Executive Summary
When a company explains a decision by pointing to AI, 41% of employees think the real reason was something else. That comes from The Harris Poll, which surveyed 2,000 non-manager employees and 1,003 managers at companies with more than 500 staff. The Grossman Group, an organizational communication consultancy, commissioned the work and released it on November 19. The study is called the Six Gap Study.
The same decision drew different reactions depending on the reason given for it. 32% of employees strongly distrusted an explanation that gave AI as the reason, and 15% strongly distrusted one that gave economic conditions. Neither 41% nor 32% is a majority. The finding is not that most employees distrust their employer, but that among the explanations on offer, the AI one gets doubted first.
Sections 1 through 4 report the published findings and what the coverage carries. Section 5 is the reading Pebblous draws from them. What the numbers point at is not the performance of the AI, but where the grounds for the decision are kept.
Key Numbers
Four numbers summarize this survey: the size of the doubt, how that doubt splits by the reason given, how far managers and employees read the same thing differently, and the point where distrust turns into an intention to leave.
Source: the Harris Poll Six Gap Study release commissioned by The Grossman Group (November 19, 2026), and the FinancialContent page carrying it.
41%
Employees who think a decision attributed to AI had another reason
Asked of the 2,000 non-manager employees
32% to 15%
Strong distrust — the AI reason against the economic one
One survey asking about the two reasons separately
26 points
Gap on job insecurity between managers (69%) and employees (43%)
Managers read the fear as larger than it is
56%
Of employees who hide their AI use, the share open to a similar-paying job elsewhere
Among those who use it openly it was 32%
Distrust More Than Doubles When the Explanation Points to AI
The survey asked about moments when a company explains a decision that lands on employees, such as a restructuring or a shift in priorities. 47% of employees believe they get the full story when a leader explains one. That is under half, and on its own it is not news. The part that stands out arrives when the reason is narrowed to AI. Only 40% believe they get the full story then.
Asked how strong that distrust is, the difference widens. 32% strongly distrust an explanation that gives AI as the reason, against 15% for one that gives economic conditions. The economy can be checked in the news and happens outside the company. AI happens inside it, and who looked at which data to reach which judgment never comes out. The reason that cannot be checked drew the heavier doubt.
What this survey covers is not wide. It ran online in the United States, among non-manager employees and managers at companies with more than 500 employees. The main wave of 2,000 and 1,003 ran in the summer of 2026, and on September 24 and 25 a further 218 employees answered a few more questions. The published write-up does not separate which figures came from which wave. The numbers here are the answers of people working at US companies with more than 500 employees, and they are hard to read as an average for workplaces in general.
What employees doubt is not AI itself. The question put to them was whether a decision their company attributed to AI was really made for another reason, and 41% said yes. That is distrust of the explanation rather than of the technology. The appeal to AI, used the same way elsewhere, also fails to line up with the numbers in layoff statistics, which this blog looked at in July (related article).
Managers Overstate the Fear and Overrate Their Own Message
This is where the separate sample of 1,003 managers earns its place. 69% of managers think employees feel their jobs are insecure as the company leans further on AI. Among employees, 43% say so. The gap is 26 percentage points, and it runs in the direction of managers reading the fear as larger than it is.
A gap runs the other way as well. 67% of managers believe employees get the full story when AI is given as the reason, while 40% of employees feel they do. That one is 27 points. Two errors sit next to each other inside one survey: the fear read large, the communication read as working.
The two gaps answer different questions, so they cannot be added together. Their direction is the same. Managers fill in the feeling coming from below with guesswork, and count what they passed along as received. The spot in the middle, the one that relays the explanation, turns out to be the darkest. It is also where employees go first: 62% say they turn to their direct manager when they suspect a leader is not telling the whole story.
Distrust Shows Up as Hidden AI Use and Ideas Kept Quiet
The distrust did not stay in people's heads. 75% of employees say they have changed how they work because they doubted a leader's explanation. 39% kept an idea or a concern to themselves, and 21% started looking at other jobs. A meeting where nothing gets said and a quiet job search grow from the same root.
The use of AI goes into hiding too. 34% of employees say they use AI heavily at work, and 27% of them say they use it without telling their leaders. Of the employees who hide it, 56% say they would move for a similar-paying job elsewhere. Among those who use it openly, that figure is 32%. The ones hiding it are also the ones more likely to go.
It tracks with how employees rate leadership. Among those who call their company's leadership excellent, 48% use AI heavily, while among those who rate it lowest, 26% do. In a low-trust organization AI use does not fall, it moves out of sight. Whatever adoption rate the organization counts comes out lower than the real one.
