A sharp turn in workplace AI sentiment, despite surging adoption signals
The latest Glassdoor survey data points to a striking contradiction at the heart of today’s enterprise AI moment: employees are talking about AI far more, but liking it far less. Mentions of AI in workplace reviews rose 240% between May 2025 and May 2026, reflecting how quickly large language models (LLMs) and adjacent automation tools have moved from experimental pilots to everyday infrastructure. Yet sentiment has flipped. Where AI-related commentary was 81% positive in 2019, it now sits at 43% positive, while negative sentiment has climbed to 53%.
This is not merely a story about “fear of the future.” It is evidence of a widening gap between AI as promised—a productivity multiplier, a creativity aid, a competitive moat—and AI as experienced by many workers: a disruptive layer added to already complex jobs, often without sufficient training, agency, or clarity on how performance data will be used.
The occupational breakdown underscores how unevenly AI’s benefits and burdens are distributed. The most negative sentiment clusters in roles where language, judgment, and process integrity are central:
- Writers, editors, and journalists: 81% negative
- Accountants: 80% negative
- Insurance claims adjustors: 98% negative
These are not fringe categories; they are knowledge-work pillars. Their pushback suggests that AI’s integration is colliding with professional identity, quality standards, and accountability norms—not just task lists.
Why employees are pushing back: displacement fears meet “digital Taylorism”
The survey’s grievances map to a set of recurring workplace dynamics that many organizations underestimate. Employees cite:
- Job displacement fears (20%)
- Pressure to adopt unfamiliar tools (14%)
- Distraction from core duties (13%)
- Heightened surveillance framed as productivity tracking (10%)
The displacement concern is the headline, but the deeper story is cultural and operational. In many workplaces, AI is arriving as a management system as much as a work assistant. LLMs are being embedded into product development, internal communications, workflow routing, and performance oversight. That breadth of integration can feel less like “augmentation” and more like a re-architecture of how work is defined, measured, and valued.
A particularly combustible element is the rise of AI-enabled monitoring—what critics often describe as digital Taylorism. When organizations layer analytics and automated evaluation onto daily activity, the relationship between employer and employee can shift from trust-based autonomy to continuous measurement. Even when deployed with benign intent, the perception of surveillance can:
- Reduce discretionary effort and experimentation
- Encourage risk-avoidance and “performative productivity”
- Erode psychological safety, especially in creative and judgment-heavy roles
For writers and journalists, the friction may be intensified by concerns about originality, voice, and editorial standards—areas where AI can be useful but also controversial. For accountants and claims adjustors, the stakes are different but equally acute: accuracy, compliance, and liability. In these domains, AI can accelerate throughput while simultaneously raising the question: who is accountable when the model is wrong?
The economic subtext: AI as cost lever, not capability investment
The sentiment reversal also reflects macroeconomic realities. In a tight labor market and margin-pressured environment, AI is frequently positioned—explicitly or implicitly—as a cost-containment tool. Even when leaders describe AI as an assistant, workers often interpret deployment as a prelude to:
- Headcount reduction
- Slower hiring and backfilling
- Role consolidation and wage compression
This is where the narrative hardens into “job killer” framing, particularly in functions where work is already heavily standardized. If employees see AI adoption paired with hiring freezes, reorganizations, or aggressive productivity targets, the technology’s intent becomes suspect regardless of its actual capabilities.
Another driver is the uneven distribution of AI gains. Digitally native roles and teams with strong enablement (training, prompt libraries, workflow redesign, governance support) can experience AI as empowering. Legacy functions—especially those with limited upskilling budgets or unclear career pathways—may experience AI as something being done *to* them rather than *with* them. That asymmetry breeds resentment and can create a two-tier workforce: those who are “AI fluent” and those who are perpetually catching up.
For employers, this is not just an HR issue. It is a strategic risk. If frontline adoption stalls or becomes performative, AI ROI assumptions—often built into transformation roadmaps and investor narratives—can quietly fail to materialize.
What durable AI transformation looks like: participation, governance, and measurable trust
The survey’s most important signal may be this: employee objections are increasingly about how AI is implemented, not whether AI exists. That distinction creates a clear path forward for organizations that want both productivity and legitimacy.
Several practices stand out as pragmatic responses to the current trust deficit:
- Participatory implementation
– Create cross-functional councils spanning frontline staff, HR, legal, and IT to evaluate tools and workflows.
– Build AI champions networks to support peer-led training and continuous feedback, reducing the “top-down edict” effect.
- Responsible AI governance employees can actually see
– Clearly communicate monitoring boundaries: what data is collected, how it is used, and what it is *not* used for.
– Establish guardrails such as bias audits, human-in-the-loop exceptions, and escalation paths for high-stakes decisions.
- Upskilling tied to real career mobility
– Pair deployments with role-specific training that improves employee market value, not just company efficiency.
– Publicize internal examples where AI led to role elevation—more analytical work, more client-facing responsibility, higher-quality output—rather than quiet redundancy.
- Metrics beyond efficiency
– Track sentiment, retention, error rates, customer impact, and innovation velocity alongside cost savings.
– Use those measures to recalibrate AI roadmaps when adoption generates cultural or operational drag.
Finally, the external environment is tightening. Regulatory regimes—from the EU AI Act to evolving U.S. sector guidance—are pushing companies toward clearer accountability, documentation, and risk controls. Organizations that treat compliance as a strategic advantage, rather than a burden, may find it easier to build internal trust: employees tend to respond better when rules are explicit and protections are enforceable.
The Glassdoor data reads less like a rejection of AI and more like a referendum on organizational change. Companies that align AI strategy with human-centered implementation—participation, transparency, and credible career pathways—are more likely to convert today’s skepticism into durable adoption, while those that treat AI as a unilateral mandate may discover that the hardest part of automation is not the model, but the workforce asked to live with it.
By
By
By
By
By
