Science Lab: Custom GPTs for Workers’ Comp Programs

The Trained A-Eye

Welcome back, classmates! Last week we spent our time in the library, learning how to organize prompts into a prompt library using the SCORE framework. Libraries make our lives easier because they keep resources consistent, efficient, and shareable. This is great and also…what if you find yourself using the same cluster of prompts over and over again, each time adding the same background information? It’s time to get more efficient. Welcome to science lab, where we experiment, innovate, and push the boundaries. Today our experiment: building custom GPTs tailored for workers’ compensation.

This project is inspired by my professor friends Chris Snider and Christopher Porter, the Innovation Profs, who created the AI Summer School series. Their work on prompt libraries and custom GPTs showed me how powerful these tools could be, and I adopted it for the world of workers’ compensation. Our industry is built on repetition and consistency including claims letters, policy explanations, RTW instructions. Custom GPTs allow us to automate the background noise so we can focus on the human impact. 

What Is a Custom GPT?

A custom GPT is a personalized version of ChatGPT that is tailored for a specific purpose. You package background information, roles, examples, and formatting into the GPT itself, so you do not have to repeat them every time. Imagine it as a lab experiment where you pre-mix all the chemicals so you do not have to measure ingredients for every test. You are coming to the table with the formula bottled and ready. For workers’ comp professionals, this means having a GPT that already understands your tone, jurisdiction, and communication style, making your daily work much smoother and faster. If you service multiple jurisdictions, you can make multiple custom GPTs.

Why Custom GPTs Matter in Workers’ Comp

Workers’ comp thrives on repeatable processes that must still be carried out with care. Enter the human elements we so direly need. Adjusters draft dozens of similar letters each week, HR managers explain modified duty policies over and over, and nurse case managers repeat treatment guidelines across multiple cases. A custom GPT bakes the background into the system, so each request starts from a place of shared understanding and reduces the chance of inconsistency. This can remove friction that drains energy and wastes time. When you streamline repetitive communication, you reclaim hours that can be redirected into empathy, listening, and problem-solving. 

Custom GPTs raise the standard for consistency across organizations. Instead of telling AI over and over, “Act as a claims adjuster writing to an injured worker, keep it empathetic, use plain language, aim for 150 words,” you build a custom GPT that takes those assumptions for granted. Now you can simply say: “Write a modified duty explanation for a warehouse worker with a sprained ankle.” The GPT delivers in the tone, format, and length you want, without you needing to spell it out. This shift ensures that every worker receives communication aligned with organizational values, not just the adjuster’s individual style that day. Over time, this builds stronger trust with injured workers and demonstrates professionalism across the board. Consistency and repeatability are key. Custom GPTs can help.

Building Blocks: From BRIEF to Custom GPTs

Remember the BRIEF framework? Background, Role, Instructions, Examples, Format. A custom GPT is like bottling BRIEF into a single reusable container so you do not have to rewrite the same details. The GPT “remembers” the Background, Role, Examples, and Format under the hood, leaving you free to only provide the Instructions each time. Think about a claims diary note. Instead of writing, “Background: Manufacturing company in Iowa. Role: Claims examiner. Examples: Use empathetic language. Format: Short paragraph for diary note,” every single time, your custom GPT already knows all of that. Remember to be mindful about what information you are putting into your custom GPT. You can do all of this without entering personal protective information. Work smarter… not harder. Custom GPTs essentially automate the setup, so you can jump straight into the work that matters.

Uploading Documents: Adding Context at Scale

One of the most powerful features of custom GPTs is the ability to upload documents. Imagine feeding your claims manual, return-to-work policy, or state-specific statutory language into the GPT. Instead of flipping through PDFs or searching intranet pages, you can ask, “What are the restrictions for temporary total disability in Iowa?” and get an answer based on your uploaded resources. This is a major time-saver, but it also reduces the risk of misinterpretation when staff are unfamiliar with a particular state’s rules. With accurate references built right into the GPT, new team members get up to speed faster and seasoned professionals stay consistent.

A Workers’ Comp Example of a Custom GPT

Want to build a Return-to-Work Assistant GPT? I know I do. What you can do is upload your company’s RTW policy and feed it sample letters you have used in the past. (Make sure to react any personal information.) You configure the GPT so that it always writes in plain, supportive language, frames modified duty as a healing support, produces letters under 200 words, and suggests talking points for supervisors. When someone types: “Draft a modified duty offer for an injured worker with lifting restrictions after a back sprain,” the GPT already knows the policy, the tone, and the structure. The result is consistent and empathetic communication, without requiring to rewrite the same background every time.

How to Build One (Without the Scary Tech Talk)

The good news is that building a custom GPT is easier than it sounds. OpenAI provides a GPT Builder that walks you through the process in plain language, like a lab partner explaining each step. You answer a series of questions: what should the GPT know, how should it behave, and what resources should it draw from? You can test your GPT as you build it, refining until it works the way you want. Paid ChatGPT users can create GPTs, but once built, they can be shared with free users as well. This means your investment benefits your entire team, regardless of who has a subscription.

Other platforms are developing similar tools, which shows how valuable this concept has become. Microsoft Copilot offers “Agents” for paid users, and Google Gemini provides “Gems,” which can be built even by unpaid users. These features allow professionals across industries to build assistants tailored to their needs. For workers’ compensation, we can experiment across platforms and identify which tool integrates best with our systems. The landscape is evolving quickly, and those who test now will be best positioned to lead later.

Class Takeaway

Custom GPTs are the next step in our AI journey. They evolve naturally from the prompt libraries we explored last week, simplifying repeated tasks, embedding context, and creating consistency across communication. In workers’ compensation, they are cultural tool that helps us care for people while protecting process. The real magic lies in how these tools save time while elevating the quality of human connection in claims management. By leaning into innovation, we shape the future of this industry with intention and heart.

Here’s your homework: think of one area of your work where you repeat the same prompts weekly. Is it claims letters? Policy summaries? Supervisor coaching scripts? Imagine how much time you could reclaim if you had a custom GPT already programmed with your tone, format, and examples. Then, take a bold step: build one. Even if it might be small, the act of experimenting is the first step toward transforming how we work. The lab is open, the tools are here, and the future of workers’ compensation is waiting for us to step in.

Class dismissed.

Next week: Math Class – Reasoning Models for Cost Prediction.

Share This Article: