Hello and welcome to our latest LinkedIn Live session with Tray.ai. My name is Michael Douglas. I'm a Senior Product Marketing Manager here at Tray and I have a heavy focus on Merlin Agent Builder and helping our customers understand the possibilities of Merlin Agent Builder and how it can truly transform their business.

What we're going to do today is, as part of Merlin Agent Builder, we have built what we call the Tray Agent Gallery and it's basically a large breadth of agents, up to 20 agent demos to help showcase the true possibilities of Merlin Agent Builder and again how it can fundamentally change your business across a multitude of use cases. And with that, I'd like to introduce my colleague, Luke Smith. Hey Luke, how are you doing?

Hey, I'm great, thank you. Yeah, great to meet everyone. I'm Luke, one of the Principal Sales Engineers here at Tray. So all of the agents that we're going to feature today are real and ones that I've built on the Tray platform. So really one of the big goals of this session is to go beyond the theory and show you what's actually possible with some of these agents. So looking forward to getting stuck in and taking you through some of the examples that we've got.

Great. Thanks, Luke. So as Luke was saying, these aren't just a generic overview of agents. These are real specific working agents that are helping to transform businesses that a lot of our customers have rolled out over the past year. And we have three demos to give you a highlight of some of these agents that are out there. One is HR, the second is a Knowledge agent, the third is a Data Analyst agent. And as I said, all of these agents are really about tackling business challenges and delivering real business results.

And so

we're going to walk through three key things of why we built it, what it solves, and how it works. And so we're going to start with the HR agent. So, everyone's been there where you join an organization or you're working in a company and you've got a lot of questions about when can I change my PTO? How do I change my PTO? How do I book it? How do I find out my parental leave? How do I find my payroll policies? There's just a whole host of questions across the organization that employees are constantly asking HR teams.

And they're usually under a lot of pressure to be able to respond to all of these requests in a timely manner. And so all of this repetitive manual work really does bog those HR teams down and really stops them from getting along with more of the critical aspects of their job. So what we're going to do is we're going to go across and Luke is going to walk you through the HR agent and all of the core capabilities that are put in place to help alleviate a lot of these repetitive tasks. So Luke, over to you for the demo.

Yeah, awesome. Let's get stuck in. So let's take a look at how I built out this HR agent. So as you'll see with all of the agents that we go through today, every one of those will have an agent scope. So an agent scope effectively tells the agent and think of that as overarching instructions that you might want to provide your agent.

So if there's specific guidance, maybe telling it what type of agent it is, it can really help it and allow you to tweak different parts of the agent's kind of behavior to help you achieve various different use cases. What you'll also have full flexibility to do is always adjust the model as well. So at the moment I'm using Tray's native model, but at the bottom here in a few clicks, I can also switch across to any of the other vendor models if I wanted to as well.

Now, different models are good for different use cases, but you have full flexibility with our agent builder to actually be able to bring in and use your own models if you so wish as well. Now, the next large part of the agent that's super important is going to be the tools. Now, tools are pieces of functionality that you add to your agent that allow it to effectively do different things. Think of these as tools in a tool belt if you're a builder. The idea being with a tool is that you have a tool description. Now, a tool description basically is a description that you provide to the agent to tell it what type of thing that tool is designed to help with.

Each tool will then also have a consistent input and output schema that you can define as well so that it has an understanding of the information that it needs to pass to that tool and then information on what it expects to get back from that. What this allows the agent to do is to take these tools into account so when you ask it a question it's going to figure out, okay, well, I've got these eight or nine tools here that I've got, which ones am I going to use to potentially come back with a response for a prompt or a question that's being asked by the user. So in this case, you can see that for my HR agent, I've got eight or so, maybe nine tools, that I've got and these are two different pieces of functionality.

So, for example, all a tool is is a low-code Tray workflow. Now, as you can see in this case, I've got my view holiday request tool. This allows my agent to, if it needed to, find out what my current holiday requests are in this case the HR system being Workday.

