回應 · 2026年10月13日 · Far West Consulting
The AI backlash is mostly a rollout backlash
Glassdoor's analysis of US employer reviews finds AI mentions up 240% in a year and now mostly negative. Job loss is one complaint in five. The rest is about how AI was put in front of people, which is the part an employer controls.
目前研究筆記的主體以英文撰寫；繁體中文版本陸續上線中。
For leaders about to put an AI tool in front of a team.
On August 27, Glassdoor’s economic research team published an analysis of how US workers talk about AI in their employer reviews.21 Mentions of AI rose 240% between May 2025 and May 2026. In 2019, 81% of reviews that mentioned AI did so only under “pros.” In 2026, 43% did; 53% mentioned it only under “cons.” This is a review analysis, not a survey: a mention counts as positive or negative by where the reviewer put it, and the report gives no review count.
The useful part is what the critical reviewers said. Glassdoor coded the detailed complaints from a year of reviews. Fear of being replaced is the largest single theme, at 20% of critical comments. Then: being forced to use AI tools (14%), AI distracting from the core business or degrading the customer’s experience (13%), AI used internally in ways that alienate people, such as surveillance or executive communications (10%), leadership expecting unrealistic gains (8%). Only 2% complained that their company gave them no tools, training or support.
Add the middle four together and you get 45% of the complaints against 20% for job loss. None of the four is a property of the technology. Each is a decision someone made about how to introduce it: mandate it, chase it, point it at staff, oversell it.
The positive comments are narrower. Glassdoor found they fall into two groups: the company is benefiting from the AI boom (44%), and the company provides tools, training and support for using AI (41%). Reviewers who listed AI as a pro were also 47% less likely to apply for another job on Glassdoor in the following month. Glassdoor calls that an association in one month of data, not a cause, and so do we.
What we make of it: people notice enablement when it is there and credit the employer for it. When it is missing, they rarely name the gap; they blame the mandate, the distraction, the surveillance. That asymmetry is why we start with an assessment of how each person actually works, choose the first workflow with the staff who will run it, and train managers before their teams. The backlash in these reviews is against being handed a tool and a target. It is avoidable, and the employers who avoid it get thanked for it in public.
One more figure from the report: at companies with fewer than 200 employees, 51% of AI comments were negative. Above 10,000 employees, 67% were. The bigger the rollout, the more of it is rollout.
People are not mainly afraid of AI. They are unhappy with how it was done to them.