AI makes it possible to produce work quickly, but it can also result in a lot of workslop—that is, work that looks polished, but does little to move the project forward. Sorting through workslop is time-consuming and undercuts the ROI on AI. It may look like productivity, but it just moves the actual work farther downstream.

The Teamwork Lab realized that workslop isn’t an AI problem, or even a people problem. We hypothesized that if teams could align on what good work looks like and how to use AI to get there, workslop would resolve itself.

Both careless and conscientious team members can produce workslop if they don’t have enough guidance. AI use isn’t the problem, either, especially when the AI did exactly as it was instructed to do.

AI polish makes problems harder to see. In an earlier Teamwork Lab study of 903 U.S. knowledge workers, reviewers were 22% less likely to catch a critical flaw in the polished version of a draft.

Workslop also costs trust. Receiving it can lead colleagues to see the sender as less reliable. According to a 2025 survey conducted by BetterUp and Stanford, 42% of people who received workslop viewed those colleagues as less trustworthy, and 37% saw them as less intelligent.

How the AI workslop workshop works

Thirty-one employees from the Workplace team—the group responsible for Atlassian’s office spaces, hospitality, and connection programs—agreed to attend small group sessions through June 2026. (We’d like to extend a big thank you to Workplace. Their candor, engagement, and participation made this pilot possible.) The workshop centered on the creation of AI-generated pages in Confluence, the internal Atlassian connected workspace.

Each of the groups completed a 60-minute workshop. The workshop itself had four phases that were designed to help the team define what a good page looked like, so that AI can help the team both save time and increase the quality of work.

The four phases were aimed at identifying:

• What does slop look like?
• What does a good page look like?
• What shared principles make a good page?
• What strategies can we use to keep quality high?

The team members also answered surveys pre, post, and two weeks after the workshop. All the statements used a five-point agreement scale to give feedback on whether they understood their team’s quality standards and whether they felt the work they produced met those standards.

Post-workshop, the teams had universal agreement on what a high-quality page looked like. 100% of people agreed that they knew what their team considers a high-quality page (up from 62%), and 97% agreed they knew how to use AI to create a quality page (up from 60%).

The teams also felt more accountable to their team members and more comfortable giving feedback. People named concrete tactics they planned to use, including recording a quick Loom to talk through their thinking before asking Rovo to draft the page.

94% said they felt more comfortable giving feedback on AI-generated pages and 97% reported feeling a stronger sense of responsibility to ensure their pages were up to standard.

Finally, there was 67% agreement that the team’s pages met a higher-quality standard post-workshop, as compared to a 40% baseline. These results are not statistically significant given that this was a small pilot study, but they are encouraging. 97% of participants would recommend the workshop to other teams.

“It’s incredible how much one 60-minute conversation can move the needle on a workslop problem that teams have lived with for months,” said Morgan Weaving, Senior Workplace Researcher on the Teamwork Lab. “The fix was surprisingly simple: say out loud what good looks like and agree on how you’ll use AI to get there, and you give the team a shared bar to work from.”