LLM for assisting users to track costs and save lives – Optum

Product Design

The need for quality care in clinics and hospitals while at the same time monitoring costs and maximizing facility use can be challenging for a busy clinical staff to handle. 

A dashboard can only provide data but not the kind of information needed to make decisions or put initiatives into action. 

At the time of this product design, Chat GPT- 4 was newly available to the public. My client was looking for new ways to provide highly actionable information that wasn’t easily gleaned from the data or business intelligence provided through standard interactivity of dashboards alone. 

As part of my proposed designs, I incorporated a generative pre-trained transformer (GPT) into the workflow and found that it was the very thing that was needed to meet the request and requirements for this product. 

I fed example-scenarios as prompts into Chat GPT-4 to create the “Drivers” for the mockup to illustrate the value of having specific information for making critical decisions and take action by “Creating Initiatives”. 

In this design, the prompts would be provided by the business intelligence and not the user in order to streamline the process. This could be later enhanced with a prompting tool in later iterations to allow the user to enquirer further and refine the drivers to a specific case. 

Through the work flow of these wireframes I was able to demonstrate how a Large Language Model could increase efficiency and improve the health of patients in hospitals and clinics while also saving money and reducing the workload on a busy staff. 

The Opportunities and Drivers screen displays the BI in the Key Indicator section, however the real value to the user is in the Opportunities section where Drivers can be seen in accordion panels which when rolled up, reduces noise for easy scanning. 

For additional assistance the number of Drivers is indicated by the numeral at the right of each panel for quicker decision making, adding to the richness of the information on the screen.
When the user chooses an Initiative, a drawer panel slides from the right side for the user to refine the Initiative and direct it to specific personnel with specific dates and actionable items. This provides a record of the Initiative which feeds back into the BI for a better analysis of the clinical operations. 
The “Initiatives” screen displays the initiatives created through the “Opportunities and Drivers” interaction. The user can keep track of the progress of the Initiatives, edit them and generate a PDF for sharing the Initiative or for hard-copy printed records.