How to Build Jira AI Agents with Atlassian Rovo (Full Transcript)

[00:00:00] Speaker 1: In this video, I'll show you how to build AI agents in Jira using Atlassian Rovo step by step. We'll start with a pre-built agent, then create our own custom AI agent, make it smarter with company knowledge and skills, and finally automate it. Atlassian sponsored and supported this video. I'm Kevin, and let's dive in. To get started, let's look at some of the pre-built agents available in Atlassian Rovo. Here I am in Jira. Now, we use Jira here at the Kevin Cookie Company to keep track of customer feedback. As you can see, customers have plenty of opinions when it comes to cookies. Let's see if an agent can help us make sense of all of this feedback. In the top right-hand corner, let's click on Ask Rovo, and this opens up Rovo, Atlassian's AI assistant. Now, I want to identify the key themes in this feedback along with how often they show up. Now, right down here, I could type in a prompt directly into Rovo, but Atlassian already includes an agent that's designed specifically for this type of task. To switch an agent, let's click on this dropdown right here, and you'll see a list of available agents. Right down here, let's click on Browse Agents, and right up on top, let's click on Browse All. Here, you'll find over 25 pre-built agents. Now, some help pull insights from meetings. Right here, you could even generate OKRs. If we scroll down a little bit more, here's an example where you could write release notes, and you have many more. Now, for this example, I want help identifying trends in customer feedback, and right here, there's the Jira Theme Analyzer. Now, as the description says, it'll find common themes from a group of issues in Jira. That's exactly what I need, so let's click on this. At the top, you'll now notice that I've switched from Rovo to the Jira Theme Analyzer agent. One of the benefits of an agent is that it's built for a specific purpose. In this case, the Jira Theme Analyzer is designed to review all these Jira issues, group them into themes, and then summarize the results. That means I don't have to figure out the perfect prompt every time I want to run this analysis. Now, right down here at the bottom, you'll notice a few suggested prompts that can help us get started. Let's try asking the agent to identify all the different themes in our Jira issues. Here, I'll type in my prompt, and then let's run that. In just a few moments, the agent analyzed all 31 customer feedback tickets and grouped them into common themes. Now, let's switch into full screen mode so we can see this better. Right up on top, I'll click on this, and here, let's click in the full screen. Here, I get a table with all of the different themes. I also get a sense of how often those themes show up. Right over here, we have a summary, and on the right-hand side, we get links back to the original customer feedback. Now, looking at all of these issues, it appears that shipping and delivery comes up most often. Now, that's probably something I wouldn't have spotted nearly as quickly by reviewing all of these issues one at a time. Now, this is a great way to quickly identify trends without manually reviewing every single piece of feedback. So far, we've looked at how you can use existing agents, but what if you have a workflow that's unique to your team or business? Let's now build our own custom agent. Back here within Jira, in the top right-hand corner, let's click on the robo icon, and then click on the agent selector. Down here at the bottom, you'll see the option to create an agent. Let's click on that. Now, alternatively, you can also navigate to the website that you see at the bottom of the screen. I'll click here. This opens up Rovo Studio, where we can create our own agents from scratch. For the Kevin Cookie company, let's create a customer support triage agent. When a new support request comes in, we'll have the agent categorize the request, determine the priority, recommend next steps, and draft a response for the customer. So over here, I'll type in my prompt, and then right over here, let's run that. In just a few moments, here, Rovo generates an agent name. We also get a description for this agent, and right down below, we have all the instructions for the agent. Now, honestly, that's a lot faster than starting with a blank page. Of course, I can always go through this, review it, and I could customize any of these instructions, but I think this gives me a really good starting point. Right down here, let's click on build agent. We've now built the agent, and this drops us into the agent details in Studio. Now, this all looks good to me, so in the top right-hand corner, let's click on publish. Now that we've built our agent, let's test it out. Back here within Jira, in the top right-hand corner, when I click on Ask Rovo, let's click on the agent dropdown, and right here, you'll notice that we now have a new agent, the customer support triage agent, the one that we just created. Let's select that agent, and right down here at the bottom, I can now prompt my new agent. Now, I can reference any one of these support tickets directly here, but let's actually make this a little bit easier. Over here, I'll open this ticket directly. This opens up the support request details, and I can see that the customer selected expedited shipping, so they were expecting it faster, but it actually arrived late. That's completely unacceptable. So we need to figure out how to prioritize this and also draft a response to the customer. Before we do that, though, let's expand this so we can see things a little bit better. I'll click here. Within the issue, right here, you'll notice that we can run an agent. I'll click on this, and there's that agent that we just created. I'll select this, and that'll kick off the work. Here, the agent reviews the request, and first off, it chooses the category. This is a delivery problem. It also assigns a priority level. Here, it's saying medium, and it also gives me the reasoning for why it chose that. Here, we have next steps, and at the bottom, it even drafts a response that we could send to the customer. Now, the beautiful thing is, instead of manually reviewing every single request, the support team now has a structured starting point that they can work from. Now, down below, let's now save this