So, as you can see, it's a really simple workflow. It's reaching into Workday to pull back my holiday request and then we'll have an input schema that's going to allow it to tell the agent what it needs to pass through, and you'll find the same with the other tools as well. So, for example, if we're in a scenario where the agent can't help the user and maybe needs to escalate it to a HR person, we've got a tool here that's going to make use of the Jira connector to be able to create that issue for us as well. And the general idea is that, because these are Tray workflows, you've got a tremendous amount of flexibility about what you can build into these tools and they're really composable in the sense that you can get started with an agent with just a few tools and start to add to it over time as you define more and more functionality that might be helpful for your agent to be able to do.

What's really cool, though, is that of course reaching into the system and pulling back data is really handy, but what you can also do is have tools that can actually take the action as well. So in this case, there's a tool that allows my agent to submit holiday requests for the user as well again using Workday as the system here, but not only is it reading information, it's actually able to take that action in those systems directly as well. But let's go ahead and actually take a look at how we can interact with this agent.

Obviously, the other important part when we're looking at agents is how are you interacting with it? Are you maybe deploying it into a Slack or a Teams for a natural language conversational type agent or is it something that maybe is sitting behind the scenes and perhaps it's just going to run automatically, maybe listening for changes in the system. You've got a lot of flexibility about where your agent sits. In this case, I've got my agent deployed over in Slack so I can chat with it.

Now I'm going to start with a relatively simple first question, so I'm going to tell it I need to fly from London to Dubai and I fly business. Now this prompt is going to get the agent to do a few different things. Now, one of the things that I've given my agent access to is a set of knowledge documents or documentation around our HR policies and guidelines.

So what we've done with this agent is basically built out with a functionality that we call our data sources that I'll show you in a moment that's reached into my, in this case, a Google Drive folder and actually ingested in all of those HR documents and guidelines for us. And what that means is that when we come back and it provides us with information, it means that if it's been or needed to based on the prompt that was being asked use some information from those that knowledge base is actually in this case pull back some information from our expense reimbursement which is one of the documents that ingested in into that knowledge base. And you can see it's actually very helpful in this case, based on the information it's provided, and according to the company's travel policy provided us with a helpful link directly to that document that I can review as well.

And so you can see what that's allowed it to do is actually have a really good grounded response in the sense that it's basically saying “Hey, your flight is over eight hours so the chances are that the flight duration probably is not going to qualify you for business class based on the current HR policies and guidelines there as well.” So what's really powerful with this is when we look at the data sources, just to show you what that looks like, it's really easy for you to be able to add a data source onto your agent. So under the agent builder, you'll have a data sources tab and you can click on add data sources, and so in this case, my HR guidelines sit in Google Drive.

So all I need to do is click on the Google Drive option here. I can then provide it with an authentication and I'm going to select the folders that I want it to sync directly as well. So in this case, it might be the HR guidelines, for exampl,e and I can go through and configure that accordingly if I needed to.

Click on save and we'll see that it's now set up a brand new data source for us and so what that does is that when I start the syncing process and it's going to look into that Google Drive folder and now ingest in any of those documents. Now what that was using and what it is using is our native vector tables or vector databases on the platform so it's effectively building out a really nice RAG pipeline for us where it's going to take those documents, create the necessary embeddings, and add that to our native vector database and so now it's a knowledge base that my agent won't have access to and one of the tools it has is the ability to reach into that internal knowledge service for us where it's actually going to go ahead and pull back that information for us directly from that data source which is great. And what that means is it all comes together really nicely so that now my agent has that extra context around those HR guidelines, so when it is deciding what to respond with, it's got that extra information that you can use to help with as well directly.

Let's go ahead and take a look at a secondary example of that. So what I'm going to do is I'm going to ask it to book some PTO. So I'm going to ask it I would like to book some PTO from, let's do Wednesday to Thursday this week, please. And so what we've got here is a few things will happen in this case is there is a PTO guideline so how often or how much in advance you need to provide the guidelines before you book. In this case, it's going to be I believe two weeks or a week we've set it to in these guidelines and what we'll see here is that the user is of course one of the things sorry the agent is able to do is actually book the PTO for us but because it's got that extra context from our knowledge base it's going to potentially push back in this scenario to say that the PTO is not actually within the guideline policies.

