Employment Law
Why it Matters
In Florida, an employer may fire a worker for a bad reason or for no reason at all, but not for a reason the law forbids. Federal and state statutes name the forbidden reasons, among them a worker’s race, sex, age, or disability; a complaint about unpaid wages; and joint action with coworkers for better conditions. The decisions that come out of those disputes are public records, and reading them is how a business learns what the law actually requires. This activity has you take a real decision from the National Labor Relations Board and use it to test how well three AI models read a legal document.
Current Context
On January 7, 2026, two new members were sworn in and the National Labor Relations Board regained the quorum (opens in a new tab) it needs to decide cases, after nearly a year without one. By July 31, a labor law newsletter counted 380 decisions (opens in a new tab) from the reconstituted Board, most of them routine, such as orders adopting judges’ decisions that no party challenged. On August 17, a fourth member was sworn in (opens in a new tab), giving the Board’s Republican members the three votes it has by custom required to overrule precedent. Closer to home, Florida’s CHOICE Act, in force since July 2025, requires a court to preliminarily enjoin (opens in a new tab) a covered high earner accused of breaking a qualifying noncompete. Only clear and convincing proof of one of three listed grounds can lift the order, so an employee free to quit at any time may still be kept from working for anyone else.
PetPals Unleashed
PetPals Unleashed is a story told one chapter per activity, and each chapter turns on that activity’s law. Read it as the outline of a page you will write: in the Project, you turn all twelve chapters into a book.
Chapter 8
PetPals’ first warehouse manager starts a competing business using customer lists and gets fired. She sues for wrongful termination and gender discrimination. Noodle realizes they never had proper employment agreements or handbooks, leading to a costly settlement and a complete HR overhaul.
Key Concepts
At-Will Employment
The default rule in Florida and most states that either side may end a job at any time, for any reason or none, unless the reason is illegal. The rule gives employers flexibility but does not shield a termination that violates a statute or a contract. Many states also bar terminations that offend public policy, a judge-made exception Florida has never adopted.
Wrongful Termination
A firing for an illegal reason, even in an at-will state. Common grounds include a protected characteristic, such as race, sex, age, or disability; retaliation for conduct a statute protects; and breach of an employment contract, though Florida courts rarely find an implied one. The employer’s usual defense is a legitimate, nondiscriminatory reason for the firing, which is why careful employers write the reason down when they act.
Employee or Independent Contractor
The classification that decides whether a worker is covered by minimum wage, overtime, antidiscrimination law, workers’ compensation, and unemployment insurance. Courts and agencies apply multi-factor tests focused on control and economic dependence, and misclassifying employees as contractors exposes a business to back pay, tax penalties, and agency enforcement.
Noncompete Agreement
A contract term in which an employee agrees not to work for a competitor or start a competing business for a set time and within a set area after leaving. Enforceability varies by state: Florida enforces reasonable restraints that protect a legitimate business interest under section 542.335, Florida Statutes. Since 2025, the CHOICE Act has also required courts to enforce a qualifying noncompete with a high earner by preliminary injunction.
Title VII of the Civil Rights Act of 1964
The federal law that prohibits employment discrimination based on race, color, religion, sex, and national origin in hiring, firing, promotion, pay, and every other term of employment. It also prohibits sexual harassment as a form of sex discrimination and retaliation against employees who complain or participate in an investigation. The Equal Employment Opportunity Commission enforces it.
Fair Labor Standards Act
The federal wage and hour law that sets a minimum wage floor and restricts child labor. It also requires overtime at one and a half times the regular rate for non-exempt employees who work more than forty hours in a week. Because employees can sue together in a collective action, one miscalculated payroll policy can create a liability to every worker who joins the suit.
Americans with Disabilities Act
The federal law that prohibits discrimination against qualified individuals with disabilities and requires employers to provide reasonable accommodations unless doing so would impose an undue hardship. The EEOC’s regulations describe an informal, interactive process for finding one: a conversation between employer and employee about the limitation and what might overcome it.
