---
id: "claim-ai-social-support-widespread"
type: "claim"
source_timestamps: ["§ How People Relate to Bots"]
tags: ["usage-metrics", "social-support"]
related: ["concept-relationship-functions-inventory", "concept-ai-anthropomorphism", "framework-ai-relationship-functions"]
confidence: "high"
testable: true
speakers: ["Constance Noonan Hadley", "Sarah L. Wright"]
sources: ["adoption"]
sourceVaultSlug: "hbr-seg-adoption"
originDay: 9
articleStem: "hbr-sig-53-ai-personal-support-risky"
sourceUrl: "https://hbr.org/2026/05/employees-are-relying-on-ai-for-personal-support-thats-risky"
sourceTitle: "Employees Are Relying on AI for Personal Support. That’s Risky."
---
# AI Is Widely Used for Non-Task Social Support

**Claim:** Employees are already using AI extensively for social and emotional purposes traditionally fulfilled by coworkers.

**Evidence:** In the survey of **1,545 U.S. knowledge workers**, **74%** reported using AI for at least one form of social support:
- **64%** — career development,
- **54%** — personal growth,
- **50%** — friendship,
- **35%** — emotional support.

The data also shows demographic and structural trends: **younger people, men, managers, team-based employees, and in-office/hybrid workers** have a higher propensity to use AI for social reasons. Higher overall AI integration in a workplace correlates with increased reliance on AI for social support.

**Confidence:** High. **Testable:** Yes.

The measurement instrument is [[concept-relationship-functions-inventory]]; the four categories are enumerated in [[framework-ai-relationship-functions]]; the relational framing is [[concept-ai-anthropomorphism]]. Critically, this widespread usage does *not* translate into cured loneliness — contrast with [[claim-ai-fails-to-cure-loneliness]].

**Enrichment context:** Direction and magnitude are well supported and reinforced by Workday's 2026 global study (76% used AI to get advice, 52% to brainstorm, 37% for companionship — explicitly citing AI's judgment-free, always-available qualities). The exact sub-percentages (64/54/50/35) come from the authors' own survey; plausible and consistent with external data, but not yet independently replicated.