
Overview:
This AI-powered assistant helps healthcare support agents navigate the complex steps of processing prior authorizations (PAs) for specialty medications. It simulates real-world job support inside a fictional healthcare tech company—CareAccess Health.
​
The Problem
​
Processing prior authorizations requires detailed, step-based workflows involving insurance requirements, documentation, and provider communication. New support agents are often overwhelmed and risk making mistakes that delay patient treatment.
​
The Solution
The CareAccess Health GPT Assistant functions as a digital mentor, walking users through real-time PA scenarios using clear, job-aligned guidance. It:
Provides step-by-step support based on medication and payer requirements
Helps agents escalate appropriately (e.g., involving case managers)
Simulates just-in-time job aid behavior without replacing formal training
​
This solution showcases the power of AI-driven learning and support tools in real-world healthcare operations.
​
Design Process
​
I used OpenAI's CustomGPT builder to create a structured, rules-based assistant:
Wrote detailed GPT logic using role, rules, and workflow prompts
Simulated realistic job scenarios using fictional data and tone
Tested scenarios using user-like questions such as:
“What documents are required for a Humira prior auth?”
“What do I do if the provider forgot to submit labs?”
Tuned language to be clear, supportive, and accurate
​
Impact
​
This project demonstrates how CustomGPTs can accelerate onboarding and support for complex roles in healthcare. It shows how AI can be:
A safe practice space
A job aid
A confidence booster
It's also an effective sample of performance support design using AI.
​
Tools Used:
​​
-
OpenAI CustomGPT Builder – for creating AI-powered learning and performance support tools
-
ChatGPT (GPT-4o) – for prompt engineering, scenario design, and simulation testing​
​
Try It
​
​
This is an interactive training module in Articulate Storyline, integrating custom JavaScript with OpenAI’s GPT API to provide AI-generated feedback based on learner inputs. The module simulated real-world tasks, requiring users to extract key details from AI-generated call transcripts and enter responses, which were then analyzed by GPT to deliver curated, context-aware feedback. This use of AI-driven evaluation ensured more dynamic and adaptive learning, enhancing accuracy and engagement. The project demonstrated the strategic application of AI and coding in instructional design, creating a scalable and immersive training experience.​​​