88% of employees say they would rather hear bad news directly, including the part AI played in it. No calm comes from not telling. What accumulates instead is usage nobody counts and an intention to leave nobody counts.
What the AI Spending Has Not Bought
David Grossman, founder of The Grossman Group, put it this way in the release. CEOs are asked to account for every dollar of AI spending, and yet they have not accounted for it to the employees who will decide whether that spending is worth anything. In the same release he added that the next obstacle to returns on AI looks less like capital and more like leadership. That is the reading of the party that commissioned the survey, which is worth keeping in mind.
There is a backdrop the diagnosis aims at. In a global survey McKinsey published in 2026, 80% of respondents said AI had raised their personal productivity, while 37% said AI had a positive effect on their organization's bottom line. Those figures come from a separate source, not from the Harris Poll study. Individuals feel an effect that does not surface as organizational performance. This survey names trust as one strand of that gap.
The numbers in the two surveys are made in different ways. The Harris Poll methodology states that the results reflect respondents' reported experiences and perceptions. In a survey about trust, what people say is the thing being measured, so it can be read as it stands. A report that productivity rose is a different kind of answer. It takes something measurable and accepts a person's memory in its place. How an answer picked in a questionnaire becomes a line in a report is something this blog looked at before with the Bank of England. The AI productivity figures in the November 5 Monetary Policy Report began as a choice among five options made by company executives (related article).
A figure of the same kind turns up once more. In the Listening Project run by the Milken Institute together with The Harris Poll, more than two thirds of leaders expected to be financially better off in five years thanks to AI, while 27% of ordinary workers said the same. Harris Poll CEO John Gerzema described the corner office and the cubicle as living two different realities inside one economy. That is a separate survey from the Six Gap Study, with a different sample and different questions, so the figures from the two should not be read together.
Why Pebblous Is Watching This Survey
AI makes decisions, and it also works as a name laid over decisions people made. That is what the distance between 32% and 15% says. Economic conditions can be checked from outside, so an explanation resting on them is verifiable. A statement that the AI judged it that way cannot be checked unless what went into the judgment and what it turned on come out alongside it. An explanation that cannot be verified gets processed as either believed or not believed, and the survey shows which side employees pick.
Pebblous has taken up the same question twice this year. The July article looked at how the AI label on layoffs fails to line up with the overall layoff statistics, from a data integrity angle (related article), and the September one looked at the gap in a Connecticut law that requires a WARN notice to state whether layoffs relate to AI without asking what that judgment rests on (related article). After whether the label matches the statistics, and whether the law demands disclosure, this survey is the third layer. Employees already do not believe the label.
The three layers point at one gap. There is no record to pull out afterwards showing what relation a decision had to AI. Without a record of which data was looked at, what the model returned, and how far a person accepted that result, the statistics and the disclosure and the in-house explanation all become statements made after the fact. This is where the question that decides whether an AI rollout works moves from model performance to whether the grounds for a decision were kept as data.
Korea has the same empty box. The filing form for a collective dismissal records the reason, the number of positions, the consultation and the schedule, and carries no place to record which technology was brought in (related article). New York State made the box first, adding technology and automation as options on its mass-layoff filing form in March 2025. Over the following year, among roughly 750 filings, not one selected them. Creating a box and having the grounds to fill it in sitting inside the company are two different problems.
This article could not open the GlobeNewswire page directly. The figures here were cross-checked against a FinancialContent page carrying the same release and against search results, and the remarks attributed to the two executives rest on secondary coverage of that release.
References
Primary sources
- 1.The Grossman Group. (2026). "New Study Suggests Employee Distrust Is a Hidden Risk to Companies' AI Returns." GlobeNewswire, 2026-11-19. Release for the Harris Poll Six Gap Study.
- 2."New Study Suggests Employee Distrust Is a Hidden Risk to Companies' AI Returns." FinancialContent, 2026-11-19. The page carrying the same release, used to confirm the figures in this article.
Background — separate surveys
- 3.Milken Institute & The Harris Poll. (2026). "The 2026 Milken Institute–Harris Poll Listening Project." The outlook gap between leaders and workers. A separate survey from the Six Gap Study discussed here.
- 4.McKinsey. (2026). "The State of AI" global survey. The gap between 80% on personal productivity and 37% on organizational bottom line. A separate source from the Harris Poll study.
- 5.INTOO & The Harris Poll. (2026-01). An earlier survey in which 69% of employees called AI-driven layoffs just an excuse. A separate survey with a different sample and different questions.