And so we can see, in this case, what the agent has come back with, is its reference that PTO policy that I mentioned and it's like hey according to our PTO guidelines requests, less than seven days in advance will require manager approval. They've actually gone ahead and created us a support ticket, SUP-55, so we briefly saw a moment ago that tool that's able to create that ticket in Jira. So what it's done is it's escalated that to our HR team because I want to book a PTO that is out of that policy so instead of just going ahead and doing it on its own, what it's done is it's pushed back and all of that is from the fact that we've ingested in those knowledge articles using that data source functionality to give my agent that really good grounding of knowledge that it can make use of as well.

Finally, let's put in some valid dates. So I'm going to bring up a final chat and I'm going to ask it to go ahead and book us. “Can you book me some PTO keys from October 1st to October 2nd?” So we'll give it plenty of time in the future to go ahead and book that and what we'll see is that the agent is actually able to take that action. So we saw just a moment ago it was taking that action by creating that escalation ticket but, in this case, it's actually able to go ahead and book us that PTO as well as long as it's valid from a guideline point of view and any necessary and extra things related to specific PTO policies as well. And so, in this case, you can see it's come back and gone “Hey, I've successfully been able to process your PTO request.” Again, it's provided us that reference back to our PTO policy but, because it meets that criteria, it’s more than happy with that so what we should see is, over on our Workday side, if I give that a quick refresh, what we should see in just a moment is that new PTO being booked for I believe it was in October that we popped that in for and you can see it's popped that request in there for the 1st and 2nd of October for us successfully.

So really, really great. We can see a few different use cases for that agent and it was able to use that knowledge base but also take into account the action for this as well. What we can also see and the final piece is that, under the logging section, all of the interactions with our agent from different periods of time, we have a full look and full visibility into every execution. And then each step and the logic that was followed through directly as well so this gives me full visibility so I can see kind of what the agent did in each and every one of these executions as well.

Great. Thanks, Luke. So when you were designing it and going to build this agent, what was kind of the end goal that you wanted to get to?

Yeah, I think like you mentioned at the start and I think one of the big things is there's the HR guidelines policies. There's often a lot of information there and a lot of them are repetitive queries. Everybody's probably got a repository of the 12 or so different HR policies and people don't always want to read through it so there's common questions that always come up. And one of the big things when designing the agent was giving it a good knowledge base that it can use so that you can ask any questions about any of those policies and get a really good helpful response back. Then there's also equally the scenario of things like booking PTO, where it's obviously a very common thing, but logging into the HR tool, go into the book PTO section, trying to fill in the dates, kind of trying to have it living in somewhere where people are working day in day out in the case of Slack. So I can just ask it a question and go “I need to book some PTO for next week” or you'll need a sick day and you need to book it straight from your your bed on your phone type thing. Just message the agent quickly and say “I need to book it” and it's going to take all of that action for you without you needing to leave those systems there. So it's just really all about giving the agent a few different things but of course the beauty here is that there's a tremendous amount of extensibility with all these agents. Where this is just a couple of examples of things you can do, but there's a lot more that you could add to your agent with those tools that you're building.

And we saw it being deployed in Slack. How did you go about making it Slack native?

Yeah, so one of the nice things that we have when we're looking at what makes up an agent is what we call the interaction channels. So it's where is your agent living. Now, a lot of people will have that they want to have an agent that's conversational. They want to chat with it so we have these interaction channels that exist for things like Slack, for example, that you can make use of and Teams that allow you to deploy your agent in multiple places equally as well. You can also very easily build agents that don't necessarily also need to rely on those conversational interfaces but Slack and Teams are one of the big ones that we, of course, see and hear all the time. So in this case, I made use of the Slack interaction channel when I was building my agents so that you can chat with it directly in there.

And I'm sure there's a lot of people out there watching this and saying “Yeah, HR, there's a lot of touchy issues around giving employees access to sensitive information, allowing agents to take action. What is the agent not doing, right?

