Hello, and welcome to this video entitled Your First Rovo Agent in 15 Minutes. I’m the Vice President of Sales and Marketing here at SPK & Associates. We’re an Atlassian Gold Solution Partner, and we’re going to try to do this very quickly so that you can build your first Rovo Agent.
What Are We Building?
The first question is: what are we building? What is Rovo? What is a Rovo agent?
Before we build an agent, it’s helpful to understand the context of what Rovo actually is. Think of it as an AI-powered teammate that’s trained to perform a specific job within your organization.
It uses Atlassian Intelligence and the Teamwork Graph, which we’ll talk about shortly, to understand the context of your project needs, things in Confluence, and other connected systems. That enables three exceptional things to happen.
It’ll find information across all of your tools.
It’ll learn from your organization’s knowledge.
And then it can act by helping you automate tasks and streamline workflows.
So instead of searching across multiple applications like Google Drive or SharePoint, you can simply ask your Rovo agent to do some of that heavy lifting.
Understanding the Teamwork Graph
I mentioned the Teamwork Graph. This is really where I think a lot of the power with Rovo comes from.
To understand that, you have to understand the underlying foundation of the Atlassian Cloud Platform. Atlassian’s AI and overall platform is optimized to support the core principles of the different types of teams on the screen here.
It helps them share and invest in the knowledge that’s shared amongst the team and bring the right information in order to plan and track the work that everybody needs, but also within context.
This enhanced AI is embedded throughout Atlassian’s platform, especially in their Rovo agent, so it’ll help teams benefit from a personalized AI experience to achieve better outcomes wherever they’re working and whatever system they’re working in.
This is really what makes Atlassian’s solution powerful: the Teamwork Graph.
The Atlassian Cloud intelligent data model helps you effectively leverage all of the data within your teams to help get work done better. Atlassian calls this model the Teamwork Graph.
It’s a unified platform where teams are sharing more information, even without having to do anything other than just doing work. That allows people to get more personalized and accurate responses from the work that they do.
The Teamwork Graph connects all of your work, your teams, your messages, and your goals through both Atlassian’s solutions and tools, along with third-party tools that your company may already be using.
Unlike a knowledge graph that just gathers data, the Teamwork Graph actually understands the context of these objects, their unique types, and, more importantly, how they’re related. It connects you to what you do, who you are, how you work, and what’s ultimately important to you.
This translates into a more tailored experience across all apps, especially when using Rovo.
How the Teamwork Graph Improves AI Responses
Let’s look at what this means in practice.
Without the Teamwork Graph, a simple query to an LLM tool that is connected to some of your systems may be able to give you some valuable output. Input goes in, generic answer comes out.
But in this Rovo example powered by the Teamwork Graph, you can start to see a lot more information being shared, painting a much fuller picture.
You get to leverage various tools to address distinct use cases effectively, and that same query, prompt, or question will give you a richer response that provides context that other tools and systems may not have.
In this case, when you ask for a summary of what happened this week, you’re getting the Jira ticket, the two GitHub deployments, and statistics about an improved cycle time.
So, again, it’s a lot more robust and richer. The best part is that the Teamwork Graph and the Rovo infrastructure are underlying all of that, collecting all that data before you even get there.
Rovo Use Cases
One other thing I want to do before we get into building an agent is give you a little bit about Rovo use cases and show how you can utilize these solutions in the real world.
One of the biggest misconceptions about Rovo is that it’s only useful for software developers. In reality, it creates value anywhere employees spend time searching for information or switching between systems.
In one case, an on-call engineer can constantly find similar incidents and resolve outages faster.
Marketing teams can reuse successful campaigns instead of starting from scratch every time.
Quality engineers can see requirements, test cases, and defects together without jumping across multiple systems.
These may seem like different use cases, but they’re all solving the same problem: helping people find trusted information quickly so that they can make better decisions.
That’s the real power of Rovo when it comes to building on a connected digital thread.
Six Steps to Build a Rovo Agent
Now we’re going to talk through what it takes to actually build a Rovo agent.
We’ll walk through these steps. There are only six of them, and we’re going to do these hands-on as well.
Step 1: Open Rovo Studio
First, you’re going to go to studio.atlassian.com, where you can also click into it from any of your Atlassian tools.
This is the central workspace for building AI agents, automations, and custom applications within your Atlassian Cloud environment.
Think of Rovo Studio as your AI development hub. It provides everything you need to create intelligent assistants that can interact with Jira, Confluence, and other connected systems.
If you plan to build multiple agents or automations, I’d recommend bookmarking that page so that you know where it is and can easily access it. Again, it is also listed on the side menu of your Jira tools.
Step 2: Log In With Your Atlassian ID
You’ll log in with your Atlassian ID.
Keep in mind that you could be part of multiple organizations, so this becomes a kind of single sign-on question.
Are you using multiple accounts with one single sign-on account, or should you log in separately for each?
Just note that there. You’ll log in with whatever credentials you have.
