Salesforce survey finds AI productivity stuck in silos, with executives calling for “multiplayer” context

As artificial intelligence moves from experimentation into daily work, a new problem is emerging: teams can end up using AI as if each person were working in a separate universe. In a Salesforce survey of 250 U.S. executives, respondents described “interface isolation” as a bottleneck that keeps AI-generated insights from spreading through organizations and can stall ongoing projects when solutions stay confined to one person’s tool.

The survey found that 68% of executives believe AI insights go unshared across the workplace, while 66% say people unknowingly use AI to solve the same problem as colleagues. That duplication shows up directly in day-to-day performance, with 59% of respondents reporting work stalls when it is trapped in a single person’s AI tool. Only 15% of executives said their AI is embedded seamlessly across multiple systems and interfaces.

Those workflow gaps translate into measurable time loss, according to the survey. Executives estimated employees spend 22 business days per year toggling between AI and other windows. Salesforce also reported that the issue compounds at larger companies: 73% of enterprise executives said they see weekly stalls, compared with 37% at small and midsize businesses.

The survey’s findings point to a broader aspiration inside organizations: AI that supports collaboration rather than parallel isolation. Salesforce said 72% of respondents described an “ideal” where AI works across a team with shared context. For many leaders, that means the value of AI is not just better individual output, but the ability for multiple people to build on the same knowledge without restarting from scratch.

Slack EVP and GM Rob Seaman framed the shift in terms of a transition from solitary to coordinated work. In remarks cited by Salesforce, Seaman said the “single-player era of AI” can make individuals faster, but the “multiplayer era” makes a company smarter together when a team tackles the same problem with shared momentum. He also argued that when AI operates in a shared workspace, it can reflect how a business actually works, including projects, debates, and decisions that live in team conversations.

The survey includes concrete examples of how useful discoveries become stranded. One respondent described a scenario in which one person learns a prompt that can cut review time dramatically, but no one else knows it, so the technique remains locked inside that individual’s AI workflow. Another IT director said engineers wasted time prompting AI to fix the same error because separate chat histories prevented the team from sharing what worked.

In software development and other technical functions, the costs of siloed AI can show up as both repeated effort and system-level friction. Salesforce reported that survey respondents cited “duplicated work” and “major systems conflicts” tied to uncoordinated AI coding. Beyond duplication, the survey also highlighted a visibility problem, where private chat logs and separate interfaces leave parts of the organization unaware of what has already been discovered or corrected.

Salesforce said the underlying math is straightforward: in larger organizations, more employees generally means more private chat histories and more places where work can go unseen. For technical teams, executives suggested that blind spots can translate into time spent troubleshooting issues that might otherwise have been prevented. One accounting director, for example, said clear visibility into AI-generated code could have saved developers hours of debugging.

There are also adoption and rollout implications. The survey suggests many organizations are beyond basic AI experimentation but have not effectively tied their AI efforts together across teams and tools. Salesforce attributed the gaps in part to a lack of coordinated implementation, where AI insights are generated in isolated channels and then fail to connect to shared execution. The survey also notes that it surveyed decision-makers over AI investments, not end users, and it did not spell out what specific platform configurations, integration patterns, or product changes would most directly solve the problem.

For executives and operations teams, the next question is how to move from individual productivity gains to shared organizational learning. Salesforce’s survey sets up a clear yardstick, centered on shared context, cross-system integration, and reduced duplication. What to watch next is whether more companies begin measuring AI success not only by output quality, but also by collaboration metrics such as reuse of prompts, speed of cross-team debugging, and the extent to which AI-generated work becomes visible and actionable across the organization.