The AI ​​of Scaling Agentic shouldn’t be nearly having the most recent tools, but additionally a transparent guide, the best context and a culture that pursues experimentation to be able to open up the actual value. At Venturebeat Transformation 2025Anu Bharadwaj, President of Atlassian, shared implementable insights into how the corporate has enabled its employees to construct 1000’s of customs agents that solve real, on a regular basis challenges. In order to construct these agents, Atlassian has promoted a culture that’s rooted in curiosity, enthusiasm and continuous experimentation.
“You hear rather a lot about AI top-down mandate,” said Bharadwaj. “Top-down mandates are perfect for a sensation, but really what happens next and for whom? Agents require constant iteration and adaptation. Top-down mandates can encourage people to make use of them of their each day work, but people should use them of their context and to realize the utmost value over time.”
This requires an experiment culture wherein short to medium-term setbacks aren’t punished, but are accepted as a springboard for future growth and applications with a high impact.
Create a protected environment
The Agent Building platform from Atlassian, Rovo Studio, serves as a playground environment for teams throughout the corporate to accumulate agents.
“As managers, it is necessary for us to create a mentally protected environment,” said Bharadwaj. “We have all the time been very open to Atlassian. Open Company, not a bullshit is considered one of our values. So we think about creating this openness and creating an environment wherein employees can check out various things, and if it fails, it's nice. It is nice because you will have learned something about how you need to use AI in your context. It is useful, very explicit and open.”
In addition, you will have to create a balance between experimenting with guidelines of security and auditability. This concludes security measures comparable to ensuring that the staff are registered when attempting to try tools to be certain that agents respect the permissions, understand the role -based access and to know answers and actions based on what a certain user has accessed.
Support for the collaboration of the team agent
“When we take into consideration agents, we take into consideration how people and agents work together,” said Bharadwaj. “What does teamwork seem like in a team that consists of a lot of people and a lot of agents -and how does it develop over time? What can we do to support it? As a result, all of our teams Rovo agents and construct their very own Rovo agents. Our theory is that the sort of teamwork as soon as the sort of teamwork becomes commonplace, the whole operating system of the corporate becomes the whole operating system.”
The magic really happens when several people work with several agents, added. Today, many agents are a player, but interaction patterns develop. Chat is not going to be the usual interaction pattern, says Bharadwaj. Instead, there are several interaction patterns that drive multiplayer cooperation forward.
“Basically what’s teamwork about?” She posted the audience. “It is multiplayer cooperation – several agents and several other individuals who work together.”
Make experiments accessible to the agent
The Rovo Studio of Atlassian provides the event of agents for people of all skills, including no-code options. A customer of the development industry arrange a lot of representatives to shorten their roadmap creation time by 75%, while the large HarperCollins, which reduced manual work by 4 times of their departments, were built up.
Through the mixture of Rovo Studio together with your developer platform, Forge, technical teams are given powerful control to adapt your KI workflows deeply – to define the context, indicate accessible sources of data, design interaction patterns and quite create highly specialized agents. At the identical time, non-technical teams should adapt and iterate, in order that they’ve built experiences within the Rovo Studio in order that users can use natural language to make their adjustments.
“This shall be the good activation, because principally, after we talk concerning the operating transformation, it can’t be limited to the code scenarios that we see today. The entire team has to penetrate,” said Bharadwaj. “Developers spend 10% of their time coding. The remaining 90% work with the remainder of the team, record customer problems and fix problems in production. We create a platform that may create agents for every individual of those functions in order that the whole loop becomes faster.”
Create from here from a bridge to the long run
In contrast to the previous shifts to mobile devices or clouds, wherein a lot of technological or increasing changes occurred, AI transformation is essentially a change in the best way we work. Bharadwaj believes that a very powerful thing is to be open and share how you alter your each day work with AI. “As an example, I share waves with latest tools that I attempted, things that I like, things that I didn't like, things that I assumed, oh, that could possibly be useful if it only had the best context,” she added. “This constant mental iteration in order that the staff can see and take a look at each day may be very necessary if we alter the best way we work.”