comment. Here, I'll scroll down. This looks good, and let's click on Save. This works really well already, and we could share it out with the team. However, we can also continue improving the agent by giving it additional capabilities. Back here in Atlassian Studio, over on the left-hand side, let's click on Agents, and right here, let's click on View My Agents, and here, I could see the customer support triage agent that we created earlier. Let's open it. Right here at the top, you'll see instructions that guide the agent's behavior, and if I scroll down, here, you'll notice that we could add additional capabilities. We have Skills, Knowledge. Here, we have Conversation Starters, and then More. Now, let's start with Knowledge. To do that, in the top right-hand corner, let's click on Edit. Then, on this screen, let's scroll down, and here, we see Knowledge. I'll expand this, and let's add Custom Knowledge. I'll check this, and then let's click on Add Knowledge. One of the easiest ways to improve an agent is by giving it access to company-specific information. So, right about here, let's add a Confluence page. I'll check this box, and let's select Content Under, and right here, I see the Kevin Cookie Company Customer Support Playbook. This includes our refund, shipping, and missing item policies. Just as an example, let's imagine that an order arrives, let's say, more than two days later than the estimated delivery date. In the policy, we provide customers with a 15% discount code. Without access to this information, the agent would have no idea about this policy. Now that I've added this page, down below, let's click on Add. Right up on top, let's also update the instructions with one additional line to make sure that the agent references the custom knowledge that we just provided. Once you do that, in the top right, let's click on Publish. At the top, we can now test the updated agent. Over here, let's click on Test. At the bottom here, in the prompt field, let's type in, my cookies arrived three days later than the estimated delivery date, and then let's run that. Earlier, the agent would have given us just a generic recommendation, but right up here, we can see that it's referencing official Kevin Cookie Company policy. Here, it offers the customer a 15% discount, and it's also providing the coupon code, SorryCookie15, just like I expected. Let's now make our agent even more capable by giving it the ability to take action. And one way to do that is by adding a skill. Back here within Studio, I have the Customer Support Triage Agent open, and up on top, let's click on Edit. And on the Edit screen, let's scroll down, and we'll see the section for Skills. I'll expand that. Here, we can see we currently have no skills. Let's click on Add. Whenever the agent sees a refund request over $100, I want it to notify the Customer Support team in Slack so they can review it right away. Right up on top, we can search for skills. Here, I'll type in Slack, and I get all of the different actions that I could take. Now here, I see the option to send a message. That's what I'm looking for, so let's select that. And down below, I'll click on Add. And right here, we can see that we added the first skill. However, we still need to tell the agent when to use this skill, just like we did with Knowledge. So right over here, let's go up to Instructions, and I'll add a sentence. If a customer requests a refund greater than $100, send a message to the Customer Escalation Slack channel that includes a summary of their request, the reason for the escalation, and the recommended next action. Now that I've entered that, right up on top, let's click on Publish. Let's now test that. Right over here, let's click on Test. Right down below, I'll type in the following prompt, my order never arrived. I'd like a refund of 150. Let's send that. Right up on top, let's send the message. Right here, it detects that the refund request is over 100, so it escalated it to Slack. Let's have a look in Slack. Here in Slack, we can see the notification. That way, the customer support team knows to look at this request right away. Running the agent manually is helpful, but we can make this even more powerful by automating the process. Whenever a new customer support request is created in Jira, I want this agent to automatically review it. Here in Atlassian Studio, over on the left-hand side, let's click on Triggers. A trigger determines when an agent should run. For example, I could run an agent on a schedule, when a work item changes status, or when a new issue is created. Right over here, let's add an automation trigger. For this example, I want the trigger to run whenever a new customer support request is submitted. So over here, I'll type in the trigger. Then for the instructions, let's enter in analyze a new support request and add your findings as a comment on the issue. Then down here, let's add the trigger. Here, Atlassian automatically builds out the automation for me. Right here, we have step number one, and let's select the Jira space that it should monitor. So over here, I'll select My Team. And next, we have the agent. And this will review any new support requests that come in. This will use the instructions, the knowledge and the skills that we added earlier. And then lastly, it'll add its findings directly to the issue as a comment. Now, this all looks good to me. So right up on top, let's click on Save and Enable. Now, let's see it in action. To do that, let's go back to Jira. Down below, let's now add a new support request. Discount code won't apply at checkout, and let's create that. Right up on top, let's click into it. Now, over on the right-hand side, we can see that the agent has added its analysis to the support ticket. Now, instead of waiting for someone to manually review every request, instead, the agent works as part of the support workflow from the moment a ticket is submitted. Before we wrap up, let's take a quick look at how you can share your agents with others. Within Studio, over on the left-hand side, let's click on Users and Collaborators. Here, I can make the agent available to everyone in my organization, or I can limit access to specific users. If you'd like to learn more about Atlassian Rovo, check out the links right down below in the description. Thanks for watching, and I'll see you in the next video.