Step 3: Create a New Agent
The next step is to create a new agent. There are some options on the left and at the top.
You can create an agent using Rovo to create a Rovo agent, which is pretty interesting. It’ll interact with you to help build that.
But it’ll also allow you to build it manually, and we’re going to show you how to build it manually today so that you can cover all the areas that you want.
You can also browse existing agents. You can duplicate those agents, or you can simply use those agents if you want to deploy what’s already there.
That’s another added benefit when you go to create an agent: you can use what’s already available.
Step 4: Define the Agent’s Purpose and Instructions
From here, you’re going to define your agent’s purpose and instructions.  This is really important.  You have to start by giving it clear, descriptive information, such as a name. That name is important so users immediately understand what the purpose is.  Then you have to give it a concise description of what it does.  The real intelligence comes from the instructions you give it, where you use natural language to tell the agent what to do, how it should behave with users, what tasks it should perform, and what’s most important when people are using that agent.
Finally, create a few conversation starters.  These are bits of information that people can simply click on to start a conversation with that agent.
Step 5: Configure Knowledge Sources and Skills
The fifth step is to configure the knowledge sources and skills.  Knowledge sources are really important because that’s where other information lives.
Obviously, Jira and Confluence are inherently built in, but you could be using Google Drive with Google Docs, SharePoint sites, Microsoft OneDrive, or any number of other solutions.  Knowledge sources are really important so that Rovo can go and find information in other places.  Then there are skills.  Skills are also really important. If you want the agent to do work, you have to give it the skills that it’s allowed to use. We’ll talk about that when we do the demo.
Step 6: Test, Iterate, and Publish
The last step is to play in the playground and test your Rovo agent.  Iterate on it, because there will always be things that you want to add back to it.
“Don’t do this.”  “I don’t like when you do that.”  You’re going to have to iterate on it, and then you can publish it to a team in a pretty straightforward manner.
Creating a Rovo Agent: Demo
Let’s dive into the demo of actually creating a Rovo agent.  When you go into Studio, it’s going to provide you with this interface.
You’ll see your agents here on the left, and you can create new agents directly from the top.  I always tell people, don’t just hit the Create button.  The Create button will prompt you with another Rovo box where you have to give it information to start building.  I like to build the first couple, at least, from scratch.
You’ll see this option when you hit the drop-down. It says Manually Create Agent, so I’d recommend doing that.
Using the Agent Template
What that’s going to do is give you a basic input instruction box.  It’s going to give you the skills area, knowledge area, conversation starters, and so on.  I’m going to throw a template in, and we’re going to briefly talk through that.  I’ll make sure this template is in the YouTube description underneath so that you can utilize it really quickly.
Agent Name
Give it a name that makes sense.  You can actually do that up here as well, but you’ll also do that in the body.
Purpose
There should be a one- to two-sentence summary of why this thing exists and who it helps.
Description
Then you have a longer description, maybe three to five sentences.  This should be used for the directory listing when people are trying to figure out what these agents do.  The description will be very helpful.
Instructions
For instructions, I always love to use this format:
“You are a…” and then the role.
“You speak in…” whatever tone.
“Your expertise is in…” whatever area applies.
That format will give the AI, this Rovo model, a better understanding of how it should interact with people.
Audience
Who is going to be using this agent?  Is it going to be very technical people?  Is it going to be non-technical people?  Maybe that dictates how information is shared.
Workflow
What’s the first thing that the agent should do when it starts a conversation?  Then what’s next? And next? And next?
Output
If you need a certain format for the output, define that.  Sometimes it’s, “I’m going to give you information, and I need you to provide it in a certain format.”  This is very helpful for that.
Constraints
Constraints are always great to use.  What you’ll find, the more that you use Rovo, is that it may start to do things that you don’t want it to do.
So you can actually say:Â
“If you don’t have this, don’t do that.”
“If you don’t have this information, don’t proceed.”
It sounds very rudimentary, but sometimes you want the AI to find a solution, so it’s going to keep looking.  But if it’s very binary, zero or one, you need to tell the agent, “Don’t do that. I don’t want you to go down a rabbit hole.”  Then define things it should always do as positive constraints.
Examples and Fallbacks
You can give it some examples.  What would somebody ask, and what should the agent answer?  Then define a fallback position. If you don’t know, say X, and then point them to a resource.  It could even be a person in the organization.
For example: “We don’t have a document for this. Let this person know so that they can create that document.”
Knowledge Sources and Conversation Starters
You can also list the knowledge sources and conversation starters.  Conversation starters are great because they’ll start people using it without having to know exactly what to type.  Sometimes they get this input box and don’t know what to do, so it’ll give them a couple of conversation starters to begin.
Example: Pipeline IQ Product Manager
I’m going to show you an existing agent rather than creating a new one from scratch here.  I have this one up already.
In our test Jira and Confluence, I’ve created a product called Pipeline IQ, and I’ve created an agent called Pipeline IQ Product Manager.
