As a software company that specializes in collaboration tools for software development and project management, it makes sense that Atlassian cites ‘making sure every voice is heard’ as a core tenet of their team culture. Their team of data scientists, however, take this principle one step further by always making sure the user’s voice is heard — through data.
This is where Tom, a senior principal data scientist with customer support services, started when his team began to build a generative AI tool for their support staff.
The goal was to proactively provide a suggested resolution for customer questions using knowledge-base articles to reduce the time required to respond to customers and, more importantly, to get a solution into customers’ hands much more quickly, Tom said, and it required pulling information from Atlassian’s 100,000+ knowledge-base articles — a Herculean task that required hard work, collaboration, and a whole lot of trust.
Luckily, Tom’s team was more than up to the task.
“Building user trust and confidence in working with generative text AI was a key challenge when designing this system,” Tom said. “Which is why we try to involve end users early and often in our development process. This allows us to iterate much more quickly on concepts and ensures that the models we develop stay aligned with user needs.”
This collaborative relationship not only helps Atlassian build trust with its users, but it also ultimately drives improved adoption and the overall success of the company because users became advocates for the system that was built, Tom said.
For Tom, the knowledge-base generative AI project established the data science team as a trusted source of truth — not to mention that it showed the team was capable of developing and scaling cutting-edge AI.
“It also helped position our team as subject-matter experts around the company and led to several of our AI solutions and best practices being adopted by other teams,” he said. None of it would have been possible, however, without a team culture rooted in collaboration, teamwork and psychological safety.
Team Culture on the Data Engineering Team
Being part of the data engineering team at Atlassian means a few things: centering remote-first communication, letting user feedback guide the project, not being afraid to try new things, and knowing that teammates always have each other’s backs.
According to Gianna Ciaccio, senior people data scientist on the People Insights team, the data team’s culture is reflected in Atlassian’s overarching focus on open communication and connection among teammates.
“Team culture is not virtual happy hours or pizza days — it’s comfort and safety in addressing difficult conversations,” Ciaccio said. This means leading with ‘candor and vulnerability,’ she added.
“Team culture is not virtual happy hours or pizza days — it’s comfort and safety in addressing difficult conversations.”
“Authenticity motivated by a common goal goes a long way toward building relationships and getting others invested in collective success,” she said. “There’s no need to suffer in silence or dance around a challenge — even acknowledging when a solution can’t be implemented immediately creates shared understanding.”
Nikita Nathan, a senior data scientist, understands the need for transparency firsthand. And something else, too: psychological safety.
She recently worked on a “fit score” project that was focused on helping the marketing team understand customer demographics in a more meaningful way.
“Think seniority, sector, and the job function that our customers have,” Nathan said. “These demographics help the sales team prioritize where they should be investing their time and efforts.”
Nathan’s team played a key role in defining the way the fit score was identified, Nathan said, but during the exploratory phase of building the model, Nathan approached the solution in a way that garnered some hesitation from partner teams.
“I had worked for weeks on this solution, so it was disappointing that it missed the mark,” she said. “I sparred with my manager and team on alternative solutions that would better solve the problem, and I was able to pivot quickly to a different algorithm that had a better ‘fit’ (no pun intended) for what my stakeholders were looking for.”
Even though she made a mistake, Nathan never doubted that the data team had her back.
“I used this as a learning moment to get better clarity on the problem from the beginning and utilize my team’s collective expertise in the brainstorming phase,” Nathan said.
This story, Nathan added, is a perfect snapshot of what it’s like to work on the data team at Atlassian — where trying new things and having the courage to spark change is all a part of making the products better.
“I know I can ask for help through any issue I have… Building those relationships is essential for our success,” Nathan said.
Solving for Distributed Work
Atlassian’s culture is built on trust — something that helps the data science team and other colleagues navigate challenges as a distributed workforce.
“I used to value in-person work earlier in my career,” Nathan said. “However, the flexibility of remote work is something that I value a lot today. I find that teams are equally as productive working remotely as in-person, and there is a level of trust that companies build with their employees when they fully embrace that.”
Atlassian employs roughly 11,000 people who are committed to building industry-leading collaboration and productivity software. The company’s popular tools like Jira, Trello, and Confluence are used by more than 300,000 businesses to streamline workflows and help teams stay on top of complex projects.
For the folks building those tools, the company is known for its “team anywhere” distributed-first approach, which allows teammates to choose where they work with the goal of fostering flexibility and inclusivity. Keeping a remote team in sync on a project while working across multiple time zones requires a team culture that is committed to both communication and continuous learning.
“You have to be willing to eliminate rituals that worked well in an in-person setting but that don’t have a good remote analog and create new remote-only rituals to fill those needs,” Tom said.
“You have to be willing to eliminate rituals that worked well in an in-person setting but that don’t have a good remote analog and create new remote-only rituals to fill those needs.”
Tom’s team, for example, has a “one remote, all remote” policy where if one person joins remotely, everyone joins via Zoom to ensure there are no uneven power dynamics.
“This allows for everyone to have a chance to participate and be heard,” he said.
According to Ciaccio, this support stands out from other teams she has worked on in the past.
“I’ve been lucky to work on some great data science teams in the past — but one factor that this team has seemed to ‘get right’ in our distributed working environment is building authentic camaraderie and friendship,” concluded Ciaccio. “No problem is your own to solve, and there are many who are eager to be with you in the thick of it.”
