AI 101: Large Language Models and the ABCs of Workers’ Comp 

The Trained A-Eye

Sharpen your pencils, grab your notebooks, and settle in because class is officially back in session! And while Drake University started back last week, CLM Claims College is in full swing this week, we have a wonderful new series to share with you to further your education this fall with your favorite professor, Dr. Claire, that’s me! 

Welcome to Back to School with AI, a series designed to explore what artificial intelligence really means for workers’ compensation. Over the next twelve articles, we will tackle everything from the basics of prompting to the ethics of machine decision-making, all through the lens of the industry we live and breathe every day.  

This project is inspired by the brilliant AI Summer School series created by two of my professor friends at Drake University: Chris Snider and Christopher Porter. These two gentlemen are known as the Innovation Profs and if you are not following them, you should. I learn so much from their weekly newsletters to keep me on the up and up, so I am going to share with you. This summer, THE Innovation Profs broke down the essentials of generative AI for a broad audience, and I could not resist taking their spark and bringing it home to workers’ compensation. Why? Because let’s be honest: our industry does not get a reputation for being at the cutting edge of technology. Still using those green screens? You get it. If there has ever been a moment to pay attention, this is it. 

Today, we are starting with the foundation: Large Language Models, or LLMs. Think of this as your “AI 101” and you are at the first day of class. The first day of class is for new markers!! And also, it is where we set the tone, meet the teacher’s pet (ChatGPT, obviously), and talk about how this shiny new tool might help us in claims, compliance, and communication. It may make us better people too… who knew? 

What Exactly Is a Large Language Model? 

At its core, a large language model is a type of artificial intelligence trained on enormous amounts of text. It learns patterns in language, which words are most likely to come after other words. LLMs learn how words and ideas connect. As a result, the LLM can generate new sentences, paragraphs, or even whole reports that sound remarkably human. 

The “large” part simply refers to the massive scale of data and computing power behind it. It reminds me of how insurance is based upon the law of large numbers. Imagine giving a student access to every textbook in the library and letting them practice writing essays a billion times over. By the time finals roll around, that student would be scarily good at predicting what a strong essay should look like. That is what LLMs do with language. They predict what the next words should be based upon patterns that have come before. 

The result? Tools like OpenAI’s ChatGPT, Microsoft Copilot, Google Gemini, and Anthropic's Claude. Chat GPT is the most popular LLM, used for everything from email drafts to complex analysis. Chat GPT went from 0 to 100 million users in two months! To give you a good basis for this technological advancement, the cell phone took 16 years to accomplish this user base. Microsoft Copilot you have probably seen as it is built right into Word, Outlook, and excel which helps workers without leaving their daily tools. Google Gemini is Google’s response which is integrated with Google Docs, Gmail, and Sheets. Claude is a quieter but powerful player in the LLM space and has been praised for its clarity and thoughtful responses. Others do exist such as Meta’s LLaMA, xAI’s Grok, and Mitrial; however the big four are the platforms you have more than likely heard of or are seeing in your day-today work right now.  

Why Should Workers’ Compensation Care? 

Now, I know what you are thinking: “Cool tech, but what does this have to do with my claims desk or return-to-work plan?” Great question! The reality is that LLMs are already reshaping how we work. Please note: LLMs are not here to replace adjusters, nurses, or human resource professionals. LLMs are here to help this space work smarter, not harder by making the administrative, repetitive, or time-consuming parts of our jobs faster and easier… and as a result, we can focus more on people. And PEOPLE are at the heart of what we do in workers’ compensation.  

Think of LLMs as the ultimate teaching assistant in your work comp “classroom.” These LLMs do not take the test for you, but they help prep the study guides, grade the practice quizzes, and free up your time to teach. This makes you more effective and efficient by providing you the tools and resources to do more with the time you have. Intentionality is key! If you’ve ever wished for a clone who could handle your paperwork while you focused on the human side of this job, LLMs are as close as it gets. Here are some ways LLMs can show up to help YOU.  

Claims Communications. Drafting clear, empathetic letters to injured workers instead of reusing cold, legal-heavy templates. LLMs can help you make this information much more consumable for your audience. Statues are hard to understand, break it down! And if you are not sure how, ahh, there is an LLM who can help you phrase it so a fifth grader could understand. (Ideal communication should be between sixth to eighth, in case you are wondering on consumablility!) 