Yeah, I think this comes really down to the design of the agent. So I think as we saw in the demo, one of the things is, because we have this agent was grounded in that knowledge around guidelines and policies, it means it's got this extra contextual awareness about things that it should or shouldn't be doing. So in this case, it was like someone trying to book a PTO in a few days time and they're like, well, no that's not within the policy and so it escalates it as necessary. So of course it's going to go to the HR team but it means that what does come through to the HR team are going to be those really important escalated issues that are maybe slightly more nuanced and because you do have complete control about the tools that you've built on the platform, it does mean that you can control what information someone may or may not be able to access, what actions can they or can they not take in line with either your policies and what your design goals are for that agent. So plenty of options and we do see people starting with that first point of “I'm going to get the agent to maybe respond where it can and then escalate that as necessary.” And then over time, they might add more and more actions onto their agent.

Yeah, super important to keep that human in the loop option versus just a an agent that is working independently, so thanks for running through that Luke. So the next agent demo that we're going to move on to is the knowledge agent. Right? The knowledge agent is very much a a big focus for a lot of organizations because there are is a lot of siloed data across multiple systems and databases and being able to unlock that knowledge is a real hard nut to crack and being able to not only do that but allow employees to self-serve in this knowledge and ingest this knowledge on their own time is also a major priority for organizations. So the problems that we're trying to solve with this agent, Luke, when you're out there talking to customers and they're starting with knowledge agents, what is their priorities?

Yeah. I think one of the big things with knowledge agents is there's a ton of knowledge that sits in so many different places that have a tremendous amount of helpful data. And one of the big things when looking at this this agent that we built was - Jira is a great example or your ticketing system - there's a lot of historical information that sits within there that is up to date. It's going to be reflective of maybe what questions are answered and, of course, you can go and manually look at it but there's a tremendous amount of value there around keeping documentation up to date trying to help it prevent it from going stale. And so one of the big things when we're looking at is how do we access this knowledge and how do we make it accessible to the agent so it can start to help make and ultimately take some action with this data that might be being pulled from multiple different places.

Yeah, so that's interesting. That's one of the critical elements, right, is we see a lot of solutions out there in the marketplace that are providing surfacing knowledge but aren't able to take action and being able to also take complex action, they might be able to take simplistic action, but being able to take complex action across the organization is really significant and I'm interested to hear from customers, are you seeing that as a priority, Luke, or what's been your view on it?

Absolutely and I think that's where, like you said, the real power comes in is, of course, great, the agent can tell me about all of this wonderful knowledge but actually how can it then make use of that to do something really helpful, right? And so in this case, in this agent, we'll see that it's going to be able to kind of take and update an FAQ page so it's a live breathing document. And all of that starts with that knowledge, but it's great, of course, we can surface that, but it's taken that next step to really start unlocking that power and actually being able to use that to have some really tangible benefits off the back of that.

Awesome. So look, let's not beat around the bush anymore. Let's just dive into the demo, Luke, if you wouldn't mind.

Yeah, absolutely, yeah, let's get stuck in.

