Salesforce survey finds 59% of executives see work stall in isolated AI tools

AI can help one employee find an answer without helping the next colleague avoid the same work. In survey findings published October 27, Salesforce says 59% of executives see work stall weekly because it sits inside one person’s AI tool. The findings put a coordination problem at the center of the company’s pitch for shared, team-wide AI.

The Salesforce Pulse Survey covered 250 U.S. executives at director level or above, all with authority over AI investment decisions. Conducted September 14–22, it captures the views of people choosing and funding workplace AI. Its findings describe their observations and estimates, rather than a direct measurement of employees’ output or time spent using tools.

A useful answer can stay invisible

The most common weekly problem was knowledge that never reached the rest of the team: 68% said AI insights or outputs went unshared or were not reused. Another 66% said colleagues unknowingly used AI to solve the same problem separately. Those findings distinguish getting a useful result from making that result useful to others.

Respondents supplied concrete examples of the disconnect. One executive described a person finding a prompt that could cut review time in half, only for it to remain inside that person’s workflow. An IT director said engineers had spent time prompting AI to fix the same error because separate chat histories prevented them from sharing the solution.

Reported delays were more common at the largest organizations. Salesforce defined enterprise businesses as those with at least 2,500 employees, and small and medium-sized businesses as those with fewer than 250.

73%Enterprise executives

Share reporting weekly delays because work remained in one person’s AI tool.

37%Small and medium-sized business executives

Share reporting the same weekly delays at smaller businesses.

Deploying agents is not the same as connecting them

Nearly half of respondents, 49%, said their organizations were deploying AI agents—software that completes tasks or workflows with some autonomy. But only 15% said agents were embedded seamlessly across multiple systems and interfaces. The survey therefore separates the presence of task-performing AI from its integration into the wider business.

Executives estimated that employees lost an average of 3.6 hours each week switching between AI tools and other work systems. Only 11% said their AI tools sat within one integrated ecosystem and cost no switching time. Salesforce translated the average estimate into about 181 hours, or roughly 22 business days, per employee annually.

That annual figure is an extrapolation, not a year of tracked activity. Salesforce’s methodology says it calculated weekly hours using the low end of each response range, then annualized the result across 50 workweeks. The calculation explains the headline-sized number, but its starting point remains executives’ estimates of employees’ switching time.

The preference is clear; the payoff is untested

Asked how they would design AI use from a blank slate, 72% preferred tools that worked across a team with shared context. Another 26% preferred individual use, with the value staying with each person. That is a stated preference for a way of working, not a demonstrated productivity gain from adopting it.

Salesforce presents those preferences as a case for what it calls “multiplayer” AI. Rob Seaman, Slack’s executive vice president and general manager, argues that agents in shared spaces can learn from projects, debates and decisions already visible in team conversations. “Every exchange with an agent out in the open becomes something the next person, and the next agent, can build on,” he said.