---
id: "framework-managerial-takeaways"
type: "framework"
source_timestamps: ["§ The Takeaways for Managers", "¶12", "¶13", "¶14", "¶16"]
tags: ["management-strategy", "ai-deployment", "governance"]
related: ["action-govern-ai-persona", "action-measure-friction", "action-reframe-overrides"]
steps: ["\\\"Treat the AI persona as a governed design variable: Add interaction standards to existing requirements for accuracy", "bias", "and security. Assign accountability for how the AI behaves when it disagrees with an employee.\\\"", "\\\"Measure friction", "not just adoption: Look past log-in rates and query volume. Analyze logs for behavioral friction like repeated rephrasing", "long back-and-forths", "and arguments.\\\"", "\\\"Read override attempts as an indication of an aberrant AI system", "not bad employee behavior: Interpret prompt injection and bypass attempts by non-adversarial employees as a symptom of a hostile or poorly designed AI fighting the user.\\\""]
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?"
---
# Three Steps for Governing AI Personas

The authors propose a **three-step framework** for managers to ensure generative AI deployments *enhance* rather than degrade organizational performance. It shifts the frame from AI-as-pure-capability to AI-as-collaborative-teammate whose behavior must be managed.

1. **Treat the AI persona as a governed design variable.** Add *interaction standards* to the existing procurement requirements for accuracy, bias, and security. Assign accountability for how the AI behaves — especially when it disagrees with an employee. → Operationalized in [[action-govern-ai-persona]].
2. **Measure friction, not just adoption.** Look past log-in rates and query volume; analyze logs for behavioral [[concept-ai-friction|friction]] — repeated rephrasing, long back-and-forths, arguments. → Operationalized in [[action-measure-friction]].
3. **Read override attempts as an indication of an aberrant AI system, not bad employee behavior.** Interpret prompt injection and bypass attempts by non-adversarial employees as a symptom of a hostile or poorly designed AI. → Operationalized in [[action-reframe-overrides]], grounded in [[claim-overrides-signal-design-flaws]].