Let's jump in and take a look at this knowledge agent in a bit more detail then. So same as our first agent that we saw, this agent has its own agent scope which is specifically giving it some guidance around the type of agent that it is. In this case, I've also tweaked around the AI model that this one's using so it's actually using one of the AWS cloud sonic models as opposed to maybe the Tray native one from the first agent. Again, you have a lot of flexibility about the model that's being used. From a tooling perspective, this agent has around eight or so tools that are going to be used to help it get information from a variety of different sources. Now, same as before, all of these tools are Tray workflows so, for example, this “get full ticket details” one is a tool that the agent can use to get a full set of information around a particular Jira ticket. If we wanted to as well, we then also have tools that are going to allow us to reach into, for example, Notion, to be able to pull back a particular set of pages that we might be interested in updating. In this case, from an FAQ document point of view, my use case is that I want to update our FAQ page with some updated information. And so the FAQ page currently lives in Notion so this tool helps it get access to that and then we also have a tool here that's going to be able to reach into Product Board to get feature requests. And so again, we're using the extensive connectivity of the platform, we're adding these as tools so that my agent has access to a wide variety of different sources, and is actually able to take action. As well with this tool here that's actually able to go ahead and update that FAQ page directly for us as well and you'll also see one down here for creating us a new Jira project too. So all of these tools are going to be used by my agent across a wide variety of systems and so my use case is that we currently have an FAQ page here with a couple of generic issues that come through and what I wanted to do is I wanted to go ahead and look at our historical Jira tickets and I wanted to recommend some FAQ updates based on common questions that we're seeing across our tickets. Now, to help with that, one of the things that this agent has ingested in and, from a knowledge-based point of view, is that it's taking a look and ingested in those previous historical Jira tickets for us. So similar to our first agent where it was looking at Google Drive, in this case, what we've set up is a data source. In this case and again, really easy to set up, you just select Jira Cloud, in this case, to point to those existing tickets. As we saw earlier with the first agent, it was Google Drive but this allows it to ingest in that knowledge for us. Of course, one of the big things to highlight here is that you do have custom data sources as well so any of the systems that you might want to ingest knowledge from outside of the ones that you see on the screen here can easily be done directly for us as well.

So I've got my Slack deployed or my agent deployed here in Slack again and I've got my agent waiting for a prompt so I'm going to ask it to take a look at the last six months of tickets related to feature X because we want to go ahead and update the FAQ page. So what we should see here in this particular circumstance is that the agent is going to make use of that knowledge base that we've ingested in those historical Jira tickets to take a look through maybe any common questions. It's also got that tool that we saw that's able to reach in and get the current FAQ page. And so all of this will come together quite nicely in this agent using those various tools we've given it to help give all the information that it needs to be able to come back with a suitable and strategic recommendation on how we should update this FAQ page.

So you can see it's come back for us quite nicely and it's analyzed over 100 tickets. It's categorized them into five major areas and there's a few top issue categories. So it's done a really good recommendation for us on all the information that it's found and that initial as for that as well. So it's provided some recommended FAQ updates, added some specific browser compatibility lists, network troubleshooting. And so you can see it's gone ahead and provided us with a question of “Hey, what do you want to do with this information?” So I'm going to go ahead and tell it to update the FAQ page please.

So because we've given it that tool, that's actually able to go ahead and take this action for us, What we should see is that it's going to be able to actually update the FAQ page for us directly as well in just a moment but as you can see because it was able to reach in and get that existing information, it's got the information in those historical tickets again, it's really good knowledge that the aid has been able to make use of to come back with some recommendations for us about some updates to our FAQ page. And so as we can see, our agents come back with a detailed analysis document that we can click on so if we give that a quick look, that should pull up a really nice breakdown of all of the analysis that it did for our reference. So the things that it looked at and some of the proposed FAQ updates and where it pulled it from which is really really great it's also provided us with an update to the FAQ page which is fantastic.

So if you take a look over on the FAQ page itself, if you get this a quick refresh, we should see this populated with a lot more information. So things like the supported browser versions was one of the recommendations. You can see it's gone ahead and taken that action to update our FAQ page for us there directly as well so not only has it been able to take all of that information, all that knowledge and come up with a suitable recommendation. Of course, we've been taking it to the next level with this agent to actually go ahead and take that action directly for us there as well in this case updating the FAQ page and creating that summary document for us that we can make use of. Of course, one of the nice things is the flexibility here is that these tools again are just those Tray workflows so wherever this data might be sitting in your circumstances for different systems, it gives you a lot of options for what you would like to connect to there directly as well.

Let's go ahead and take a look at a second example then for this particular agent and we'll see how it's able to help with that as well. So in this case, what I'm going to do from a prompt point of view is I'm going to ask it the question of I wanted to go ahead and take a look at our analysis briefings for the last few months and I wanted to see how Acme Corp is doing against competitors and potential areas for improvement, taking a look to see if we have similar Product Board feature requests as well. So again, another quite loaded prompt here with a lot of different things that I'm going to ask for but, again, using those tools that we saw just a moment ago, it's going to be able to reach into some information that we've ingested into our knowledge base from analysis briefings. It's going to be able to take a look at that and use that Product Board tool to get our current feature requests. And again, give the agent all of the context that it needs to come up with some recommendations and some project plans for us around some potential product improvement areas.

