Outright Refusal
Ignoring company-mandated AI tools entirely. Completing tasks the old way and never opening the tool.
50% actively resist
Research briefing · August 2026
Employees Are Sabotaging AI. Here's What the Data Says.
14 studies, 20,000+ workers surveyed. A synthesis of the research on why employees resist, sabotage, and fake their way through AI adoption — and the signals managers can actually watch for.
29%
deliberately sabotage their company’s AI strategy
Writer × Workplace Intelligence
53%
submit work slop — AI output that looks passable but lacks substance
BetterUp Labs
50%
describe themselves as actively resisting AI tools
Software Finder
66%
use unauthorized AI tools, bypassing corporate guardrails
PagerDuty
The sabotage spectrum
It's not just people refusing to use AI.
The research reveals six distinct behaviors, ranging from passive avoidance to active undermining.
Faking Usage
Pretending to use the tool but actually doing the work manually. The tool stays open on screen; the work happens elsewhere.
13% fake AI use
Work Slop
Submitting AI-generated output that looks polished at a glance but lacks substance. Shifts cognitive load to the recipient.
53% submit slop
Shadow AI
Using unapproved public AI tools instead of sanctioned ones. 89% of workplace AI use escapes governance.
66% use unauthorized tools
Data Leaking
Entering proprietary company info — customer records, financials, strategy docs — into public AI tools.
35% leak data to public AI
Hiding AI Use
Using AI effectively but concealing it — to keep a competitive edge, avoid extra work, or dodge stigma.
57% hide AI usage
Why they do it
Fear of replacement is the headline, but it's not the only driver.
The research shows overlapping motivations that feed one another.
60%
Fear of job elimination
77%
Increased workload
49%
AI diminishes my value
75%
Strategy is performative
28%
Security & privacy worry
The doom loop.
Employees fear job loss → sabotage AI tools → executives threaten layoffs → more fear, more sabotage.
The wage compression angle.
Companies capture productivity gains through wage compression, not layoffs. Bottom-quartile workers lost 10.7% in real wage growth.
The visibility gap
Managers know something is wrong but can't see what.
When managers model AI use themselves, reported AI value lifts 17 points, critical thinking 22 points, and trust 30 points.
6%
of executives believe managers accurately understand employee AI usage
54%
of managers notice receiving work slop
75%
of C-suite admit their AI strategy is more for show
Signals a manager could actually watch for
Mapping research behaviors to observable data points.
These are the signals a platform can surface.
| Signal | What it looks like | Severity |
|---|---|---|
| Low prompt quality | Vague, one-word, or copy-paste prompts that produce generic output. | High |
| Zero agent engagement | Tasks completed manually with zero or near-zero agent messages. | High |
| Agent ≫ user ratio | Many agent messages but almost no user engagement. | Medium |
| Sabotage indicators | Repetitive prompts, padding, or test giveaways. | High |
| High usage + declining quality | Many messages while output quality trends down. | Medium |
| Task completion imbalance | Agent works many tasks but the user completes very few. | Medium |
| Manager not modeling | A manager mandates AI while their own usage is zero. | High |
| Peer contagion | An entire team has low AI usage, not just one person. | Watch |
All 14 studies referenced
Sorted by relevance.
Each entry highlights the single most actionable finding.
Writer × Workplace Intelligence
2,400 workers · 6 countries · April 2026
29% sabotage AI; 44% of Gen Z. 35% leak proprietary data into public AI. 75% of C-suite admit strategy is for show.
BetterUp Labs & Stanford Social Media Lab
U.S. workers · March 2026
53% submit work slop; 54% of managers notice it. 60% believe technology will eliminate their jobs.
Software Finder Survey
1,005 U.S. workers · June 2026
50% actively resist AI. 13% fake usage. Only 16% believe adoption is for genuine value.
PagerDuty Shadow AI Survey
1,250 office workers · June 2026
66% use unauthorized AI tools. Corporate guardrails are being routed around, not followed.
Microsoft 2026 Work Trend Index
20,000 AI users · 10 markets · May 2026
67% of AI impact depends on manager behavior, not tools. Manager modeling lifts value, critical thinking, and trust.
Apollo Global Management
Macroeconomic analysis · 2026
AI-exposed roles see slower wage growth. Bottom quartile lost 10.7% in real wage growth.
Upwork Research Institute
2,500 workers globally · July 2024
77% say AI added to workload. 88% of top AI performers are burned out and twice as likely to quit.
Gartner HR Survey
2,986 employees · 2025–2026
37% do not use AI because co-workers do not. A 24-point gap separates manager experimentation from employee adoption.
Ivanti Everywhere Work Report
Global survey · 2025
32% hide AI usage; 36% hide it for a competitive edge. 46% use non-employer tools.
HBR — AI Workslop Series
Academic research · Jan–Mar 2026
Competence, autonomy, and relatedness are threatened. 45% report fixing AI-dependent colleagues’ work.
Salesforce Slack Workforce Index
5,000 desk workers · 2025–2026
Daily AI usage rose 233% in six months, but fewer than 48% say their workplace is prepared.
BuiltIn — Resistance Taxonomy
Multi-source compilation · 2026
64% plan to avoid AI as long as possible. 95% of AI pilots fail to produce measurable savings.
Deloitte Switzerland + PwC
Global · 2024–2026
61% use generative AI daily, often without manager knowledge. Engaged users are also the most anxious.
Shadow AI Multi-Source
Multiple · 2025–2026
89% of workplace AI escapes governance. Average cost of a shadow AI data breach: $4.2M.