Atlassian Teamwork Lab Experiment Demonstrates How AI-Focused Retrospectives Drive Team Adoption and Workflow Alignment

The prevailing narrative surrounding the integration of artificial intelligence into the modern workplace is frequently dominated by large-scale, high-level concepts, such as massive enterprise rollouts, platform-wide transformations, and top-down corporate initiatives. While these structural shifts are undeniably significant, researchers at Atlassian suggest that the actual day-to-day behavior of employees is shaped more profoundly by the consistent rituals and regular conversations that occur within individual teams. Rather than launching a separate, siloed AI program, Atlassian’s Teamwork Lab recently conducted an experiment to determine if upgrading an existing team ritual—the retrospective—could more effectively operationalize AI adoption and behavior change.

The experiment involved 33 diverse teams across Atlassian, serving as a focused study to capture and measure real-world results of AI integration. The retrospective, a standard practice in agile methodologies where teams reflect on their recent work to improve future performance, was identified as the ideal vehicle for this intervention. By focusing on how teams talk about their work, the Teamwork Lab sought to move AI from an abstract concept to a practical tool embedded in the team’s standard operating procedures. The results of the experiment indicated significant shifts in how teams perceived and utilized AI, with participants reporting higher levels of alignment, increased confidence, and a more nuanced understanding of where AI adds the most value.

At the core of this experiment was the realization that AI adoption often stalls because it is treated as a solitary endeavor. When employees explore AI tools independently, they are more likely to abandon them if they encounter initial friction or fail to achieve immediate, high-quality results. To counter this, the AI-focused retrospectives were designed to make AI learning a social, collective experience. The data gathered from the 33 teams showed that after just one AI-focused retrospective, 82% of participants reported learning a new AI use case that they could immediately apply to their work. Furthermore, teams reported feeling more aligned on how to use the technology and more comfortable discussing both the successes and the limitations of the tools at their disposal.

The methodology developed by the Teamwork Lab is framed as a repeatable "play" that any team can implement using their existing retrospective schedule. This approach eliminates the need for additional meetings or the creation of new, burdensome "AI initiatives." The process is divided into four distinct steps designed to ground AI in practical application, foster social learning, document institutional knowledge, and drive continuous improvement through small, actionable commitments.

The first step in this process is to anchor the conversation in real work rather than abstract AI capabilities. Because retrospectives are already grounded in the events of the previous sprint or work cycle, facilitators are encouraged to surface specific moments where AI was utilized, where it was intentionally skipped, or where it could have potentially provided assistance. This involves asking targeted questions such as where AI helped save time, where it failed to meet expectations, or what tasks felt tedious enough that an AI solution should be explored in the future.

Ben Ostrowski, a researcher at Atlassian’s Teamwork Lab, noted that this line of conversation is essential for surfacing practical and reusable patterns that are directly relevant to a team’s specific workflows. By encouraging team members to reference specific tasks—such as using AI to draft or proofread an email, or summarizing a 20-page document into key talking points—the retrospective reveals the practical reality of AI usage. Ostrowski highlighted that this transparency also helps to identify the quiet fears or gaps in knowledge that may be pervasively slowing down adoption and eroding team confidence. The experiment found that the most impactful "aha" moments occurred during this phase, often involving simple but highly effective use cases like transforming messy meeting notes into customer-ready summaries.

The second step emphasizes making AI learning a social activity. The Teamwork Lab’s research indicates that when teammates, and particularly team leaders, openly share their misfires and failed attempts alongside their wins, it creates a psychologically safe environment. This openness invites less-confident colleagues to experiment and grow without the fear of failure. During the AI retrospectives, facilitators are encouraged to frame stories of AI frustration as learning opportunities for the entire group. This shift in perspective helps teams problem-solve together, swapping tactics to overcome common hurdles. One participant in the experiment noted that the ability to "vent" about AI’s shortcomings was just as valuable as learning new tips, as it helped the team calibrate their expectations and feel more inspired to continue integrating agents and tools like Atlassian’s Rovo into their daily flows.

The third step involves the creation of a living AI use-case document. To ensure that the insights surfaced during the retrospective are not lost, teams are encouraged to co-edit a shared document that serves as a lightweight AI playbook. This document is typically divided into two sections: "AI helps us with…" and "We avoid AI for…" This practice turns ephemeral conversation into a permanent organizational asset.

During the experiment, teams utilized these documents in various innovative ways. One team integrated their AI use-case list into their onboarding process for new hires, clearly communicating where AI usage is expected rather than merely allowed. Another team used the "avoid" list to resolve long-standing debates regarding data privacy and security, identifying specific gray areas that required escalation to leadership for clearer guidance. The researchers emphasize that this document does not need to be complex; it simply needs to be accessible, easy to edit, and treated as an evolving artifact that is updated as the team’s relationship with AI matures. When paired with formal AI Working Agreements, these documents provide a clear roadmap for how a specific team operates in an AI-augmented environment.

The final step in the repeatable play is to end the session with one small, specific AI commitment. The Teamwork Lab posits that while reflection is valuable, actual behavior change only occurs when a team commits to a clear next step. Before the retrospective concludes, every team member or the team as a collective is asked to make a micro-commitment for the upcoming sprint. These commitments are most effective when they are small, specific, and measurable—for example, a team member might commit to using AI to draft one project update or to experiment with a specific prompt template for a recurring task. The goal is not immediate perfection but rather the accumulation of small, forward-moving actions that eventually stack into significant behavior change.

The overarching takeaway from the Teamwork Lab’s experiment is that AI enablement does not require a reinvention of team management. Sara Gottlieb-Cohen, another researcher at the Teamwork Lab, likened the approach to adding "AI-powered LED lights" to the spokes of an existing wheel. The strength of the AI-focused retrospective lies in its ability to be integrated into everyday rituals without adding a single new recurring meeting to an already crowded calendar.

For teams looking to begin this process immediately, the researchers suggest a "quick and easy" starting point: adding just two questions to the next scheduled retrospective. These questions—"Where did AI help you this sprint?" and "Where did you try to use AI but it failed?"—are enough to spark the necessary dialogue. Over time, the goal is for AI reflection to stop being viewed as a special "initiative" and instead become a natural part of how a team regularly inspects and adapts its way of working. By operationalizing AI through existing team rituals, organizations can move beyond the hype of enterprise-wide transformations and focus on the practical, incremental changes that define how work actually gets done.