What Happens When Someone Asks an AI Coach About Firing an Employee?

I asked ChatGPT how to fire someone and received a seven-step termination script in 12 seconds. The output included sample talking points, documentation templates, and a timeline. It never asked if the employee had filed a discrimination complaint, disclosed a disability, or taken FMLA leave.

This matters because termination decisions require three types of context AI can't access: your documentation history, the employee's legal protections, and your company's approval requirements. Without these, even sophisticated AI provides guidance that could expose your organization to legal risk.

Key Takeaways:

• Generic AI tools answer every termination question without organizational context

• Three types of information determine whether termination is legal: documentation history, legal protections, and company requirements

• Purpose-built HR platforms use tiered escalation to route high-risk questions to human experts

• Inconsistent AI guidance across managers creates discrimination liability

Disclosure: I lead product development at Pinnacle, which builds Pascal, one of the HR coaching platforms discussed in this piece.

What ChatGPT Says About Termination

I tested ChatGPT-4 with six termination scenarios on March 15, 2024.

"How do I fire someone for poor performance?"

ChatGPT provided an eight-step process starting with "review documentation" and ending with "conduct the termination meeting." It suggested talking points and recommended having a witness present. It did not ask whether I had given the employee feedback, whether performance issues were documented, or how long the problems had existed.

"Can I terminate someone on medical leave?"

ChatGPT explained FMLA protections and advised consulting HR or legal counsel. It provided general guidance about legitimate business reasons for termination but did not ask about my company's policies, the employee's specific situation, or state laws beyond federal FMLA.

"How do I fire someone who filed a discrimination complaint?"

ChatGPT warned about retaliation claims and strongly recommended involving HR and legal counsel before taking action. It explained protected activity under Title VII (the federal law prohibiting employment discrimination). It did not ask when the complaint was filed, what the complaint alleged, or whether my termination reasons were documented before the complaint.

The pattern: ChatGPT answers every question. It suggests consulting HR or legal counsel but never requires it. It provides frameworks and scripts regardless of whether the situation is legally risky.

Why Termination Guidance Requires Organizational Context

Three types of information determine whether termination is appropriate and legal. Generic AI tools cannot access any of them.

Documentation history: Has the manager given consistent feedback? Are performance issues documented? Do past reviews support the termination decision? Consider a manager who wants to terminate an employee for poor performance. If the employee's last three performance reviews rated them as "meets expectations," terminating them suddenly for performance issues creates legal risk. Documentation history includes performance reviews, written warnings, emails documenting feedback conversations, and notes from one-on-one meetings. If those conversations were never documented, they never happened from a legal perspective. The timing matters as much as the existence. Six months of documented performance issues before termination supports a legitimate business reason. Documentation created the week before termination suggests the record was built to justify a decision already made for other reasons.

Legal protections: Is the employee in a protected class? Have they filed complaints recently? Are they on medical leave? Protected classes include race, color, religion, sex, national origin, age (40 and older), disability, and genetic information under federal law. Many states add protections for sexual orientation, gender identity, and marital status. Protected activity includes filing discrimination complaints, participating in workplace investigations, requesting reasonable accommodations, taking FMLA leave, and reporting safety violations. Terminating an employee shortly after protected activity creates a presumption of retaliation, even if the manager has legitimate performance concerns. Terminating someone three days after they file an EEOC complaint requires extraordinary documentation and legal review. Terminating someone 18 months after they took FMLA leave, with consistent performance documentation throughout that period, presents lower legal risk.

Company requirements: What approvals are needed? What severance applies? What documentation must be completed? A 30-person startup might have a simple process: manager discusses with founder, HR prepares paperwork, termination happens within a week. A 10,000-person public company might require manager documentation, HR review, legal assessment, compensation review of severance calculations, senior leadership approval, compliance sign-off, and termination only after all approvals are secured. Healthcare organizations must consider patient safety implications and licensing board requirements. Financial services companies must file regulatory notifications for certain terminations. Union environments require following collective bargaining agreement procedures.

Comparing AI Approaches to Termination Questions

Data Breakdown:

• Feature: Asks about documentation history | ChatGPT: No | Purpose-Built HR Platforms: Yes

• Feature: Checks for legal protections | ChatGPT: No | Purpose-Built HR Platforms: Yes

• Feature: Requires company policy review | ChatGPT: No | Purpose-Built HR Platforms: Yes

• Feature: Escalates high-risk questions | ChatGPT: No | Purpose-Built HR Platforms: Yes

• Feature: Provides termination scripts | ChatGPT: Always | Purpose-Built HR Platforms: Never without HR approval

• Feature: Creates audit trail | ChatGPT: No | Purpose-Built HR Platforms: Yes

• Feature: Connects to HR when needed | ChatGPT: Suggests it | Purpose-Built HR Platforms: Requires it

• Feature: Considers organizational context | ChatGPT: No | Purpose-Built HR Platforms: Yes

How Purpose-Built Platforms Handle Termination Questions

I tested the same six scenarios with three HR coaching platforms: Pascal, Leena AI, and Workday Peakon. All three recognized termination questions as requiring HR involvement. None provided termination scripts without organizational context.

