---
type: "synthesis"
tags: ["shadow-ai", "clandestine-use", "tension"]
articles: ["a036", "a040", "a052"]
id: "cross-shadow-ai-three-readings"
sources: ["adoption"]
sourceVaultSlug: "hbr-seg-adoption"
originDay: 9
articleStem: "hbr-seg-adoption"
sourceUrl: "(unified vault: 11 sources)"
sourceTitle: "HBR — People Ⅲ-A · Adoption / trust / literacy / psych-safety"
---
Employees secretly using unsanctioned AI is described in at least three articles — but each gives it a *different causal story*, and the differences matter for the fix.

1. **A036 (Chamorro-Premuzic)** — [[concept-clandestine-ai-use]] is a **rational response to input-based evaluation**. Workers cut effort 30–40% but *hide the gain* to avoid being loaded with more work. The remedy is to reward output not input (see [[cross-incentives-metrics-redesign]]).
2. **A040 (Deloitte)** — [[concept-shadow-ai-solutions]] signals **distrust of the employer's specific tools, not of AI**. Because workers actively seek out shadow tools ([[claim-shadow-ai-preference]]), the diagnosis flips from technophobia to a co-creation failure ([[contrarian-shadow-ai-trust]]). It raises an unmanaged security risk ([[question-shadow-ai-security]]).
3. **A052 (Hermann et al.)** — [[concept-shadow-ai]] is a form of **[[concept-maladaptive-coping]]** driven by frustrated psychological needs; motivations include seeking a 'secret advantage' and fear of being fired (data: [[entity-ivanti|Ivanti]], BCG).

Synthesized: shadow AI is simultaneously an *incentive artifact* (A036), a *trust signal* (A040), and a *coping mechanism* (A052). All three agree it is evidence the workforce **wants** to use AI — the problem is the conditions of use, not the will. That converges with the corpus's dominant prescription (co-creation, safe incentives) and complicates the pure 'ban it' governance instinct.