And so as we can see it's done some competitive analysis. So it's taken into account those analysis scores, it's given us a breakdown into key performance areas, and it's also given us some recommended project on focus areas. As you can see, it's also asking “Hey, I can do the action of creating this as a project plan for you.” So let's go ahead and get it to this as a project plan so again just as the previous example, of course all of this knowledge is great it's given us great recommendations but now it's taking it to that next level and actually making this into a tangible outcome, in this case, by creating us a project plan over in Jira with this analysis.

Now again, this is all coming from the knowledge base so we ingested in some information that we have pulled from our analysis briefings that came through in a variety of different document formats and again using that data source functionality that we've seen across the two agents that we've got today is really powerful functionality to help give your agent that extra knowledge. But then, of course, as we're seeing it taking that next step with the actions being taken there as well.

And so as you can see, our agents come back and it's created us a nice project plan. So it's given us a detailed project plan in Google Drive so if we give that a quick link, we should be able to see what those recommendations are from a project plan point of view for us as well. So as you can see, it's given us an executive summary, broken it down into sprints. So just before it's created that and then it's also created us a Jira project. So, again, if we select that one, let's head on over to Jira.

What we can see is it's created us some timelines and some work items and actually gone ahead and created us a couple of sprints related to those key identified areas that our knowledge agent was able to pull. So security, compliance, data handling, and the infrastructure support. So again, really cool to see it taking this action across this these various systems. In this case, creating a project plan for us and again all that stems from the tools that we've given our agent access to here to reach into these various systems to be able to take that action as necessary and then also making use of that data source functionality to really be able to easily bring in that knowledge from a variety of different sources to really help the agent with its recommendations for us.

Awesome. Thanks so much, Luke. I really appreciate that. Great demo. I'm sure our viewers got a lot of great knowledge from that, no pun intended. So moving on to our third and final agent demo, we have the data analysis agent. So there's a ton of manual work that goes into building all these dashboards and reports across the organization. It can be quite tedious and time consuming to actually go and extract a lot of that data across the organization and try to make sense of it, how is this agent different from a simplistic BI tool or dashboard or not simplistic. Some of these can be pretty complex. What's your take on it?

Yeah. I think one of the big things when I was making this agent specifically, of course, there's tons of visualization tools out there and they're really great if you're pointing it at one source of data, right? You've got some data sitting somewhere and you want to visualize. The tricky thing is that, okay, that's really great but what if I want to have my visualization reflect data from five different places and have some sense between that? That's where I think when I was building this one out is thinking about where all these different places. Again, where that data is sitting, how can we bring it together, and make a visualization that can look at some of the relationship between it, do some analysis on that, and figure out some good things, bad things, when all of that data isn't all necessarily sitting in the same place and still be able to make a powerful visualization from that. I think of course, visualizations can actually also be quite a tricky thing to build if you're not familiar with it and so bringing it to something that's just, hey, ask it some questions, get some data from a few places, and ask it to make a visualization for you. It can't get any more simpler than doing it in natural language.

Yeah and that's a great example of what kind of actions can the agent take beyond the analysis. Building those visualizations on your behalf is, when you think about it, the amount of time it goes into building it with a BI tool or some sort of dashboard is really considerable time savings So with that, let's just jump in.

Awesome. Yeah. Let's get going.

Let's jump into the final agent of the day. So we've covered well in the previous two agents, the agent scope and the model. This agent, specifically, because it is a data analysis agent, has a few tools to connect to a variety of different systems to help it pull different pieces of data so that it can do that analysis on top. So in this case, we can see that it's got a tool to be able to reach into Salesforce, our CRM.

