Claude asked me if I wanted to create an artifact. I had no idea what that meant.
What was an artifact? What are we building here? A mini time capsule? A futuristic résumé? A portal to another dimension?
Nope. We’re building interactive tools. Custom, shareable, AI-powered tools that can actually help people make decisions.
Here’s the official-ish definition:
Claude artifacts are standalone apps, tools, or documents created within a Claude project using your instructions, data, and context. Think of it like building a tiny, purpose-built GPT that can do something specific—based entirely on what you give it.
And hive, once I understood that? Game on.
How I built an Open Enrollment Tool in Claude
I started by uploading:
Our company’s health care Summary Plan Descriptions (SPDs)
A 70+ page pharmacy drug list (that the SPDs reference)
2025 employee premium costs
Then I gave Claude my goal:
“I want a tool that helps employees choose between health plans. Use the drug list and SBCs to compare costs.”
And it made a price comparison tool.
From there, I just kept going:
Added brand fonts and colors
Explained pre-authorization rules
Asked it to read and code the entire drug list
Added a toggle so people could find drug alternatives [using AI]
Fixed the math when it got deductible rules wrong
Asked Claude to optimize and shorten the code (it did!)
Built in mental health care scenarios and examples
Simulated everything from minimalist care to chronic conditions
Asked it to suggest HSA vs FSA savings based on inputs
And now? We have a custom-built, company-specific tool to help employees navigate open enrollment.
Want to try this yourself? Here’s how:
Open Claude and start a new Project
Upload relevant documents or datasets (pdfs, docs, etc)
Use a clear prompt that explains your goal—what do you want this tool to do?
Hit “Create Artifact” when Claude offers (you can’t miss it)
Keep building: give feedback, fix errors, test, iterate
When it works? Share it.
Oh! Here’s the kicker.
I have zero coding skills. None.
And hands-on? This took about five hours.
The trick wasn’t in the tech. It was in the thinking. What info to feed it. What decisions people actually need help with. What problems we could solve if the right data was structured the right way.
That’s when it hit me: This is the HR of the future.
I really want to chat Claude artifacts with peers- anyone trying anything cool at there? Let me know. Please.