National Labor Relations Act
The federal law governing labor relations in the private sector. It protects the rights to organize, join a union, and bargain collectively, and it protects concerted activity, meaning even non-union employees who act together over wages or working conditions are shielded from retaliation. The National Labor Relations Board enforces it, and its administrative law judges write the public decisions this activity is built on.
Resources
- Textbook: Chapter 50, Employment Law (opens in a new tab) (The assigned reading on discrimination, wage and hour law, and at-will employment. Its at-will section describes judge-made exceptions that Florida courts have never adopted, and its disability case, Toyota v. Williams, sets out the narrow reading of disability that Congress rejected in the ADA Amendments Act of 2008.)
- Textbook: Chapter 51, Labor-Management Relations (opens in a new tab) (The assigned reading on the NLRA, the NLRB, and unfair labor practices.)
- Employment Law, LawShelf (opens in a new tab) (Ten short free videos, among them employee versus independent contractor, workplace discrimination, retaliation, and five on workers’ compensation.)
- Administrative Law Judge Decisions, NLRB (opens in a new tab) (The database you draw your decision from in Part 2; filter by date and read the full text.)
- National Labor Relations Act, NLRB (opens in a new tab) (The statute the administrative law judge applies; check the sections your decision cites.)
- Concerted Activity, NLRB (opens in a new tab) (The Board’s explanation of protected concerted activity, including when one employee acting alone is protected and how protection can be lost.)
- Title VII of the Civil Rights Act of 1964, EEOC (opens in a new tab) (The text of the main federal antidiscrimination statute.)
- Disability Discrimination and Employment Decisions, EEOC (opens in a new tab) (The EEOC’s guide to the ADA’s employment provisions and reasonable accommodation.)
- ADA.gov, U.S. Department of Justice (opens in a new tab) (The government’s ADA portal, including the statute and regulations.)
- Fair Labor Standards Act, Wage and Hour Division (opens in a new tab) (The Department of Labor’s overview of minimum wage, overtime, and recordkeeping.)
- Misclassification of Employees as Independent Contractors, Wage and Hour Division (opens in a new tab) (The Department’s page on who counts as an employee under the FLSA. It still describes a 2024 rule that the Department no longer applies in its investigations and proposed in February 2026 to replace.)
- Employment-at-will doctrine, Cornell Wex (opens in a new tab) (The default rule and the exceptions states recognize, which vary from state to state.)
- Section 542.335, Florida Statutes: Valid restraints of trade or commerce (opens in a new tab) (Florida’s general noncompete statute and its legitimate business interest requirement.)
- Florida Statutes Chapter 542, Part II, Florida CHOICE Act (opens in a new tab) (Sections 542.41 through 542.45, the 2025 law on covered garden leave and noncompete agreements.)
What to Do
This activity works differently from the earlier ones. Instead of drafting documents or working through fact patterns, you benchmark three AI models, giving each the same real legal decision and judging how well it handles it. Legal and business professionals increasingly use AI tools to analyze long documents, and evaluating that output critically is a professional skill in its own right.
Part 1: Build Your Benchmarks
Work through the first three prompts in order before you run your analysis. Start with the Benchmark Primer to learn what AI benchmark testing is and why it matters. Next, use the Benchmark Designer to generate a set of benchmarks tailored to reading legal decisions. Finally, use the Benchmark Refiner to revise and finalize the benchmarks you will test with.
Part 2: Retrieve the Public Record
Go to the NLRB Administrative Law Judge Decisions page linked in the Resources and download one decision issued within the past twelve months. Choose a decision that involves a fact-specific dispute, such as a discharge, retaliation, or another unfair labor practice. Read enough of the decision yourself to understand the basic facts and legal issues before you begin testing. You cannot evaluate whether an AI is analyzing the decision accurately if you have not read it yourself.
