands that it needs to ask me for this. And so what I'm going to do is I'm going to drop in my fake order number here uh for my new 911. And voila. You can see here apparently I've ordered a Carrera 4S uh and you can see all the information about it here that it's all been returned. Now what I can also do is I can ask specific questions about it. I can say where is my order being delivered and when is it going to be there? And notice I asked these follow-up questions and it's going to tell me it's going to be delivered at Porsche of Nashville and here's the date that it's going to be delivered. And again, it's getting all of this from this connector that we have. But because

of the fact that we maintain the context of the whole conversation, I can ask something like what were the interior trim color options for my model and know that we're talking about the Carrera 4S because that's the one I had. So you can see it goes back into the knowledge and actually figures out what were the trim options for the uh Carrera 4S. So this is just a great example of how we can use C-pilot Studio to both do retrieval and task oriented uh capabilities. And

so this is really sort of the key capabilities of Copilot Studio that's going to be a little bit more advanced than what we're going to talk about in the next stage. So deer, if you want to bring back up the slides. Uh, yep. Sure. Uh, do you want to show first the uh demo on uh making that available in Microsoft 365 or No, we'll go to that in the next section. So, if we'll just go ahead and uh bring up this. So, the main

thing that I wanted you to be able to see here is that uh and uh you can ask questions into the question uh system and we'll get those answered and I'll answer how you did that. Um so now if we take a look at the key takeaways of what we saw here, you're seeing that information workers can go out and build agents using a simple description. You can see that we grounded it inside of enterprise data and know that we also I think the we're we're too far in advance. I think we're on the ones that are takeaways for the next one.

Here I can show my screen if if it helps. Yeah. Yeah. Okay. And what I would tell you guys is just be aware that it's included. We do include if you are M365 C-pilot, know that you are having the ability inside of the M365 C-pilot to be able to use both um both teams. So, the next thing I want to show you guys is going to be the M365 C-Pilot extensibility capabilities that come in.

So, we kind of talked about that at the very end just a second ago about how it's it can be included as part of your M365 co-pilot. But what I want to show you is I want to go into what does that mean and how how does this sort of work. So, what I wanted to show you is I'm going to show you a couple of things. One is that in this is the M365 copilot and you can come in and create an agent on the right rail and this is the M365 co-pilot studio inline builder experience that lets you build what we would refer to as like a declarative co-pilot or a co-pilot agent that is basically you can come in and you could use something like I could say I want to use this career coach as a start and you can go into configure and you can adjust your um your icon on and the descriptions and the instructions that you want it to use to follow. Uh just right here. You can also come in and

browse for knowledge and and use knowledge that's in the actual Office 365 graph. And you can even see like I can go into operations and just say all I want for this one is the human resources uh data that's inside of it. You'll also see that you can choose whether you want to use web search and I encourage you guys to look at these little icons here. they're uh there to help you understand what each of these things do. So, this will add additional

web content. You'll even see code interpreter which lets you go in and do interpret interpreting and doing uh things like math problems. And then you even have an image generator if you want it to be able to do that. And if you notice down here, we've got all these starter prompts. And you'll see the starter prompts over here on the right. So, I can say something like networking strategies or things like that. And what

it's going to do is you can test this thing before you actually create it. And you can see here it's answering the question. It's created uh some input for us to be able to see what we can actually do uh with this particular thing. And I could just hit create right here and then it would automatically add it to that right rail for me for myself and then I could decide to share it with my organization. Now, you'll notice that

some of the things about this is that we don't really have the ability to access APIs or do things like that. And that's where we go into Copilot Studio again. And when we look at this and we'll play with that same one we were just dealing with before, which is the this 911 manual. And if you notice as you scroll

further down inside of Copilot Studio, you'll see starter prompts just like you saw before. And those starter prompts are going to give you the ability to interact and make part of what you can do with the channels and be able to make it where I can now share this directly into M365 copilot by going into the channel for teams and M365 Copilot. You just check this box and then you can say availability options and you can see here that I've actually shared it with everyone in in my org and I even made it available in the app store. Now, that can take some time for it to do it. It takes a little while for it to populate and get your admin approvals and all those type of things. But once you've

done this, uh what will happen is it'll become available in the get agent section. So that that way you could go in here and find it in your own corporate library of different ones. And you can see right here that I've actually already got this 911 manual one published and it's available and it's the same one that we've been playing with before. And if I ask about color options and things like this, you're going to see that it's going to go out and it's going to connect to that to that agent that we built inside of uh Copilot Studio. That is a full agent.