Training & Education. Creating supervisor handouts, safety reminders, or FAQ sheets tailored to specific industries. I don’t know about you, but I am not a spur of the moment creative, and this is where LLMs can help you be more inspiring through visual communication to help emphasize educational points or reminders.  

Fraud Detection Support. Spotting unusual language patterns across reports that might flag inconsistencies. Words and verbiage choice matter. People are trained to detect language nuances and if you are not one of those trained professionals, LLMs can help you along the way, assisting you to work smarter, not harder.  

Policy & Procedure Writing. Turning complex regulation updates into plain-language explanations for teams. Sometimes we think what we are communicating makes sense, and it does, to us, the drafter of the communication. LLMs can help with how written communication can be received. This is like having a personal editor always within reach. 

Working with People. People are hard. We are multi-faceted, multi-functional, multi-dimensional, multi-emotional beings. Two things can be true for any one of us at a given time. With this complexity, sometimes we have no idea how to handle a situation and LLMs can offer suggestions on what to say next, how to approach a situation, or when to simply be quiet and say nothing. (Yes, I have had it tell me this before. If you don’t know what to say in a situation, maybe nothing is the answer for this moment. And. My LLM was right.) 

Picking Your Platform: The Big Four 

Like choosing a calculator for math class, the LLM you use depends on what you need. The truth? You do not have to master them all. Start with one that integrates with your existing tools. The learning curve is less steep than you think, and most people are amazed by how quickly they get comfortable once they start experimenting! ChatGPT is flexible, intuitive, and great for brainstorming, drafting, and role-play exercises. (No one likes doing those, so I prefer to call those simulations.) Microsoft Copilot is basically a tutor sitting right in the margin of your spreadsheets and works well for people already living in Excel or Outlook. If you are old enough to remember, Clippy was present WAY back in the day with Word where the avatar could help you start a new document or share a resume template. Google Gemini seamlessly fits into organizations running on Gmail or utilizing Google Docs. Claude stands out for summarizing long, complex documents with accuracy. Which one to use? Try one and see how you feel.  

How Do LLMs Work? (Without the Boring Math) 

We could dive into “transformers” and “neural nets,” but this is workers’ comp school, not computer science class and therefore, I will leave this with the Innovation Profs for further explanation there. What you need to know is that LLMs don’t “think” the way humans do. They predict the next likely word or phrase based on patterns. As predictors, LLMs do not know truth from fiction as they function around knowing probability. (This is why fact-checking their work is critical!) The more you interact with a LLM, the better they get. Prompting, the way you ask a question, shapes the quality of the answer you will get from your LLM. Think of it like teaching a brand-new adjuster. If you say, “Handle this claim,” they will stumble until they figure it out. If you give step-by-step guidance such as “Review medical notes, summarize treatment, draft a letter”, the new adjuster has a better chance to succeed. LLMs need that same clarity. 

The Human Factor 

The biggest misconception about AI in workers’ comp is fear of replacement. Adjusters, HR professionals, risk managers, providers, safety professionals ... we are not being erased! Please note that what is being erased is the busywork. LLMs can pull data and draft text, but they cannot sit across from an injured worker, hear the tremble in their voice, and know when to soften the conversation. They cannot walk a supervisor through a tough return-to-work decision with empathy. They cannot see the bigger cultural context of a workplace. LLMs build from language patters. We grow through experiences with people. That is where we shine. AI doesn’t eliminate the need for the human element: it highlights it, improves it, and shows us we CAN be better… we simply need to make the CHOICE to do so. 

Class Takeaway 

Large language models are the ABCs of the AI world. LLMs free us to do what matters most: care for people, support recovery, and connect with empathy, not about making workers’ comp colder or more robotic. As we head into this school year together, I challenge you to think: Where could an LLM give me back 10 minutes today? An hour this week? A day this month? That’s time you can spend where it matters, with your people. For me? Chat GPT and I are going to be in the kitchen as I finally learn to cook… maybe. Or it can help me organize my schedule with time blocking so I can get that down to a science! That is all for now.  

Class dismissed. ✨ 

Next week? Pack your art supplies, because we are heading into Show & Tell: Multimedia Tools for Training & Injured Worker Support. 

Share This Article: