Daily AI Users Feel Safer at Work—Weekly Users Report the Reverse

A CNBC and SurveyMonkey survey released on August 17 found a sharp divide in perceived job security: daily workplace AI users were the only usage cohort in which more respondents felt safer rather than less safe. CIO Dive’s August 18 report says the survey covered 1,686 workers and students in the United States who were polled in July.

The August 17 release establishes an association between frequency of use and workers’ confidence, not evidence that daily AI use protects a job. The gap could reflect familiarity with the technology, differences in job roles and organizational support, or the tendency of already-confident workers to adopt AI more often.

The security gradient changes sharply after daily use

Among daily users, 27% said AI made their current job feel more secure and 22% said less secure—a net balance of five percentage points. Weekly users reversed that result: 10% felt more secure and 25% less secure. Among people using AI once a month or less, the split widened to 3% versus 24%; non-users registered 4% versus 10%.

The detailed CNBC and SurveyMonkey findings also show that 21% used AI several times a day and 8% once a day. Another 34% used it a few times a week or less often, while 37% never used AI at work. Among AI users, 68% said the technology saved them time, rising to 84% for daily users; 56% said it increased confidence in their skills, rising to 73% among daily users.

The wider labor-market question produced a similar pattern. Daily users were nearly evenly divided, with 35% feeling more optimistic because of AI and 33% more pessimistic. Weekly users split 16% optimistic to 35% pessimistic, occasional users 7% to 52%, and non-users 5% to 31%.

Daily users encounter more of AI’s value—and its limits

Daily users employed AI across more types of work than weekly users, including research, brainstorming, writing messages and documents, analyzing data, summarizing material and planning tasks. Repeated use can give a worker more direct evidence about where AI saves time, where its output fails and where human judgment remains necessary. That is one plausible explanation for the confidence gap, but the survey did not test it as a causal mechanism.

Weekly and occasional users may occupy a less comfortable middle position. They have enough exposure to see that AI can perform meaningful tasks, yet may have fewer chances to integrate it into a stable workflow or show how their own judgment complements the tool. The non-user result complicates any simple familiarity theory: non-users were less negative about their own security than occasional users.

Hesitation also varied by experience. Privacy and security were the leading concern for both regular users and occasional users, cited by 37% and 39%, respectively. Occasional users more often cited ethical issues, unreliable or unhelpful results and environmental concerns, suggesting that access to a tool does not automatically produce either trust or frequent adoption.

Policy and training can reduce uncertainty, not guarantee safety

The organizational setting remains unsettled. Fifty-five percent of respondents said their workplace had no official AI policy, while 34% said use was optional. Only 6% worked where AI was required and 5% where it was prohibited. Those categories describe formal rules, not the quality of training or whether employees understand how the rules apply to their work.

For employers, the relevant response is not to promise that heavier use will preserve jobs. Clear policies can identify approved systems, prohibit the entry of sensitive information, establish when human review is required and explain who remains accountable for AI-assisted work. Task-specific training can then show employees how those rules operate in drafting, research, analysis or other workflows they actually perform.

A risk-based rollout also addresses the divide without treating usage frequency as the goal. Lower-consequence tasks can provide structured practice before AI is introduced into decisions that affect customers, employees or regulated data. A useful evaluation would examine whether workers recognize errors, protect information and understand responsibility—not merely how often they open an AI tool.

The survey measures sentiment, not job retention

Correlation is the central limitation. The published results compare self-reported AI use with self-reported feelings about job security at one point in time. They do not show that respondents were randomly assigned to usage groups or followed over time to determine who retained or lost a job.

Self-selection could account for part of the pattern. Workers in technology-intensive or better-resourced roles may be more likely to receive approved tools, useful training and managerial support. People who already feel adaptable may choose to use AI daily, while employees whose tasks appear more exposed to automation may use it only occasionally and feel more threatened.

The evidence therefore supports a precise conclusion: daily users had a small net-positive security balance, while weekly and occasional users had substantial net-negative balances. Establishing whether familiarity changes confidence—or whether confident, well-supported workers become daily users—will require longitudinal research that compares similar employees, workplace conditions and actual employment outcomes.