This has all of the same things that I just showed you in that template: the purpose, description, instructions, workflow, constraints, examples, conversation starters, acceptance criteria, and all the things that I just mentioned.
Configuring Agent Skills
It also has all of these skills.
If you’re creating a new agent, you’ll click Add Skill, and then you’ll have all these skills that you can add.
There are ones just for Jira and just for Confluence. You can add them both. There’s Bitbucket, Compass, and others.
This gives your agent permission to do certain things in Jira, Confluence, or any of those products.
So if I want it to be able to update a sprint or get editable fields, I have to give it that skill.
If I don’t give it that skill, it can’t help me in those areas.
This is a very common area where people miss something and then say, “Rovo doesn’t work.”
Well, Rovo doesn’t work because you need to give it that skill.
Think of skills like permissions.
Configuring Knowledge Sources
At the bottom are the knowledge areas.  You’ll see I have selected all organizational knowledge.  With custom settings, you could say, “Don’t look at these spaces or pages in Confluence,” or things like that.  I can also allow it to search the web or not, which is beneficial if you want it to go out and find information elsewhere. Â
Example Use Case: Associating a Jira Issue With an Epic
I’m going to show very briefly one use case of this particular Rovo agent that I built.  This ticket in Jira has no Epic.  I can come in here, click Agents, and select my Pipeline IQ agent.  It will create a new session here at the bottom.  I can click into the chat itself and say:
“Can you find what Epic this issue should be associated with?”
It will come back and look at all the Epics, analyze what’s in this ticket, and try to determine which Epic it belongs in.  It will give me an answer. Here’s the one that it’s recommending: it should be in Pipeline Orchestration and Standardization.  It explains why and actually gives you that information.
Then I can say:
“Let’s attach it. Can you associate this issue with that Epic?”
Because I gave it the skill to edit parent issues, I’ll get a button here to click Accept.  It’s going to add that parent work item, and you’ll see that it updated the issue to the Pipeline Orchestration and Standardization Epic.  And that’s all we have for this demo.
Best Practices for Building Rovo Agents
That was creating a Rovo agent.  We’re going to summarize here with some best practices.  As you begin to build more and more agents, there are a few best practices that I think will dramatically improve the results that you get.
Be Specific With Your Instructions
First, be very specific about your instructions.  Giving clear guidance about what the agent should and should not do consistently will produce better results.
Start With One Focused Use Case
Start with a single focused use case.  Validate it.  Validate that it works well.  Then expand it from there over time.
Be Intentional About Knowledge Sources
Be intentional about the knowledge sources you connect.  More data isn’t always better, especially if it’s data that’s not really helpful or maybe provides bad information.
Use Conversation Starters
Use conversation starters to guide users toward the agent’s strongest capabilities, making it easier for them to realize the value immediately.
Test Before Publishing
Before publishing, test that agent.  Make sure it’s able to answer real questions. Make sure that it’s actually giving good information, not just performing well in ideal scenarios.
Continuously Review and Update the Agent
Finally, remember that the AI agent isn’t a one-time project.  As your business processes, documentation, and applications evolve, regularly review and update that agent so it continues to provide accurate, relevant, and trustworthy assistance.
Final Thoughts on Building a Rovo Agent
That’s all I have for you today, and hopefully you’ll see how easy it is to build a Rovo agent in just about 15 minutes.  Hopefully, this gives you a good starting point.  But also remember, the real value comes from connecting your agent to the right knowledge sources, giving it clear instructions, and then continuously refining it as your business evolves.
Rovo Manual Template
NAME: [Agent Name]
PURPOSE: [1–2 sentences: why it exists, who it helps]
DESCRIPTION: [3–5 sentences for the directory listing]
INSTRUCTIONS:
Role: You are [role].
You speak in [tone].
Your expertise is [domain].
Audience: [Who uses this agent and their context]
Workflow:
1. [First thing the agent should do when a user starts a conversation]
2. [Next step]
3. [Continue…]
Output Format: [How to structure responses] Constraints:
- Do NOT [guardrail 1]
- Do NOT [guardrail 2]
- Always [positive constraint]
Examples:
User: “[sample question]”
Agent: “[ideal response]”
Fallback: If you don’t know, say [X] and point to [resource].
KNOWLEDGE SOURCES:
[Source 1 — what it contains]
[Source 2 — what it contains]
CONVERSATION STARTERS:
“[Starter 1]”
“[Starter 2]”
“[Starter 3]”
How SPK Can Help
If you’re looking to build more advanced agents, connect Rovo to other engineering tools, or explore how AI can actually accelerate your product development cycle, SPK would be happy to help.
Feel free to reach out to us using the contact information in the description, and we’ll be happy to have a conversation.
If you found this video helpful, be sure to like, subscribe, and follow the SPK & Associates YouTube channel for more practical tips on AI, Atlassian, and engineering transformation.
Thanks for watching, and we’ll see you in the next video.
