FEHB Plan Renewal Helper

Health insurance plans are confusing. They are packed with unfamiliar terms, long explanations, and important financial details that can be easy to miss. The U.S Office of Personnel Management says that in 2026, the FEHB, or the Federal Employees Health Benefits Program provides coverage to 8.2 million Americans. The problem is that millions of these people don’t understand their healthcare plan. In the 2023 KFF Consumer Experiences with Health Insurance Survey, KFF found that “Half of insured adults struggle to understand at least one part of their insurance.” This gap between how important healthcare is and how many people even understand what it does for them is what inspired us.

Our product is targeted to those federal employees, retirees, and families who use the FEHB program, but do not fully understand what their insurance plan is. We aim to make sure those people will never have to pick a plan and spend thousands of dollars a year while not even fully understanding what the plan can do for them, and we aim to make sure that the plan they pick truly is the best for them. Instead of forcing users to read through long documents alone, our app helps organize plan information, estimate expected costs, and explain why one plan may fit a user’s needs better than another.

We built our project through an iterative process. First, we outlined the problem, the user flow, and found actual plan details for UnitedHealthcare’s Choice Primary plan on their website. We then made a list of everything we wanted our app to do in order to target plans like United Healthcare’s. Then we used tools like Claude Code and ChatGPT in order to build the website. After every new iteration, we would go through and jot down places that did not work, what we wanted to change, and anything that users would find confusing. We also made sure to periodically check in with the rubric and questions to consider in order to make sure our project was not just functional, but also relevant, responsible, and clearly connected to the problem we were trying to solve. After roughly fifteen iterations, we reached a version that felt much closer to what a real user would need.

Throughout this process we grew. First of all, we learned what all the different healthcare terms meant like copay and coinsurance. We also had to figure out all the different rules like whether or not the deductible applied before or after the copay to make sure that our app accounted for all of these when calculating a price. It was as if we got a glimpse of what the average adult has to do in order to find the best healthcare for them and their family, and that was when our hearts truly started empathizing with the issue.

One of our biggest challenges was scope. At first, we wanted to make our app help with all different kinds of healthcare like Marketplace and ACA, but then we realized that we were overambitious. Marketplace costs can vary based on age, income, subsidies, location, household size, and other factors, which would make accurate cost estimates much harder. Instead of building a larger but weaker tool, we narrowed our focus to FEHB plans. This made the project more realistic because FEHB plans are used by millions of people and do not depend on some of the same pricing variables, such as age and income.

Another major challenge was using AI responsibly. We did not want the app to simply tell users, “This is the best plan,” and expect them to trust it blindly. Health insurance decisions are too important for that. To reduce this risk, we designed the app so that calculations are kept separate from AI explanations. The structured table handles the cost comparison, while the AI helps explain the recommendation in clearer language. We also require users to review important information, such as costs, network rules, and prescription assumptions, before relying on the final result.

Accomplishments that we are proud of include turning complicated insurance documents into plain-language explanations, building a recommendation system that users can actually understand, creating responsible AI safeguards instead of treating AI as a black box, and helping users leave with clear next steps, not just numbers. We learned that trust matters just as much as accuracy. Building with AI isn't just about generating answers—it's about helping people understand where those answers come from. We also gained experience with AI integration, prompt engineering, and designing systems that keep humans in control of important decisions. Next, we'd like to add real-time prescription drug and provider coverage checks, more advanced "what-if" scenarios, and additional plan comparison tools. Our goal is to make health insurance easier to understand so people can make confident decisions without feeling overwhelmed by paperwork.