A manager sits down to write a performance review.
It’s tedious and time-consuming. They have five reviews due by the end of the week.
So they open ChatGPT, paste in some notes about the employee, and ask it to draft a professional performance review.
The AI generates polished text. The manager makes a few edits and submits it to HR.
HR has no idea AI was involved. Legal has no idea. The employee has no idea.
Three months later, that employee is terminated based partly on that performance review.
Six months after that, the wrongful termination lawsuit begins.
During discovery, the manager admits they used AI to write the review.
Now the company is defending why AI was used to evaluate this employee without any oversight, testing, or policy.
Welcome to the newest source of employment liability.
What’s Actually Happening Right Now
Managers across every industry are using AI for HR decisions.
They’re not using official HR systems approved by legal and IT. They’re using consumer AI tools like ChatGPT, Claude, and Gemini, whatever’s easiest to access.
They’re using AI to write performance reviews, draft termination memos, summarize employee feedback, make promotion recommendations, evaluate candidates, and write disciplinary warnings.
They think they’re being efficient while creating massive legal exposure.
HR departments have zero visibility into it because it’s happening on individual managers’ laptops and phones.
Why Managers Are Doing This
Performance reviews are tedious and time-consuming. Writing five or ten reviews takes hours.
AI can draft a review in 30 seconds that sounds professional, uses proper HR language, and saves time.
Termination memos are stressful to write because managers worry about saying the wrong thing.
AI provides templates and makes the language sound official and defensible.
Promotion decisions require documentation, and managers need to articulate why one person was promoted over another.
AI helps structure the reasoning and makes weak justifications sound stronger.
Nobody told managers this was prohibited. Most companies have no policy about using AI for HR decisions, so managers assume it’s fine as just another productivity tool.
Why This Creates Massive Liability
Employment decisions are the highest-risk AI use case.
Hiring, promotion, performance evaluation, and termination are exactly the decisions that discrimination laws regulate.
When AI influences these decisions, even indirectly, it creates legal exposure.
The bias problem: AI trained on internet data reflects societal biases. When managers use it to write performance reviews or evaluate employees, those biases can seep into the language through descriptions of female employees’ leadership styles that differ from descriptions of male employees’, coded language that correlates with protected characteristics, and emphasis on different qualities based on age, race, or gender.
The manager doesn’t intend bias, but the AI suggests language, the manager uses it, and discrimination gets embedded in official HR documents.
The consistency problem: Different managers using different AI tools with different prompts creates inconsistent evaluation standards.
One manager’s ChatGPT-generated review emphasizes certain qualities while another manager’s Claude-generated review emphasizes different qualities, with no consistent methodology.
When employees compare reviews and see inconsistencies, it suggests arbitrary or discriminatory treatment.
The documentation problem: AI-generated HR documents become evidence in litigation.
If a manager used AI to write a termination memo, opposing counsel will ask for the AI prompts to understand what the manager told the AI about the employee, what options the AI suggested, and why the manager chose this language over alternatives.
The AI’s role in the decision-making process becomes part of the legal record, and companies can’t defend what they didn’t know was happening.
Reason WHy HR and Legal Don’t Know This Is Happening
HR departments are focused on official systems like applicant tracking systems, HRIS platforms, and performance management software.
They govern those systems, test them for compliance, and train managers on proper use.
Consumer AI tools aren’t on HR’s radar because they’re not enterprise software but free websites managers access on personal devices.
Legal departments are focused on AI the company deploys, vendor AI systems and AI integrated into products.
Individual managers using ChatGPT for performance reviews doesn’t show up in legal’s risk assessment because it’s not a formal AI deployment.
The risk is invisible until litigation forces it into the open.
What Companies Should Have (But Don’t)
Most companies have zero policies on managers’ use of AI for HR decisions.
Here’s what they need:
Clear AI Use Policy for Employment Decisions: “Managers may not use AI tools (including ChatGPT, Claude, Gemini, or similar) for any HR-related purpose, including performance reviews, termination memos, promotion decisions, or employee evaluations, without prior approval from HR and Legal.”
Required Disclosure: “Any use of AI in employment decisions must be disclosed to HR with AI-generated content clearly marked and original AI outputs retained for documentation.”
Training for Managers: “All managers must complete training on AI risks in employment decisions before using any AI tools for HR purposes.”
Alternative Approved Tools: “If managers need assistance with HR documentation, they should use company-approved tools with built-in compliance safeguards.”
Consequences for Violation: “Unauthorized use of AI for employment decisions may result in disciplinary action.”
Most companies have none of this, while managers are using AI freely without any oversight.
What Needs to Happen Immediately
Issue a clear policy: Immediately communicate to all managers that AI cannot be used for HR decisions without approval.
Conduct amnesty review: Give managers a short window to disclose if they’ve used AI for HR decisions, then review those decisions for bias or problems.
Provide training: Educate managers on why AI in employment decisions creates legal risk and make it clear this isn’t about efficiency but about liability.
Offer alternatives: If managers genuinely need help with HR documentation, provide approved resources like HR templates, writing guides, and access to HR for consultation.
Implement monitoring: Work with IT to detect if managers are using consumer AI tools for work purposes and flag HR-related use for review.
Update discovery holds: In any employment litigation, specifically ask if managers used AI tools and preserve any AI-generated content and prompts.
Ask Yourself This
How many managers in the organization right now have used ChatGPT or similar AI to help write performance reviews, termination memos, or promotion justifications?
The answer is probably “we don’t know,” and that lack of knowledge is the problem.
When the wrongful termination lawsuit arrives, and discovery reveals AI was used without oversight, “we didn’t know” won’t protect the company.
Schedule a 30-minute Manager AI Use Assessment where we’ll:
- Review current policies (or lack thereof) on manager AI use
- Show what the AI use policy for employment decisions should include
- Identify gaps in HR oversight of manager AI use
- Provide manager training materials on AI employment risks
- Help create disclosure and review processes
We’ve helped dozens of companies address shadow AI use in HR before it becomes employment litigation. Managers using ChatGPT to write performance reviews feel like a productivity hack until the discrimination lawsuit reveals what the AI actually suggested.
Brian Alexander
Founding Partner, VisionAI+ Consulting Group
Former General Counsel, AI Engineering Firm
Certified ISO 42001 Lead Auditor