AI is writing your performance review - and maybe your layoff email

AI Is Writing Performance Reviews. HR Must Respond Now

AI has moved from helpful assistant to default co-writer in workplace communication. A new ZeroBounce survey of 1,000 US professionals shows rising use of AI for emails, performance reviews, and even layoff messages. Efficiency is up. Trust and empathy are at risk.

What the data says

• 24% of employees use AI daily to draft or edit emails, with tech workers leading adoption.
• 35% use AI for sensitive messages, and many copy AI text without edits.
• 20% have caught identical AI-generated emails being sent by colleagues.
• 26% suspect their performance review was written by AI, especially younger and tech employees.
• Among laid-off workers, 16% believe AI wrote their termination email; 20% said the messages felt robotic and made them cry.
• 41% of managers have used AI for performance reviews; 17% used it for layoff emails.

Translation for HR: AI is now embedded in high-stakes communication. Employees feel the gap when empathy is missing.

The risk for HR

AI can improve clarity and consistency, but it can also flatten voice, remove context, and standardize language to the point of alienation. When feedback sounds generic, people assume the process is generic. That erodes credibility, engagement, and psychological safety.

Left unchecked, this trend pushes HR toward automation at the expense of judgment. That's a brand problem and a legal risk.

Policy guardrails to implement this quarter

• Define approved use cases: AI can draft early versions of routine notes and summaries. Human-only for final performance reviews, PIPs, promotions, terminations, and investigations.
• Require human ownership: Manager must fact-check, personalize with specific examples, and sign off. No copy-paste without edits.
• Disclosure standard: For formal reviews, include a short note if AI assisted the draft and confirm the manager authored the final version.
• Tone and empathy checklist: Plain language, name the achievement or gap, cite evidence, state impact, offer support, confirm next steps. Ban robotic phrasing.
• Anti-duplication rule: Prohibit identical language across reviews. Include at least three role-specific examples per review.
• Data protection: No PII or sensitive data in consumer tools. Vet vendors, sign DPAs, enforce access controls, and retention limits.
• Bias controls: Use job-related criteria, run bias checks, and keep documentation. See guidance from the EEOC on AI and fairness and the NIST AI Risk Management Framework.
• Version control: Store prompts, drafts, edits, and final messages. Maintain a clear audit trail.
• Manager training: Teach prompt hygiene, bias awareness, feedback models, and tone editing.
• Employee care in layoffs: Live conversation first, then written follow-up. If AI assists, the final message must be human-edited and delivered with support options.

Use AI in performance reviews without losing trust

• Feed facts, not feelings: Provide goals, KPIs, projects, peer feedback, 1:1 notes, and examples. No vague inputs.
• Structure the prompt: Ask for a strengths summary, evidence-based gaps, business impact, development plan, and next-quarter goals.
• Force specificity: Require numbers, dates, and named projects in every section.
• Apply a rubric: Map language to your competency levels and ratings.
• Human edit pass: Personalize voice, remove clichés, and check for unintended bias.
• Read it out loud: If it sounds cold or generic, rewrite before sending.

Layoffs and sensitive notices

• Human-first delivery: Live conversation, then a clear, compassionate written summary.
• AI as a draft only: Use it to tighten wording and ensure clarity. Final language must reflect context and support resources.
• Consistency without sameness: Use a framework, not a template. Every message should reference the person's role and circumstances.

Metrics that keep you honest

• Employee trust and fairness scores post-review cycles.
• Percent of reviews with unique, role-specific examples.
• Time-to-draft vs. satisfaction with clarity and empathy.
• Grievances and appeals related to reviews or terminations.
• Bias audit results and remediation actions.

What to do next

• Audit now: Where is AI used in comms, reviews, and terminations? Who is using which tools? What data goes in?
• Set the guardrails: Publish policy, templates, tone checklists, and an approval flow for sensitive messages.
• Upskill managers: Run short workshops on prompts, tone, bias checks, and evidence-based feedback. For structured learning, see practical AI upskilling by job.

AI can make HR faster. Your job is to keep it human. Clear standards, trained managers, and proof of fairness will decide whether AI strengthens your function or replaces it.