TeamViewer and Sapeo Research’s “Path to the Autonomous Workplace” report finds that senior executives are more willing than middle managers to let AI agents make enterprise security decisions. In a survey conducted in May among 4,200 managers and employees across the US, Australia, Brazil, Canada, France, Germany, Italy, Japan, Singapore and the UK, 24% of senior executives wanted a human involved in those decisions, compared with 46% of middle managers. The survey predates the “rogue AI” stories that later reached newspaper front pages.

The report says autonomous workplace adoption is accelerating, especially in IT: more than one-quarter of respondents described their departments as at least partially autonomous, and 99% of IT leaders were willing to delegate at least one task to AI. However, 61% of all respondents preferred AI to take no independent action. Among those open to greater autonomy, 71% were comfortable only with clearly defined tasks; 5% of all 4,200 respondents were comfortable with AI operating without a human involved.

Respondents said they would be more comfortable with autonomous actions if they could intervene when needed: 70% accepted AI acting autonomously under that condition, while 40% wanted it to detect and fix issues early while keeping them informed. Other safeguards cited were strong security and privacy protections (39%), notifications before major changes (36%), limits on AI access (36%), visible activity logs (33%) and rollback options (31%).

The report also links more frequent AI use with more positive experiences, while cautioning that the association does not prove frequent use alone produces better outcomes. The article notes that the report’s explanation does not address the possibility of growing dependency and cognitive debt, which it says academic studies have identified in the past two years. For 2030, 33% of respondents expect AI to assist them while they continue doing most tasks, 28% expect to delegate more tasks and focus on higher-value work, and 2% expect their role to be eliminated. In a post-publication update, Anthropic warned that agentic AI may soon be able to self-replicate and occupy Internet of Things devices, and suggested securing those devices against that possibility.

TeamViewer and Sapeo Research’s “Path to the Autonomous Workplace” report finds that senior executives are more willing than middle managers to let AI agents make enterprise security decisions. In a survey conducted in May among 4,200 managers and employees across the US, Australia, Brazil, Canada, France, Germany, Italy, Japan, Singapore and the UK, 24% of senior executives wanted a human involved in those decisions, compared with 46% of middle managers. The survey predates the “rogue AI” stories that later reached newspaper front pages.

The report says autonomous workplace adoption is accelerating, especially in IT: more than one-quarter of respondents described their departments as at least partially autonomous, and 99% of IT leaders were willing to delegate at least one task to AI. However, 61% of all respondents preferred AI to take no independent action. Among those open to greater autonomy, 71% were comfortable only with clearly defined tasks; 5% of all 4,200 respondents were comfortable with AI operating without a human involved.

Respondents said they would be more comfortable with autonomous actions if they could intervene when needed: 70% accepted AI acting autonomously under that condition, while 40% wanted it to detect and fix issues early while keeping them informed. Other safeguards cited were strong security and privacy protections (39%), notifications before major changes (36%), limits on AI access (36%), visible activity logs (33%) and rollback options (31%).

The report also links more frequent AI use with more positive experiences, while cautioning that the association does not prove frequent use alone produces better outcomes. The article notes that the report’s explanation does not address the possibility of growing dependency and cognitive debt, which it says academic studies have identified in the past two years. For 2030, 33% of respondents expect AI to assist them while they continue doing most tasks, 28% expect to delegate more tasks and focus on higher-value work, and 2% expect their role to be eliminated. In a post-publication update, Anthropic warned that agentic AI may soon be able to self-replicate and occupy Internet of Things devices, and suggested securing those devices against that possibility.