Oct 15, 2026 · 23 min
AI-written reviews put managerial trust under pressure
Ghostwritten Feedback: Managers' Hidden Use of AI in Performance Conversations, and How Leaders Can Protect Trust and Development
When managers outsource feedback, polished language can conceal weaker judgment, amplify bias, and damage the human relationship employees need for development.
- 1Undisclosed AI use can make accurate feedback feel less trustworthy once employees learn a machine produced it.
- 2Automation bias may erase contributions, accept hallucinations, and flatten evaluations into generic organizational language.
- 3Managers should disclose AI use, retain judgment, and use tools for rehearsal, evidence gathering, and bias auditing.
Don't miss
The episode’s sharpest turn comes when technically accurate AI feedback is compared with having someone else write a personal apology, exposing why correctness cannot substitute for human effort.
The brief
A polished review can trigger suspicion when an employee realizes the manager may not have written it. The episode uses that scenario to frame AI’s central performance-management problem: efficiency without trust.
Managers are turning to AI because feedback is time-consuming, difficult, and increasingly encouraged by executives, yet disclosure to employees remains limited. That gap makes adoption a trust issue, not merely a workflow change.
AI-generated feedback can be accurate and data-rich while still weakening motivation, especially when employees shift from learning about their work to wondering whether their manager cared enough to engage.
The episode challenges AI’s supposed neutrality: biased historical evaluations can produce longer, more critical, personality-focused reviews of women, while automation bias hides omissions and hallucinations behind authoritative prose.
The proposed safeguard is human-centered: disclose AI use, keep ratings and judgment with managers, use secure tools for rehearsal and bias audits, and reinvest saved time in substantive coaching.
The closing warning extends beyond compliance: if employees trust neither machine-perfect feedback nor human mistakes, they may hide errors and stop taking the risks that support development.