Part 3: Run the Comparative Analysis
Once your benchmarks are finalized, use three different AI models, such as ChatGPT, Claude, Gemini, or Gemini Notebook, to analyze the decision. Ask each model to perform the same tasks so the comparison is fair, and upload the decision directly if the model allows it or paste the relevant portions. As you work through each model, note how it performs against your benchmarks. Look specifically at whether it correctly identifies the legal standard the administrative law judge applied, because that connection to employment law doctrine is what makes this a legal exercise rather than a software test. Then use the Comparative Evaluator to process and reflect on your findings.
Part 4: Post Your Deliverables
Create a single new discussion thread with two sections. The first is a case summary: a concise description of the decision you selected, including the parties, the core legal dispute, the judge’s reasoning, and the outcome, written entirely in your own words. The second covers your benchmarking results: explain your benchmarks, describe the tasks you asked the models to perform, present your ratings, and declare which model performed best with concrete examples.
Part 5: Share Chat Link
Include one AI chat link with a 1–2 sentence explanation of what the conversation shows and why you chose to share it.
Suggested AI Prompts
Use these prompts as a starting point, then adjust them to fit your goal. Strong prompting develops through trial, revision, and testing. It’s a foundational skill that grows into more advanced AI work such as context engineering and agent-based workflows.
Benchmark Primer
Act as an expert in AI model benchmark testing. Explain to a college student what AI benchmark testing is, why it is important, and how it works in plain terms. Describe the main steps involved in testing and comparing different AI models and highlight the most important things to remember when designing fair, clear, and useful benchmarks. Use concrete examples to show how well-designed benchmark tests help people understand which AI model performs a task better, and identify the most common mistakes to avoid when evaluating AI performance.
Method before measurement. Understanding what makes a benchmark fair, such as identical inputs and measurable criteria, keeps your comparison in Part 3 from turning into a matter of taste.
Benchmark Designer
Act as an expert in analyzing National Labor Relations Board administrative law judge decisions and as an expert in designing tests and benchmarks for AI models. Create five clear, practical benchmarks that a college student can use to evaluate how well any AI model analyzes an NLRB administrative law judge decision. Each benchmark should reflect something a real legal reader would care about, be specific and measurable, and include a brief explanation of how a student can use it to compare different AI models. At least one benchmark must test whether the model correctly identifies the legal standard the judge applied, and at least one must test whether every quotation the model attributes to the decision actually appears in it. Make the benchmarks realistic, objective, and suitable for students using tools like ChatGPT, AI Studio, or Gemini Notebook.
Two required benchmarks. A model reading a legal document can misstate the governing standard or invent a quotation, and the prompt builds a check for each into your test.
Benchmark Refiner
Act as an expert in AI model benchmark testing and in analyzing NLRB decisions. Ask me to share my proposed benchmarks for evaluating how well an AI model analyzes an NLRB decision. Once I share them, review each one carefully, provide constructive feedback and suggested improvements where needed, and explain your reasoning for any changes. Then guide me step by step on how to use my refined benchmarks to test and compare different AI models so that my evaluation is clear, fair, and meaningful.
Revision with a rationale. Asking the AI to explain every proposed change makes you decide whether to accept it, which is how the benchmarks become yours rather than the model’s.
Comparative Evaluator
Act as an expert in both AI model evaluation and employment law. I am going to share the results of my benchmark testing across three different AI models based on my analysis of a real NLRB administrative law judge decision. After I share my results, help me evaluate which model performed best overall and on each individual benchmark. Identify any patterns in where the models succeeded or struggled, and push me to think critically about whether any model correctly identified and applied the specific legal standard the judge used in the decision. Ask me follow-up questions that help me connect my AI evaluation findings back to the employment law doctrine the decision involved.
From scores to doctrine. The evaluator’s follow-up questions turn your benchmark results back to the legal standard itself, so the activity ends in employment law, where it began.