AI helps employees get faster—why haven’t companies become stronger? Atlassian reveals three blind spots behind organizations that can’t turn a profit

Artificial intelligence (AI) is sweeping across the global workplace, but many companies have found that overall efficiency has not improved as expected. According to a report by VentureBeat, Dr. Molly Sands, head of the Atlassian Teamwork Lab, recently said bluntly at a public event that most companies’ approach to adopting AI has “gone backward”—they only focus on optimizing individual productivity while ignoring teamwork collaboration, making it hard for organizations to see tangible return on investment (ROI). She emphasized that successful teams must build shared context, redesign work processes, and cultivate a corporate culture that dares to experiment.
(Background: The Jacobian conjecture in mathematics was refuted, OpenAI researcher: Codex “with no internet assistance” also worked it out)
(Additional background: Is running AI models self-hosted cheaper? Cline tests Kimi K3’s cost advantage; netizens rebut: it’s more expensive than Fable 5 when priced by task)

Table of contents

Toggle

• Individual productivity improves, but the organization doesn’t see ROI
• Only 6% of managers see real value; successful teams have three key traits
• Create “AI working agreements,” exposing hidden organizational fault lines

With the widespread adoption of generative AI tools, many employees’ day-to-day task speeds have indeed improved significantly, but this efficiency boost does not seem to translate smoothly into companies’ overall competitiveness. On July 21, 2026, Taipei time, according to a VentureBeat report sponsored by software giant Atlassian, Dr. Molly Sands, head of the Atlassian Teamwork Lab, delivered a deep analysis of this seemingly contradictory workplace phenomenon at the VB Transform 2026 event.

Individual productivity improves, but the organization doesn’t see ROI

During a fireside chat hosted by VentureBeat senior technology contributor Sam Witteveen, Sands pinpointed the core issue in one sentence: most companies adopt AI “in reverse.” Companies often only focus on optimizing how individual employees use AI, without looking to optimize the way teams collaborate at the teamwork level.

This imbalance leads to an awkward outcome: even though individuals are moving faster, because team workflows are not adjusted, these different approaches end up “colliding quickly,” preventing the organization from clearly seeing investment returns (ROI). Sands’ team of behavioral scientists and psychologists studies not only the impact of AI on collaboration, but also works with companies to practically redesign work processes.

Only 6% of managers see real value; successful teams have three key traits

According to the Atlassian 2026 State of Teams Report survey of 12,000 knowledge workers worldwide and about 200 senior executives at Fortune 1000 companies, as many as 89% of managers admit that after employees use AI, their speed does indeed increase. However, surprisingly, only 6% of managers can point to clear ROI examples; and only about 14% of teams successfully turn AI usage into real value, showing that even within the same company, performance gaps between teams can be extremely wide.

Sands noted that organizations that successfully turn AI into team advantages typically share three key traits:

• Context: Successful teams build a “context graph,” systematically recording goals, decisions, and organizational knowledge in shared digital platforms such as Jira or Confluence, so AI can access organization-level context rather than relying solely on isolated individual prompts.
• Workflows: They don’t just speed up single tasks—they redesign the entire end-to-end workflow.
• Culture: Leaders clearly encourage learning and experimentation, and are able to tolerate failures during the experimental process.

Create “AI working agreements,” exposing hidden organizational fault lines

To help enterprises cross this gap, Sands suggests that teams can learn quickly through setting deliberate constraints—such as breaking tasks down into the smallest units, or challenging “no handwritten code at all within a week.” In addition, it’s also crucial to establish clear “AI Working Agreements” when starting a project, including deciding what kinds of matters should be handled by AI, what should be avoided, and which agents and skills the team should share.

The article ends with a deeper revelation: AI hasn’t actually created entirely new management problems—it has simply ruthlessly exposed fault lines that already existed within organizations, such as hidden assumptions and different mental models. The intervention of AI amplifies the consequences of these gaps, making “shared context” and “explicit ways of working” more urgent and more important than ever for future enterprise operations.