New Wharton AI Survey: Bosses Use AI Daily, Hype Fades

A new Wharton AI survey shows managers are using artificial intelligence more than ever and enjoying it less. The fourth annual Wharton study, reported by Fortune and first covered this week, found sixty-two percent of managers and higher-ranking employees at large United States companies now use AI every day, up from forty-six percent the year before. Over the same stretch, the share calling themselves excited about AI fell eight percentage points to fifty-one percent. This is the first wave of the four-year study in which daily use and enthusiasm moved in opposite directions.

The gap tells its own story: about a quarter of managers called themselves skeptical, and forty-four percent said they felt cautious, while optimism held steady at fifty-nine percent. The Wharton AI survey is the first wave of the four-year study in which daily use and enthusiasm moved in opposite directions.

Returns are positive, just not strongly so

Eighty percent of the managers in the Wharton AI survey said AI had produced positive financial returns for their organizations. Only twenty-eight percent described those returns as strongly positive. The researchers were careful on one point: the findings do not establish that AI caused those returns, according to coverage of the study. Departments that measured growth outcomes reported strongly positive returns far more often than departments that measured efficiency alone, thirty-five percent versus thirteen percent.

Wharton professor Prasanna "Sonny" Tambe summed up the mood when he said the amazement is kind of wearing off a little bit, according to TokenPost's coverage of the release. He added that the technology is moving faster than organizations can keep up with. That mismatch between speed and readiness runs through the entire survey.

Managers are becoming AI supervisors

The job itself is changing shape. Seventy-two percent of managers said their department's focus has shifted from creating outputs to reviewing, approving, and governing AI-generated work. Among vice presidents and above, that figure reached eighty-two percent, compared with sixty-three percent among managers and directors.

The day-to-day breakdown shows how much human judgment still sits in the loop. Forty-eight percent said AI performed work that required human approval before execution. Thirty-five percent said AI provided suggestions or drafts while a person made the final decision. Seventeen percent reported AI operating autonomously with monitoring or guardrails.

Accountability remains fuzzy. Asked who would be responsible for a consequential error made by an autonomous AI system, twenty-eight percent named a governance or risk team, twenty-six percent named the deploying employee's manager, and twenty-five percent named the employee. Tambe said organizations must determine what it is that can be verified that agents can do, according to TokenPost's coverage of the Wharton AI survey release. Jeremy Korst, a Wharton lecturer and partner at GBK Consulting, framed the management challenge as figuring out what to prioritize day to day.

The entry-level door is narrowing

This is the part that matters most for Gen Z. One-third of managers and higher-ranking employees said their department had slowed or frozen hiring because of AI. Twenty-four percent expected fewer entry-level hires over the next two to five years, compared with fourteen percent who expected fewer executive hires. The junior roles young workers are applying for are the ones most likely to shrink.

Training makes the picture more awkward. Managers were more likely than entry- and mid-level employees to say workers had received adequate AI training, seventy-eight percent versus fifty-six percent. The researchers noted the two groups were not matched within the same companies, so that split reflects perception as much as reality. Half of managers said AI had changed their department's management structure, and seventy-three percent said workflows had changed significantly.

A separate Wharton experiment suggests why the skepticism is warranted. Across three experiments involving more than thirteen hundred participants and nearly ten thousand trials, access to correct AI answers increased accuracy by twenty-five percentage points. Access to incorrect answers reduced accuracy by fifteen percentage points compared with participants who had no AI access. Wharton associate professor Gideon Nave called the pattern "cognitive surrender," describing it as adopting an AI decision without recognizing that the transfer of judgment has occurred.

Gen Z figured this out first

Fortune framed the Wharton AI survey findings as managers learning what Gen Z already figured out, and the claim holds up. Young workers have spent years as the heaviest AI users while staying realistic about its limits. A September Deloitte survey found that most workers hide their AI use on the job, worried that using it openly invites disapproval. The new Wharton AI survey confirms the exhaustion on the other side of the org chart: the people who set the policies now feel the same fatigue.

None of this argues for ditching the tools. The Wharton AI survey shows daily AI use becoming as ordinary as email, which is exactly when the skill stops being a differentiator and starts being table stakes. Young workers who treat AI output as a first draft that needs checking will have an edge over managers who either trust it blindly or distrust it reflexively.

What the Wharton AI survey keeps circling back to is verification. Amazement fades, but the habit of checking AI output before acting is the part that pays. For Gen Z entering the workforce, that means the advantage is not who uses AI the most. It is who keeps their judgment while everyone else hands theirs over. The AI productivity tax and the shift toward cognitive work describe what this transition feels like from the employee side.