The Ambivalent AI Employee
Some employees saw AI’s potential to improve their work while also worrying that it could replace their tasks or role.
In this article
In Elation’s study, 48 of the 115 employees who reported elevated fear of AI or automation replacing their work also endorsed the belief that AI would enhance it. That is about two in five within this higher-fear group. A positive view of AI’s potential did not necessarily mean that concerns about replacement were absent.
The same person can hold both views
The assessment asked about two kinds of expectation. One concerned automation replacing tasks or AI making a role obsolete. The other concerned whether AI would enhance the employee’s work rather than threaten job security.
Across the whole sample, replacement fear and enhancement belief generally moved in opposite directions. Yet they still appeared together for a meaningful group of employees. Averaging the questions into one overall AI-attitude score would make that combination harder to see.
“Ambivalent” is a convenient description of this pattern. We did not measure whether these employees felt internally conflicted, and we did not identify a permanent employee type.
Two expectations, held together
The overlapping group met the study’s descriptive threshold on both measures.
48 of the 115 employees with elevated replacement fear also endorsed enhancement.
Replacement concern
Worry about automation replacing tasks or AI making a role obsolete.
Enhancement belief
Endorsement of the belief that AI would enhance work rather than threaten job security.
Two views in the same group
The 42% refers to employees in the higher-fear group. It does not mean that 42% of all employees held both views.
The exact proportion depends on where the threshold is set. The important discovery was the overlap itself: concern about replacement could coexist with a positive expectation about what AI might offer.
An encouraging answer about AI’s potential can sit beside a concern about replacement.
An AI attitude is more than a side to pick
I would be cautious about dividing employees into people who are “for AI” and people who are “against it.” Those labels suggest a simplicity the responses did not have. A person may see useful possibilities while remaining uncertain about where their own role fits.
That is one possible interpretation, not a mechanism this study proved. We did not measure AI use, adoption, resistance, or actual job changes. Fear alone cannot tell us whether someone will use a tool, and enhancement belief cannot tell us whether they feel secure.
Elation found the overlap because these expectations were measured separately. The discovery was the combination: an encouraging answer about AI’s potential could sit beside a concern that deserved to remain visible.
Discover what a single AI score can miss
Explore how Elation’s assessment and AI-supported analysis can reveal combinations of employee expectations that an overall attitude score may obscure.
A little more about the research
The source sample contained 706 completed assessments from 691 people across 14 organizations and 16 assessment collections, gathered from July 8, 2025, to March 6, 2026. Fifteen people contributed twice. This was an observational convenience sample, not a representative workforce survey.
Replacement fear was the mean of two questions about task replacement and role obsolescence (α=.820). Enhancement belief was a separate question. Stored favorable-direction scores were used consistently; higher fear scores indicate greater concern. The enhancement wording also mentions job security, so it is not a pure measure of expected productivity benefits.
The primary rule required replacement fear ≥60 and enhancement belief ≥60. There were 115 elevated-fear records from 115 distinct people, with 48 records from 48 distinct people meeting both thresholds: 41.7% of the elevated-fear group and 6.8% of all 706 records. “Ambivalent” is editorial shorthand; internal conflict was not measured.
Latest-response-only analysis gave 47 of 113, or 41.6%. Omitting each organization in turn gave conditional proportions from 39.0% to 45.9%. The overlapping group appeared in 12 organizations. These checks do not establish population prevalence.
The exact proportion depends on the threshold: requiring enhancement ≥80 while keeping fear ≥60 yielded 10 of 115, or 8.7%. Replacement fear and enhancement belief were negatively correlated overall (r=−.393). The views can coexist; they are not statistically independent.
The full study (technical brief), including the measures, checks, and limitations, is available on request.
Request the technical brief ↗