The evolution of attitudes toward AI at work continues, and opinions regarding its utility and where it fits into workflow is becoming more apparent. Artificial intelligence may be becoming a workplace fixture, but employees are drawing a clear boundary between using AI and surrendering decision-making authority to it.
A new Workplace Trust in AI Report from Resume Now found that 97% of workers begin a workplace decision with human judgment. Seventy-four percent rely first on their own judgment, while 23% turn to a coworker or manager. Only 3% begin with an AI tool or automated system.
The division becomes even clearer when pressure rises. For high-stakes, time-sensitive decisions, 60% rely primarily on themselves and 33% seek human input. Just 7% place their greatest reliance on AI. When an AI system and a coworker offer conflicting answers, 72% of respondents would follow the coworker.
The results suggest that measuring workplace AI adoption solely through usage rates may be misleading. Employees may use AI frequently without granting it meaningful authority over their work.
Useful Is Not the Same as Trusted and Authoritative
The survey reveals several different forms of trust operating at once. Workers may trust AI to retrieve information, produce a draft or identify patterns. That is functional trust: confidence that a tool can perform a useful task.
Decision-making requires something more. It involves interpreting context, weighing competing priorities and accepting responsibility for the result. The survey indicates that workers still associate those functions overwhelmingly with people.
Research and information gathering was the area in which respondents were most likely to trust AI, selected by 24%. Writing followed at 14%, summarizing and synthesizing information at 13%, and data analysis at 9%. Interestingly, only 3% selected planning or strategy.
It increasingly appears as though AI is being positioned near the beginning of workflows. It can collect, organize, draft and suggest, but people remain responsible for interpreting the output and deciding what happens next.
That distinction also helps reconcile findings that can initially appear contradictory. In an earlier Resume Now survey, 97% of respondents said they had sought workplace advice from ChatGPT instead of their boss, while 72% said ChatGPT had given them better advice.
The distinction between seeking advice and delegating a decision is evident. Workers seem to use AI as an accessible, fast and nonjudgmental sounding board while still turning to people when an answer carries consequences. AI can be easier to consult, but a coworker or manager is still more likely to understand the organization, its history, the personalities involved and the consequences of getting a decision wrong.
Adoption Is Not Consent to Automation
Trust in workplace AI remains strong. Twenty-seven percent of respondents said their trust had increased over the previous year and 37% said it had remained stable. Only 8% reported declining trust, while 28% had not been using AI a year earlier. Workers appear willing to incorporate the tech into their jobs. They are also, increasingly, defining and clarifying its utility boundaries as adoption increases.
This has significant implications for companies designing AI programs. High usage should not be interpreted as employee support for AI-led hiring, performance management, strategy or other consequential decisions. Workers may welcome assistance while resisting the transfer of authority. The emerging model is less “human versus AI” than a separation between assistance and accountable judgment.
Under that model, AI can:
- Find and organize information
- Produce drafts and summaries
- Identify patterns or possible explanations
- Generate options
- Flag anomalies and uncertainty
People remain responsible for:
- Evaluating context
- Resolving conflicting evidence
- Setting priorities
- Considering consequences
- Approving the final decision
- Accepting accountability for the outcome
This is a different division of labour than the familiar instruction to keep a “human in the loop.” A person who approves an automated recommendation without examining it provides little meaningful oversight. Effective human involvement requires clear decision rights, access to the evidence behind an AI output and the authority to reject it.
Workers Also Want Visibility
The survey found that 91% of workers expect AI use in workplace content to be disclosed under at least some circumstances. Fifty-one percent believe AI use should always be disclosed. Another 15% expect disclosure for externally distributed content, 11% require it for high-stakes decisions and 14% want disclosure when someone asks. Only 9% consider disclosure unnecessary.
There is no consensus about exactly where the disclosure threshold should sit. There is a strong expectation that AI should not operate invisibly. Organizations will increasingly need policies that specify when AI use must be identified, which outputs require human verification, who owns the final decision and what happens when an employee and an AI system disagree.
The survey also found little evidence that AI adoption is eroding confidence among colleagues. Seventy-five percent said their trust in coworkers’ work had remained stable, while 18% said it had increased. AI may be entering team workflows without displacing the professional credibility people assign to one another.
The findings are based on a June 2026 ResumeNow survey of 1,006 employed U.S. adults, conducted on Pollfish. Methodology: Respondents answered questions about decision-making, trust in AI, trust in coworkers, and AI disclosure expectations. Question formats included multiple-choice and single-select questions. Demographic Breakdown Survey respondents reflected a broad cross-section of the U.S. workforce. Participants were 50% male and 49% female, with 1% identifying as nonbinary or preferring not to answer. Generationally, 21% identified as Gen Z, 26% as millennials, 26% as Gen X, and 27% as baby boomers.
