Executives Trust AI Over Themselves—32.8% Admit It

Charter released its AI Pulse research on October 21, 2026, finding that executives who use AI at work were about twice as likely as individual contributors to trust it over their own judgment. That difference matters most when a leader’s judgment determines what a team is asked to deliver.

In Axios’s coverage, 32.8% of executives, 34.8% of managers and 16.3% of individual contributors expressed greater trust in AI than in their own judgment; report coauthor Brian Elliott called AI “really good at giving you the executive summary” and described technology executives making AI-assisted product prototypes over a weekend, then expecting teams to turn them quickly into working products. The August survey covered 5,002 U.S. office workers who use AI at organizations with at least 100 employees.

Managers edge past executives on the trust question

The comparison shows a divide between people in management roles and those without them. It also adds a qualification to the focus on executives: managers had the highest share choosing AI over their own judgment, although the difference between managers and executives was small.

• Executives: 32.8% expressed greater trust in AI than in their own judgment. This group contained 503 vice presidents, C-suite executives, owners and founders.
• Managers: 34.8% expressed the same preference. The survey included 1,547 managers.
• Individual contributors: 16.3% expressed that preference. They formed the largest group, with 2,952 respondents.

The sample consisted of people already using AI at work, so the figures should not be treated as a measure of trust among all U.S. workers or all executives. Job level also tells only part of the story: the survey question compared trust in AI with trust in one’s own judgment, regardless of which tools respondents used, how often they used them or the kinds of decisions they faced. The striking result is the concentration of expressed trust in the roles that direct other people’s work.

Trust in a survey is not a record of decisions

The response captures a general preference, not what happens when a specific AI answer conflicts with other evidence. An executive might express broad confidence in AI and still reject a recommendation after consulting the people responsible for delivery. Another might prefer personal judgment in principle while routinely accepting a generated summary without examining the material behind it. The survey cannot distinguish between those patterns.

Actual reliance would be visible at the decision point: what the tool proposed, which competing information was available, what the decision-maker accepted and whether the final choice changed because of the AI output. Those details matter because using a tool to draft an idea and allowing its answer to settle a disputed question are different acts. A trust percentage alone cannot establish how often either occurs.

The distinction is especially important for managers, whose survey response was slightly more favorable to AI than executives’. Their position between senior leadership and individual contributors can put them in charge of translating a proposal into schedules, assignments and performance expectations. The data identify an attitude at that level; they do not show whether managers gave AI output more weight than the knowledge of their teams.

Why an executive summary can become a decision risk

Cognitive offloading means letting an external tool reduce the mental effort a task requires. For a senior leader, an AI summary can compress unfamiliar information into options that appear ready for a decision. The risk arises when the compression drops delivery constraints, uncertainty or disagreement that would change the choice. A fluent account can make an incomplete proposal feel more settled than it is.

The prototype example makes that gap concrete. A quick AI-assisted demonstration may be useful for exploring an idea, while a functioning product also requires integration, testing, support and decisions about who will maintain it. Treating the demonstration as proof that delivery will be quick can turn an executive’s optimistic timetable into a team’s obligation. The example illustrates a possible mechanism; the survey did not observe how often its respondents made such calls.

Authority changes the scale of the potential consequence. An individual contributor may use an AI answer within a bounded task, while an executive can use one to set a priority, approve resources or impose a deadline across teams. The survey does not measure errors or damage from those decisions. It does show why confidence in AI among decision-makers warrants a closer look at the evidence that reaches them before approval.

The safeguard belongs at approval

For a consequential proposal shaped by AI, the useful check is whether the approver can identify which claims came from the tool and which are supported independently. A review can ask the relevant specialist to test the operational assumptions, particularly where a polished summary has shortened a complex piece of work. That check addresses the gap illustrated by a fast prototype followed by expectations of rapid delivery.

An approval record can also name the human decision owner and preserve material disagreement from people with direct knowledge of the work. These are proposed safeguards drawn from the decision risk, not practices measured in the survey. They would make a future assessment of reliance more meaningful: investigators could compare an AI recommendation with the evidence considered and the decision ultimately authorized.

For employees affected by a leader’s AI-assisted proposal, the consequential question is whether its timetable and assumptions survive scrutiny by the people expected to carry it out. The new trust figures cannot answer that question, but the answer will determine whether confidence in a quick summary becomes a workable decision.