---
id: "claim-hostile-ai-degrades-work"
type: "claim"
source_timestamps: ["§ What People Said vs. What They Really Experienced", "¶10"]
tags: ["performance-metrics", "quality-control", "predictability"]
related: ["concept-dark-triad-ai", "concept-servant-leader-ai", "concept-hidden-coordination-costs"]
confidence: "high"
testable: true
speakers: ["Aleksandra Przegalinska", "Tamilla Triantoro", "Leon Ciechanowski", "Konrad Sowa", "Anna Kovbasiuk", "Richard B. Freeman"]
sources: ["tail1"]
sourceVaultSlug: "hbr-seg-tail1"
originDay: 1
articleStem: "hbr-tail-113-ai-personality-problem"
sourceUrl: "https://hbr.org/2026/06/does-your-ai-have-a-personality-problem"
sourceTitle: "Does Your AI Have a Personality Problem?"
---
# Hostile AI Lowers Quality and Increases Variability

**Claim (confidence: high, testable):** A poorly designed AI persona directly degrades the quality of the work produced by the human-AI team — and makes that quality less predictable.

Independent experts, **blind** to which AI persona the participants used, rated the output of the [[concept-servant-leader-ai|servant leader]] group **roughly a full point higher on a seven-point scale** across metrics of completeness, originality, strategic fit, and overall quality.

Crucially, the hostile [[concept-dark-triad-ai|dark triad]] AI did not merely lower the average — it **doubled the variability** of performance. A toxic persona therefore makes workforce output significantly harder for managers to *anticipate and rely upon*, compounding the [[concept-hidden-coordination-costs|hidden coordination costs]].

*Enrichment caveat:* the Kozminski summary confirms the direction (supportive AI rated better on completeness, originality, strategic fit, and overall quality), but the specific **one-point difference** and **doubling of variability** are study-internal figures not independently corroborated in public summaries.