And now keep in mind that the information that was stored and where it's getting this answer is not inside the graph. That was actually uploaded into data versse to get all this information. Now, the other thing to be aware of is if you're not familiar with uh M365 copilot's capabilities, you can also at call out the uh specific agent that you're wanting to talk to and you can ask it other questions and directly directed to that particular agent. And so in this case, I can actually ask a question about that order that we were talking about before. And you can see I

actually provide the order number in. So, we'll see slot filling take place where it doesn't have to ask me a follow-up question. It's just going to answer the question. And you can see I got the answer about that order that we had before. Now, that was calling a

back-end API and using actions and everything. So now with the power of using Copilot Studio and publishing to the M365 Copilot as a channel, I have the ability now to bring these advanced capabilities and dynamic chaining and all the greatness that comes within Copilot Studio directly into the M365 C-pilot as a co-pilot agent and not just doing declarative co-pilot. So that's a major huge improvement in things that we are able to do. And so data, let's go back over and I think we've got some polls that we wanted to do as well. And let's do the takeaways first here. So what we're going to talk if you can see what we just saw is that I could easily go extend agents across any channel using C-pilot Studio. You could

see that the agents can do more and answer questions. They can act on uh your behalf. They can interact with line of business applications. You can see that there are over 1,500 connectors out of the box uh for your agent and all of these things are able to be brought into the M365 co-pilot capabilities as well. And so just be aware that this is really sort of key capabilities that I wanted to call out for you guys. So I think

there did we have a poll here? Oh no, we're going to jump right into autonomous agents. So with autonomous agents, autonomous agents is a different concept guys where what we're doing there is is not the same as business process automation because what we're going to be able to do is proactive and independent like solutions that are going to become very adaptive in the way that they work. So it's not a matter of if this then that type of scenario. Matter of fact, deer, if you'll go to the next slide, I want to kind of showcase what we're talking about and what I mean by this. So, when we talk

about a declarative or a a top a business process automation, when you're dealing with business process automation, what happens is it goes through the same steps every single time, every time you run it. A business process automation is actually a lot of times a piece of an autonomous agent because an autonomous agent can actually go through and do a lot of different things. Like it could actually make a decision based upon what it's what it's getting or the information provided as the input to decide to run one way one time and run a different way a different time. And then maybe there's a human in

the middle that needs to do something and then based upon that it will go do what it needs to do. The way I like to explain this is if you think about it in the terms of imagine autonomous agent is you writing a job description. You going to a person and you saying this is what your job is. This is how you do your

job. Here are the tools that you have to do your job. Here is the information and the and like manuals and information that you're going to need to be able to do your job. And then oh by the way this is when to do your job. When this happens go do your job. And so what I'm

going to do is I think we're going to just jump into the demo here and yeah let me show you this part. So if we look at autonomous agents and an autonomous agent which this is an example of one and in this case what we're talking about is someone goes and builds this thing we said hey what if someone wanted to book a cruise and they said hey I'm interested in booking a cruise and they come to my website and fill out a form saying I would like to get the latest specials for me to do this in an effective way what I want to do is I want to be able to say here's the instructions of what I want you to do to be able to help out being able to figure this out. Now, I need to be able to turn on this generative orchestration because generative orchestration is the lifeblood of an autonomous agent. So, now when you go and you look further down, I don't really have any additional topics, but I've got a whole bunch of different pieces of knowledge that I've given it from deck plans to cruise itineraries, uh, protocol excursion guides, all of this stuff. so that it

has all this knowledge about all the different offerings that we have. And then when we go into actions, which I like to kind of expand this a little bit for you to see, I can create a new lead in CRM. I can generate booking links out of Salesforce to be able to say, "Hey, I want to get a unique booking link." So when someone clicks through, I know that that's how they got that. And then we

can also pull their uh their booking history. We can go and also uh retrieve leads from CRM. We can send emails responses to them and we can update the customer information as well inside of the CRM. And the beauty of this is that

there's this new capability called a trigger. And you'll see here that when I talk about this this trigger, it's going to say when do you do your job? And when I do my job is when the actual uh form gets filled out. Now you guys don't want to sit and watch me fill out a form. So, luckily we already have one that has been filled out. And what I'm going to do is I'm going to kick this off like this form was filled out by someone and we're going to watch this thing work.

So, we're waiting for the form to get uh submitted. Now, the form submitted, it goes and pulls the leads from CRM. It got the booking history and everything. Uh, and then what it's doing now is it's looking at, oh, well, what is the different knowledge sources? what are the itineraries and things that I might want to recommend to this person based upon their booking history. And then once it gets done with that, it's then going to create the booking link for us. It's going to send an email with the information that it found and then it's going to go into CRM and it's going to update the CRM record for us on the fly.

Now, to see this actually work, the easiest way to see it is if we go into Outlook And inside of Outlook, this is the email that we were sending out. And you'll actually see here, here's an example right here of one that just where it created the email and it sent everything out with the options. Now, the beauty of this is I could even go in and we talked about triggers. I could even come in here and just add additional triggers and do something like email and say when maybe they came back and replied to that email and I need to monitor this inbox to be able to see okay when this thing came back in. Maybe somebody replies back to it and says hey I actually would like to go to Hawaii this time. Do you have any specials for

that? And then they can ask even more questions and be able to interact and all of those type of things. And they're doing it with this autonomous agent. And we could even potentially go into, okay, let's go through and do a full booking or be able to and all of this interaction could be updating that CRM record on the back end so that you can see what's going on with this. We really want to thank you all for joining this AI webinar and hopefully you had um an immersive experience in a short amount of time and fun along the way. Thank you

all. Cheers. Cheers. Thank you. [Music]

2025-05-18 06:42

Show Video

Other news

Сравним питание предов от сети и DC-DC конверторов 2025-06-01 04:37
Nvidia Tie-Up Helps Vietnam Tech Tycoon Bet Big On AI 2025-05-30 06:08
Nvidia Opens AI Ecosystem to Rivals, Apple AI Struggles | Bloomberg Technology 2025-05-25 18:40