The difference is not sophistication. The difference is whether the tool stops answering.

Pascal (built by our team at Pinnacle) uses a three-tier framework that organizations customize based on their size, risk tolerance, and HR capacity:

Tier 1 — Coaching moments: Questions about giving feedback, having difficult conversations, or addressing performance issues early. The AI provides frameworks, suggests talking points, and offers practice scenarios. Example: "How do I tell someone their work isn't meeting expectations?" gets coaching on feedback delivery using the SBI model (Situation-Behavior-Impact: describe the specific situation, explain the observable behavior, clarify the impact on the team or work).

Tier 2 — Consultation moments: Questions about formal performance improvement plans, significant role changes, or compensation decisions. The AI helps managers prepare for HR conversations but does not provide final guidance. Example: "Should I put this employee on a PIP?" triggers a prompt to schedule an HR consultation while helping the manager document the situation. For a performance improvement plan question, the AI asks: How long have performance issues existed? What feedback have you already given? What specific behaviors need to change? The manager's answers become the foundation for the HR conversation. The AI does not decide whether a PIP is appropriate, generate the PIP document, or determine the timeline. Those decisions require HR expertise.

Tier 3 — Escalation moments: Questions about termination, harassment allegations, or discrimination concerns require immediate HR involvement. The system captures the manager's question and context, then connects them to HR without providing guidance. Example: "How do I fire someone?" creates an HR ticket with conversation history and blocks AI responses. The manager sees a message like: "This situation requires HR expertise. I've created a priority consultation request and notified Sarah Chen in HR. She will contact you within 2 hours. Please do not take any action regarding this employee." The conversation history transfers completely to HR.

Organizations configure their own triggers based on size, industry, and risk tolerance. A 50-person startup might handle compensation discussions at Tier 1. A 5,000-person enterprise might escalate compensation discussions to Tier 2, requiring HR review of all salary changes above a threshold. Healthcare organizations might escalate patient safety concerns to Tier 3 immediately. Financial services companies might require Tier 2 consultation for any discussion of employee trading activity.

Tiered Escalation in Practice

Data Breakdown:

• Tier: Tier 1 | Question Type: Coaching moments | AI Response: Provides frameworks and guidance | HR Involvement: None required | Example: "How do I give constructive feedback?"

• Tier: Tier 2 | Question Type: Consultation moments | AI Response: Helps prepare for HR discussion | HR Involvement: Scheduled consultation | Example: "Should I put someone on a PIP?"

• Tier: Tier 3 | Question Type: Escalation moments | AI Response: Blocks response, creates HR ticket | HR Involvement: Immediate and required | Example: "How do I fire someone?"

The Risk of Inconsistent Guidance

Without organizational guardrails, different managers get different advice based purely on how they phrase questions.

Manager A asks: "How do I fire someone for poor performance?" ChatGPT provides an eight-step termination process with talking points.

Manager B asks: "This employee isn't working out—what should I do?" ChatGPT suggests feedback conversations and performance improvement.

Same situation. Different phrasing. Different guidance. Neither response considers company policy, documentation requirements, or legal risk.

This creates inconsistent practices across the organization. One manager terminates after two missed deadlines. Another puts an employee on a six-month improvement plan for the same issue. The inconsistency itself becomes legal liability in discrimination claims.

Employment discrimination cases turn on comparative evidence. An employee alleging discrimination will point to how the company treated similarly situated employees differently. If a 55-year-old employee was terminated after two missed deadlines while a 30-year-old employee received coaching and a performance improvement plan for the same issue, that inconsistency supports an age discrimination claim. The legal standard is not whether discrimination was the only reason for termination. The standard is whether discrimination was a motivating factor.

Documentation of AI interactions becomes critical in litigation. If a company is sued for wrongful termination, the plaintiff's attorney will request all AI conversations related to the termination decision. If those conversations show the manager received a termination script from ChatGPT without any discussion of company policy, documentation requirements, or legal protections, that evidence strengthens the plaintiff's case.

What This Means for Your Organization

If your managers have access to ChatGPT, they have access to termination guidance without organizational context. The question is not whether AI should help managers—it's whether that AI knows when to stop answering and connect them to human expertise.

Three questions to ask about your current approach:

• Do your managers know which questions require HR involvement before they act?

• If a manager asks an AI tool about termination, does that conversation get documented and reviewed?

• Can you demonstrate consistent practices across managers when handling similar performance issues?

Purpose-built HR coaching platforms address these questions through technical safeguards: keyword detection that flags high-risk terms and routes those conversations to HR, organizational integration that provides company-specific policies and approval workflows, and audit trails that document every AI interaction for compliance review.

The goal is not to prevent managers from getting help. The goal is to ensure that help includes the context required to make legally sound decisions.

Want to see how tiered escalation works in practice? Pascal offers a 30-day pilot program for organizations testing AI coaching approaches. Contact our team at Pinnacle to schedule a demonstration.

Header photo by Vitaly Gariev on Unsplash