Unlocking the Power of Rovo Agents – Atlassian Team ’25 NA

This is Techron tv. We are back at Atlassian, team 25. I'm here with Shihab Hamid, who's the head of product for Rvo agents and Shahi.

I, I understand Rvo agents were a, for a very big part of the topic or the keynote presentation on, on Wednesday. Maybe you can tell me a little bit what's new about them and, um, you know, what are other things that your customers should need to know? Yeah, One of the biggest announcements that we're really excited about is we've made Robo available to everyone.

And so it, there's no excuses to getting started. And really with robo agents, it's really giving our customers the ability to have virtual teammates on deck that they can hand over some of their, uh, more mundane tasks, uh, too. And we've got a bunch of out of the box agents that help you in a variety of, uh, different contexts throughout, uh, the teamwork journey.

So one of the things that we're really excited about is, uh, you know, having Rover agents that help you with brainstorming and ideation. We've got agents that help you whiteboard content all the way through to taking those ideas into user stories and those user stories through to code, through to implementation. And then also on the other side, when you're having to support products, we've got agents that help you provide dynamic support for both self-service and also for your teams, uh, that are providing that help.

I actually, you talked about the brainstorming. Can you maybe give me kind of an example of how that might be used? Because I, it's very interesting to me because we have, where I work, we have a lot of weekly meetings where we just kind of throw around ideas, stories, some story ideas, some just ideas about, uh, promoting things on social media.

What, how would say typically like a brainstorming session work and how would the agent work within that session? Yeah, so you can, uh, bring along agents as you're creating content together. So we've seen with the rise of remote work, the ability for people to just collaborate on a page dynamically throughout all areas of the planet.

And so what you can do is actually invite an agent onto a page. You can get them to generate content with you in a collaborative manner. And so whether that's brainstorming about, you know, uh, marketing content, now, Are the notes taken by an agent at the same time when you're brainstorming?

Yeah. Yeah. So, oh, cool.

The agent can, uh, insert the notes. So before you even publish the page, you can, uh, correct what the agent is, uh, producing. You can change it, you can tweak it, you can also have a conversation with the agent.

And so you can really, uh, customize the content before, um, Uh, Kim, like I'm visualizing this and how are people reacting to interacting with the agent? Is it something that they're, they're finding useful or is it something that takes a little bit of getting used to? Or is it, yeah.

What is that experience like? Yeah, Honestly, we've seen a lot of people play around and dabble with, uh, a lot of the LLM technologies that are available out there. And having, uh, the power of that in context as you're creating content, as you're working through, you know, stories in Jira, um, issues on your backlog all the way through to production, we're seeing that, bringing that power in context, but also tailored to your organization with rules that you can define for your agents, give that power up to organizations.

That's, That is, that is very interesting. How, how can users customize robo agents to meet their specific needs? So we've got a bunch of agents that are available out of the box, and so that helps a lot of the teams get started.

And what we've seen with almost every single customer is once they get started editing and tuning agents is just addictive, right? So you start with an agent that can do one or two things. You start small, you give it access to, you know, natural language instruction.

Like what's one of the things out of the box you can do, for instance? Yeah. So one of the most common ones we've seen is for, uh, a service use case.

We've got, uh, a triage agent. So imagine a new ticket comes in, it can, based on your past ticket handling history, so you don't have to encode any of those rules. It can look at how you've behaved in the past and just apply that to new tickets coming in.

And so that's a very simple example of an agent that can start out small, but think about like on-call rosters, people that are available, workload schedules. You can encode all of that in into your agent description and into the instructions and apply that to every ticket that comes in. Is it pretty easy to, to create these agents?

Yeah, absolutely. You can get started. So the natural language process for building an agent, you can just describe what you want your agent to do.

Okay. And we've seen like teams get started, like non-technical teams get started off the bat. And one of our most popular agents internally in Atlassian is our onboarding agent.

And that was entirely created by our HR team. So, you know, not super technical teams getting started to say, Hey, when a new starter comes on board, here's some policy documents, here's some employee onboarding guides, and you could just hit go. How many, uh, agents are used within atla, Atlassian?

I've heard it was like 2000, is that sound Right? Yeah, there's an incredible number because once we get started with, uh, these agents, like I mentioned earlier, it's super addictive and anyone can clone someone else's agent. Got it.

So if you like something you can, that that's available within your organization, you can duplicate it and make it your own. And so we've got a bunch that help you get started, but if you see something, uh, from someone else, like a, a peer in your organization, you're like, Hey, I really like that agent, but I wanted to do two or three more things. Um, it's very easy to riff off other people's ING agents as Well.

Alright. How, so how does Atlassian plan to expand the role of Roag agents in the broader ecosystem? Yeah, so we've got agents available on our marketplace as well.

So it's possible for our ecosystem vendors, third parties to build custom agents. So for example, we've got agents from Canva that can help you, uh, you know, create beautiful designs in context. We've got other agents from, uh, other ecosystem vendors that help you integrate with HRIS system.

So you can, you know, request leave, you can log actions. And so we are making that all available as part of our Forge developer platform. So if you wanna really customize the, uh, the agents and bring in new actions, you can get into code and make that happen.

Oh, cool. Um, so what I'm gonna ask you about something in particular is VO Agents Studio. And how does that help customers implement agentic ai?

Yeah, so what we've seen with, um, uh, the adoption of agents is that it's very easy for people to get started and create their own conversational agents, but the true power of agents is connecting them up to a process. And so with Studio, we're, uh, enhancing the interoperability between automation and agents and other capabilities that we've got on our platform. So it's really aimed at the low code, no code developer to get started, and you can customize all the way through to more sophisticated solutions in code.

