Scaling agentic AI isn’t nearly having the most recent instruments — it requires clear steerage, the suitable context, and a tradition that champions experimentation to unlock actual worth. At VentureBeat’s Remodel 2025, Anu Bharadwaj, president of Atlassian, shared actionable insights into how the corporate has empowered its staff to construct 1000’s of customized brokers that clear up actual, on a regular basis challenges. To construct these brokers, Atlassian has fostered a tradition rooted in curiosity, enthusiasm and steady experimentation.
“You hear rather a lot about AI top-down mandates,” Bharadwaj mentioned. “Prime-down mandates are nice for making a giant splash, however actually, what occurs subsequent, and to who? Brokers require fixed iteration and adaptation. Prime-down mandates can encourage folks to start out utilizing it of their each day work, however folks have to make use of it of their context and iterate over time to comprehend most worth.”
That requires a tradition of experimentation — one the place short- to medium-term setbacks aren’t penalized however embraced as stepping stones to future progress and high-impact use circumstances.
Making a protected atmosphere
Atlassian’s agent-building platform, Rovo Studio, serves as a playground atmosphere for groups throughout the enterprise to construct brokers.
“As leaders, it’s necessary for us to create a psychologically protected atmosphere,” Bharadwaj mentioned. “At Atlassian, we’ve all the time been very open. Open firm, no bullshit is considered one of our values. So we concentrate on creating that openness, and creating an atmosphere the place staff can check out various things, and if it fails, it’s okay. It’s superb since you realized one thing about how you can use AI in your context. It’s useful to be very express and open about it.”
Past that, you must create a steadiness between experimentation with guardrails of security and auditability. This contains security measures like ensuring staff are logged in once they’re making an attempt instruments, to creating certain brokers respect permissions, perceive role-based entry, and supply solutions and actions based mostly on what a selected person has entry to.
Supporting team-agent collaboration
“Once we take into consideration brokers, we take into consideration how people and brokers work collectively,” Bharadwaj mentioned. “What does teamwork seem like throughout a group composed of a bunch of individuals and a bunch of brokers — and the way does that evolve over time? What can we do to assist that? Because of this, all of our groups use Rovo brokers and construct their very own Rovo brokers. Our principle is that when that sort of teamwork turns into extra commonplace, your complete working system of the corporate adjustments.”
The magic actually occurs when a number of folks work along with a number of brokers, she added. In the present day quite a lot of brokers are single-player, however interplay patterns are evolving. Chat won’t be the default interplay sample, Bharadwaj says. As a substitute, there can be a number of interplay patterns that drive multiplayer collaboration.
“Basically, what’s teamwork all about?” she posed to the viewers. “It’s multiplayer collaboration — a number of brokers and a number of people working collectively.”
Making agent experimentation accessible
Atlassian’s Rovo Studio makes agent constructing obtainable and accessible to folks of all ability units, together with no-code choices. One building trade buyer constructed a set of brokers to scale back their roadmap creation time by 75%, whereas publishing large HarperCollins constructed brokers that decreased guide work by 4X throughout their departments.
By combining Rovo Studio with their developer platform, Forge, technical groups acquire highly effective management to deeply customise their AI workflows — defining context, specifying accessible information sources, shaping interplay patterns and extra — and create extremely specialised brokers. On the similar time, non-technical groups additionally must customise and iterate, so that they’ve constructed experiences in Rovo Studio to permit customers to leverage pure language to make their customizations.
“That’s going to be the large unlock, as a result of basically, after we discuss agentic transformation, it can’t be restricted to the code gen situations we see right now. It has to permeate your complete group,” Bharadwaj mentioned. “Builders spend 10% of their time coding. The remaining 90% is working with the remainder of the group, determining buyer points and fixing points in manufacturing. We’re making a platform by which you’ll construct brokers for each single a kind of features, so your complete loop will get sooner.”
Making a bridge from right here to the long run
Not like the earlier shifts to cell or cloud, the place a set of technological or go-to-market adjustments occurred, AI transformation is basically a change in the way in which we work. Bharadwaj believes an important factor to do is to be open and to share how you might be utilizing AI to vary your each day work. “For instance, I share Loom movies of latest instruments that I’ve tried out, issues that I like, issues that I didn’t like, issues the place I believed, oh, this may very well be helpful if solely it had the suitable context,” she added. “That fixed psychological iteration, for workers to see and check out each single day, is extremely necessary as we shift the way in which we work.”
