Why You Shouldn't Use LLMs to Draft Employee Termination Letters (And the Ethical Way to Handle Redundancies)

The Temptation of the Clean Slate

There is no getting around it: letting an employee go is miserable. Whether it is a performance-related termination or a company-wide redundancy, the process is a cocktail of awkward silences, legal tightropes, and genuine human distress. It is the kind of task that sits on a manager’s to-do list like a lead weight.

So when you are staring at a blank document, trying to find a way to tell someone their livelihood is changing, the temptation to open up Claude or GPT-4o and type, "Write a compassionate but firm redundancy notification letter for a software engineer" is incredibly strong. It promises to take the emotional edge off. It offers a clean, professional, friction-free draft in three seconds flat.

But using an LLM to write a termination letter is a profound ethical misstep. It is a failure of leadership, a regulatory hazard, and a betrayal of the basic human relationship between employer and employee.

The Synthetic Empathy Trap

LLMs are master mimics. They can generate sentences that sound apologetic, supportive, and solemn. But there is a distinct, uncanny valley of machine-generated sympathy. When an AI drafts a layoff letter, it relies on a statistical average of corporate euphemisms: "we appreciate your dedication," "navigating changing market conditions," "wishing you the absolute best in your future endeavours."

To the person receiving the letter, this reads as cold, clinical, and deeply insincere. They can spot the algorithmic veneer from a mile away. When someone is losing their job, they deserve raw, honest human communication—not a calculated string of tokens designed to sound like a human who cares.

Using an AI to bypass the emotional discomfort of writing these words means you are prioritising your own comfort over the dignity of the person being dismissed. Understanding what makes us tick as leaders means leaning into that discomfort, not outsourcing it to a server farm in Oregon. If you cannot find the words to tell someone why they are being let go, you need to step back and examine your own management process, not your prompt engineering.

The Illusion of Legal Compliance

Beyond the moral argument, there is a massive operational risk. Employment law is a highly regional, rapidly shifting minefield. What is legally compliant in California will get you dragged before an employment tribunal in the United Kingdom or Germany.

LLMs do not understand the law; they predict the next likely word. When you ask an LLM to draft a termination letter, it will happily hallucinate legal justifications, notice periods, or severance terms that may violate local labour standards. For example, it might draft a "termination for cause" letter that fails to meet the strict statutory definitions required in your jurisdiction, exposing your business to a costly wrongful dismissal lawsuit.

If you find yourself trying to fix a poorly drafted prompt that hallucinated legal terms, you will end up spending more time in the OpenAI troubleshooting hub than you would have spent writing a straightforward, legally vetted document from scratch.

The Privacy Leak You Didn't See Coming

To write an accurate termination letter, you have to feed the model context. This often includes the employee’s name, job title, performance history, specific reasons for termination, and salary details.

Feeding personally identifiable information (PII) and highly sensitive HR data into commercial LLMs is a massive breach of trust, and potentially a violation of data protection laws like GDPR. Unless you are running a completely sandboxed, zero-data-retention enterprise API, you have no business pasting an employee's performance struggles into an external prompt. Your employees have a right to have their career setbacks kept between them and human resources—not digested into a model's training weight.

How to Ethically Draft Offboarding Communications

If you shouldn't use LLMs to write the letter, how should you handle the documentation? Ethical, professional offboarding requires a clean separation of human empathy and structured legal templates.

1. Use Vetted, Human-Written Templates as Your Base Do not start with a blank page, and do not start with a prompt. Start with a standard, lawyer-approved template designed specifically for your jurisdiction. This ensures that notice periods, benefit continuations, and legal protections are perfectly accurate.

2. Handwrite the Contextual Elements Every termination letter needs a brief explanation of the context (e.g., departmental restructuring, budget constraints, or documented performance milestones). Write this section yourself. Keep it brief, factual, and free of corporate jargon. If you cannot explain the decision in two simple, human sentences, you are not ready to have the conversation.

3. Deliver the News Verbally First The written letter should only ever be a confirmation of a conversation that has already happened. Never let a document—and certainly never let an AI-generated document—be the first way an employee hears that they are losing their job.

Save the LLM for your marketing copy, your database migrations, or your code refactoring. When it comes to the heavy, human moments of leadership, do the work yourself. Your team deserves nothing less.