Um, speaking of agentic ai, but there's, you know, Google did a, a number of series of, of announcements, even like agent to agent type of capabilities. I'm wondering just kind of in general, what you think of this kind of explosion or tidal wave of, of AI agents and how they go seep into enterprises. Do you th I I'm just kind of wondering, do they, are we gonna see an era where people are using different agents from different vendors, or is there gonna be an all in one solution?

I was just wondering, or is it gonna be a hybrid of some sort? Yeah, we, we definitely think it's gonna be a hybrid approach because everyone's got the concept of agents. The idea for us to interact with a lot of these other platforms, you can see a lot of protocols getting developed, like the MCP protocol to share actions, uh, agent protocols to, uh, enhance agent to agent communication.

So I think, think of one way to think about this and simplify it is just, it's another way of interacting between systems and it's happening at a higher level. So some of these agents can plan, they can communicate in plain English, um, it's just a more natural way of, uh, interacting. Do you think, Uh, in, in terms, and I think Atlassian is on the right, is like understands this, but do you see the adoption or even the acceptance and use of AI agents starting kind of as like a bottom up type of movement where people who feel comfortable using them in certain tasks then expand and then kind of move onto more complicated processes?

'cause it seems as if what you, what you've been offering as a company this week is something, something along the lines of, here's how to collaborate in a way to, to, to reduce, uh, wasted time. Here's, here's things that you're comfortable doing, and then you build upon that level of trust to do more and more complicated things. Yeah, absolutely.

Uh, the message out there is just to get started, every customer I've sp spoken to has said, once we get that aha moment of just building my first agent, the iteration is so addictive and it's just natural because you just want it to do more. And honestly, our customers are always pleasantly surprised by what's possible. And so I think that incremental approach is something that lends itself to experimentation, uh, to, to really trialing what's possible and then pushing the envelope there.

Was there before the aha moment, was there some trepidation or what was kind of their general mindset? For the most part? Were they, they kind of, people are scared of new technology.

Yeah. And it always seems as if they're scared or resistant to it until they use it, then they realize, wow, like, look what I missed out on. Yeah.

And then they become true evangelists or true believers. Yeah. I think honestly what we've seen is a lot of customers get started with search, so just being able to find content.

And so one of the experiences that we've got baked into search is if you ask it something that looks like a question, it'll actually generate an answer for you. It'll show you the knowledge sources that it's producing that answer from. And then after the answer that we generate, we also generate subsequent questions that you might be interested in asking.

And so once you hit that, it actually takes you to a conversational experience with robo, and then you have the back and forth conversation, and that really, uh, gets customers thinking about the power that you can have just with that simple interaction. And so once people have, uh, a couple of those interactions, they start asking, how can I make that process or that question answering more standard? How can I really encode that into, um, a process that I can repeat?

And that's where agents really come to life. Yes. It's, it's interesting, like there's this whole, this whole process, um, and there's not a lot of training involved with the VO agents, right?

It's just pretty standard nonthreatening, non intimidating. And I think the reason I'm bringing this up is because I keep coming across these research studies where I see two things. I see a lot of push by the CEOs and the C-Suite to adopt and use AI as quickly as possible.

They're not very specific about how they want it to be used. They just want it used. Yeah.

And then at the same on the flip side, there's a level of, of uncertainty or fear, uh, among the rank and file and how that's going to impact them and how it's gonna work within their, their workflow. Yeah. So, um, it seems as if this is kind of somewhere where you're looking from at, at the, at the employee's point of view, more so, but it's also practical for the CEOs.

Yeah, absolutely. So even from the, uh, the knowledge worker that's working day in and day out with a lot of different knowledge sources to get their day job done, they may start with something as simple as chat to have those conversations. Like it's almost bringing their documents to life.

So, you know, you have reams and res of content, you don't know where the contents are stored. That's, that would keep coming across that robo search is so good at that, right? Yeah.

Looking across the enterprise and finding things. Exactly. And, and you're not looking just for a search result.

You're often looking for an answer to a question or you're looking for some analysis. And one of the, uh, you know, headline features that we're really excited to announce is deep research. So the ability to really ask a general broad question and have the system generate subsequent questions and actually answer those questions for you to compile a beautiful report on a particular area.

So that could be, you know, previous, uh, you know, iterations of development that you've done, previous roadmap ideas, uh, insights and analysis on, uh, exploration, uh, that has happened in the company. So really bringing that, you know, stale knowledge back to life. Cool.

Hey, uh, Shia, very nice to meet you. This is very interesting. Um, I think this is, uh, fast, quite fascinating.

It's gonna work really well within what everyone else is doing, I think, um, is, and I, I think that really maybe agentic AI will be 2025, the big year for it. It'll, it'll gain some sort of traction. Yeah, we're absolutely excited about this.

This is one of the biggest moments that we've had. And, um, just to see the power of, uh, the collaboration between, uh, agents and, uh, team members coming together to really make work happen, uh, and teamwork happen at an accelerated rate. Alright.

Thanks. Thanks again. Thanks again.

So that's, um, that's yet another interview for us at, uh, Atlassian, team 25. Uh, we're right across, again, we're right across the street from Disneyland. We're near where the Mighty Ducks play, the Anaheim Angels, they're all here and, uh, so is Atlassian.

So stay tuned for more content on Techstrong tv.