Using AI to Increase Efficiency for Physicians

Case Study
Brand Design
Button Text
Starting from a small project in a hackathon, we decided to bring our product to it's full potential by building a start-up from it. Check out how I design a product while utilizing large language models and natural language processing to boost productivity in private practices
Initial prototype placed top 50 in a hackathon
Invited by Skydeck to pursue a full time startup
Working with multiple pilot customers

⛰ The Challenge

Private practice physicians easily have their hands full due to healthcare inefficiencies such as insurance billing and administrative tasks on top of their usual responsibilities.
‍
As a past chiropractic assistant, I handled patient care, administration, and insurance billing and spotted a big inefficiency that affected our business. Varying policies and led to billing errors and delays, affecting practice finances and patient satisfaction, creating more work for physicians.

🧗🏻‍♀️ My Role

I'm at the forefront as the lead designer, guiding us through interviews, testing and the developing wireframes ad prototypes. I also manage ongoing projects while ensuring smooth communication and understanding with engineers and pilot customers.

I also bring warmth and creativity to our brand identity and marketing efforts with a business mindset.


Final Product

As the frontend is currently being built out through the electron app, the final product (for now) is an integrative widget. Read more to see my process!

How it Works

How I Did It

Testing my Hypothesis

Though I discovered the problem while working for a private practice, I needed to prove that others had similar pain points as well.

📝 Medical scribes and taking notes added cost or time for the physician.

“My medical scribe takes too long to finish writing and documenting, which is why I just end up doing it myself. Yes, it is more administrative work but less money spent.”   -Interviewee
To solve this, I put myself in the user's shoes...

Speech to Scribe Transcription

Thinking from the user's perspective:

Physicians don't have their hands free, but are still able to converse.

💸 Insurance billing not only is tedious but also is prone to error.

“I always need to search up ICD code because there’s too many. It takes so much time because it has to be accurate.” -Interviewee
To solve this, I looked at the resources and tools available to me...

Suggestive Billing Codes

I assessed tools and skillsets available to me, which were engineers that could create large language models (LLMs).

💊 Slow medication search up.

“As a primary doctor, there are so many prescriptions that I search for but the system takes so long to load.”
To solve this, I took a step back and identified common problem themes...

Bringing it back to user needs..

I reassessed the user needs and put them in a general perspective. When it comes to multiple needs, I zoom out and try to find a general problem pattern or ways they can be organized.

The First Prototype

This was the first prototype that we built which got us to the top 50 out of 700 projects at CalHack's AI Hackathon in August of 2023.

Due to a lack of time, this was built out on Flutterflow for easy frontend and back end integration.
View Prototype

Learning Space

As we started to fully develop this product, our original team of 3 engineers was down to 1 due to their own personal passion projects.
Because of the lack of developers, I had to redesign a product that was suitable with the Streamlit app but also suitable for the users, the physicians.
Now, the single screen platform is structured based off information that the physician would use most to least from bottom to top.
View Prototype

Business Retrospection

Of course, the platforms above is not the best possible solution for customers. The first product was too much like an EHR system which made me wonder why private practices would be willing to transfer all their patient data into our platform just for a couple features--and it definitely would not be for the minute use of the AI trend.

While our company looked for more engineers to build out our MVP, I started to conduct more interviews and brainstorm the best possible solution through the user needs.
Secondary interviews further proved that a new EHR system is not what customers are looking for.

Final Product

After long sessions of further research and brainstorming, I created a product that would be similar to a Chrome plug-in. Our platform is small but mighty, where it is responsive and can be minimized while viewing integrated EHR systems.

Our system will read and analyze the EHR system for patient records but will not store the data, ensuring HIPAA compliance. Instead, the selected diagnosis, billing codes and generated patient notes will be sent to an email or phone number for the user's records.
The now integrated plugin organizes dense information into cards and sections to promote a smoother user experience.
View Prototype