We've got one here that's able to reach into Redshift to pull back things like NPS scores about particular users. We can also grab campaign data with a tool here that's connecting into HubSpot. And all of this is coming together quite nicely within our agent to give it a lot of different places it's able to pull the additional information from. What this is also able to do is actually create this as a visualization for us as well.

So there's a tool that's actually able to hook into Grafana to create a visualization on the fly based on what the user is asking directly. Now, we saw in the previous two agents as well that we deployed those into Slack. What you can also do, which is really nice with an agent, is you can actually chat directly to it when you are building them out. So you don't have to immediately jump into deploying it into Slack or Teams just to chat with it.

There is a nice test interface that you can use, and I'm gonna use that for this last agent of the session here. So, I'm gonna ask it a question. I'm gonna ask it to take a look at the open opportunities that we have, and I want it to include information and summarize it for me into a table. So this one's gonna reach into things like our CRM and look up the contact information.

It has provided a nice, simple summarization for us.

And so the agents come back and provided us a nice, helpful look at the current open opportunities from our CRM and put it into a table for us directly. He's also provided us some summary statistics. And you can see it's included that source of our CRM lookups at the bottom here. So that's really great.

Obviously, that's just pulling it from a single place. So let's try and pull some data from a few extra places as well. So I'm gonna ask it, in this case, to add in the interactions with the contacts for each of these opportunities with any campaigns that we've run. I also wanted to provide some data on the page viewed from our web analytics, and then also flag any provided NPS scores by the relevant contacts on these opportunities that might be at risk.

So basically, adding on three or four extra data sources that I want it to consider in its analysis here. And what's gonna happen is because we've got those tools that we've given it that can reach into these systems and pull back our information and search for the relevant information, it gives our agent that information from a variety of different places to help it perform that analysis.

And so in this case, you can see it's come back and has added us a lot more information to our table. So it's been able to pull in the NPS scores, the recent campaign activities as well, and then also taking a look and highlight those that might be at risk from those low NPS scores. And if we scroll down to the bottom, what we can see is it's pulled in the Salesforce lookup. It's pulled in that HubSpot campaign data, the campaign Intelligence.

It reached into Redshift for those web analytics and NPS scores. And so it's pulled in from a much larger amount of data sources, again, all from those different tools that we've given it access to. As you'll see as the final point, it's also recommended in asking here if I wanna go ahead and create the Grafana dashboard from this information. So I'm gonna go ahead and say, yes, please.

And we'll take a look to see how the agent is now able to actually take that action and create a nice visualization for us with this information in. And we can see how easy it was to pull data from five or six different places, run some brief analysis on that, and then get a really nice, helpful visualization dashboard at the end of it. And so as you can see, it's come back and created us that dashboard successfully. So it's got the total opportunity value, the at risk ones.

This also provided us with a nice link here. So let's go ahead and click that, and we'll just take a look to see what that Grafana dashboard looks like for us.

And so as we can see over on this tab, we've got our nice Grafana dashboard that was created on the fly directly there. So not only did we start with a relatively simple query, we added to it iteratively with our prompts, added on a few extra systems we wanted to check. And it's been able to take that action by creating us that opportunity dashboard over in Grafana for us, which is great so that you can summarize all that information for us on the fly there too.

Super. Thanks so much, Luke. Great way to finish up our third demo, and I'm sure a lot of people are gonna be looking at this agent instead of a lot of the dashboards and BI tools that are out there. So super exciting.

So you've seen three different demos for three different types of use cases within three different business settings, but there's a whole host of other agents that we have built for sales, for marketing, for IT service management, and beyond. And so there's really much more functionality and business value that you can go and see if you wanna check out the Tray Agent Gallery. I'd really encourage you to do that.

And if you have any other questions, feel free to contact your CSM if you're a customer or, fill out the form online if you're not. But I'd really like to take this opportunity to thank Luke for building out our demos and walking us through it. And also thank you so much to you for giving us your time today and tuning in and seeing the demos. We look forward to talking to you on the next LinkedIn Live. Thanks, Luke.

Thanks, everybody. Cheers. Bye.