Designing a generative AI app trained on healthcare knowledge.
Designing a generative AI app trained on healthcare knowledge.
Client: AVIA
Role: Product Design Lead
Skills: Product Design, UX Research, Information Architecture, Visual Design, Team Leadership, DesignOps
Team: Product Design Lead (me), Product Designer, UX Researcher, Product Manager
Timeframe: 3 months
The Problem
Health professionals need quick research for digital strategies to improve patient care and reduce costs. They struggle to keep up with digital changes while staying competitive in healthcare.
The Solution
As Product Design Director, I led a 3-month project to create an AI Chatbot powered by ChatGPT and AVIA’s proprietary content. The chatbot provides easy access to digital health strategies and product information, helping healthcare professionals make better decisions.
Understanding the User
To ensure our AI Chatbot met user needs, I conducted interviews with 15 health system employees as part of a “Voice of the Customer” project. These conversations provided invaluable insights into their organization’s tech selection approach, guiding our design process to better serve their needs.
We shared an outline of the tech selection process and asked them to compare it with how things work in their organization.
Exploration
In our exploration phase, we dived into competitor research while also creating a Human-Centered AI canvas. It wasn’t just about understanding the competition; we aimed to ground our approach in a deep understanding of human needs, and the canvas played a huge role in guiding our journey.
How might we leverage AI to simplify research for healthcare professionals, improving digital strategies, patient experiences, and operational efficiency?
Competitor Research
We conducted competitor research, thoroughly examining platforms such as ChatGPT, Gemini, Chatsonic, and Claude to gain valuable insights and inform our approach to creating a distinctive AI Chatbot.
Human-Centered AI canvas
With our Human-Centered AI canvas, we made brainstorming a team effort, blending insights from real conversations with health system professionals. It guided us to create an AI Chatbot that truly understands and meets their needs.
Ideation
Next, I led our team in an ideation session, where we got creative and used affinity mapping to organize diverse ideas. This collaborative process laid the groundwork for our AI Chatbot design.ion
Wireframing
Beginning with basic Figma wireframes, we tested our ideas using a low-fidelity prototype. These initial outlines served as our starting point, guiding us through the testing and refining phases of the design process.
Prototyping and testing
I teamed up closely with our engineers to create a quick beta version, tested with our product advisory group. Users loved how helpful and efficient it was for research, even preferring it over consulting with professionals for quicker answers. This positive response showed our collaboration hit the mark in meeting user needs.
The Final Design
Listening to our beta users, we fine-tuned designs to better address their organizational needs, simplify pre-defined prompts, and improve the user experience of saving and deleting prompts. Working closely with our engineering team, we made the AI Chatbot even more user-friendly.
We created a final design that met health system user needs and testing feedback, with a clean, modern look. It used familiar AI design patterns, letting users quickly start using it without much effort. We even gave it a cute name, “Hugo”.
Quick links for common questions to streamline user experience.
Thumbs up/down feedback option for quick and easy response evaluation.
Easy access to previous chats with the option to search your history.
Results
Our journey through beta testing has yielded encouraging outcomes, marked by positive feedback and strong interest from potential subscribers.
Positive feedback
from our second round of beta testers
Intent to purchase
verbally confirmed from several testers