Form Automation
Designing the interface that will allow AI to perform data-entry for doctors.
Company
IntelliCentrics
Timeline
2 months
(Feb - April 2023)
Role
UX Designer
Cross Functional Partners
Product Management, Development
How can we utilize AI to expedite the process of filling out forms?
Context
IntelliCentrics digitizes and stores verified credential data for doctors, nurses, and medical vendors. The company’s overall goal is to onboard medical professionals to healthcare facilities as quickly and accurately as possible.
Insurance registration is typically handled by hospital administrative staff who manually fill out forms on behalf of doctors. The company sought to automate this process by leveraging existing data and emerging AI technology, eliminating hours of administrative work.
Process
-
When this assignment came to me, it was an ambitious task riddled with ambiguity. Slowly but surely, we uncovered the scope and limitations of the project by asking questions and doing competitive analysis.
-
Whiteboarding, sketching, wireframing potential design options.
-
After receiving feedback from the Product Owner and Development team and working through ambiguity, we iterated until we found the best solution.
-
The product was implemented, staying loyal to the final designs and an interactive Figma prototype. Using the Form Automation system, 700 doctors were enrolled with insurance payers in minutes, 33% faster than competing products.
-
This project taught me the importance of open communication and persisting through daunting obstacles.
Before
On average, a doctor has about 21 insurance contracts, which are typically manually filled out by administrative staff. That translates to around 14,800 new insurance contracts for a mid-sized hospital.
After
Using the form automation tool, AI fills out insurance contracts for an entire roster of doctors, eliminating the tedious task for administrators, and enabling doctors to practice medicine as fast as possible.
Information Gathering
When this assignment came to me, it was an ambitious task riddled with ambiguity. The road to the final product was a group effort to uncover details and requirements.
Ideation
The deadline for a fully implemented minimum viable version of this product was two months. Since this system was unlike anything the company had done before, we needed to hit the ground running. I created several early ideas for this interface, gradually increasing in sophistication.
An Early Sketch
Resourcefulness is key, so I always try to design interfaces using components already written and implemented by the Development Team.
This sketch is modeled after the company’s existing Credentialing interface, a tool used by internal specialists. Those specialists would be trained to use this Form Automation tool, so a familiar interface will help ease the learning curve of new software.
First Wireframe
Operating mainly on assumptions, we anticipated the need for organizing form fields into smaller categories, from Sections to Groups to Fields.
We based this concept on forms wherein each character in a field got its own space, which we dubbed Discrete Text fields.
Depending on the expected content, each field was designated as a Text Field, Option, Date, or Code. Each designation had a corresponding color to clearly distinguish it from other field types.
There wasn’t a definitive answer on how connecting data would occur, so we held off on designing that panel in this wireframe.
Later Wireframe
To avoid clutter and information overload, items in the Form Field panel were collapsed into nesting accordions.
The previous wireframe opened a discussion about the specific information each field needed to capture for the automation to work properly. For an MVP, the metadata for each field is:
an editable Field Name
the Field Type
By this point, the Chief Product Officer had given the team insight into how data points would connect to form fields. This initial design communicated the appropriate information but was difficult to parse quickly.
Iteration
After a month of presenting several rounds of wireframes, uncovering new acceptance criteria, and juggling other stories and tasks, I began incorporating components from the company’s design system to create high-fidelity mock-ups.
High Fidelity Concept
In this iteration, the form fields are color-coded by field type, a concept that was pulled back in from our earliest wireframe. The three colors seen here are:
Green for Free Text Fields
Purple for Date Fields
Grey for fields that have not been assigned a field type
There was no way for the system to verify that the user had correctly identified all fields before completing the form so, the grey color code is a visual way to indicate to the user when a field was missing information.
Below the Form Fields panel, three buttons allow users to add a New Field, New Page, or New Group. The automated system has no use for Pages or Groups, but I thought it was important that the user could organize the items in this panel by grouping fields into smaller sections. The New Field button is the most prominent because it is the most frequently used of the group.
In the Data panel, I updated the Entity selection tool to be a dropdown menu instead of radio buttons to allow for other Entities to be added to this section in the future.
Final Iteration
At this point in the process, we were told that color-coding the fields based on their field type was not a priority. We campaigned to keep the unfinished fields grey as an error prevention measure until the Product Owner saw the value and allowed it to stay in the design.
After playing around with a few options for selecting an Entity in the Data panel, we determined that the selection should be made when the form is uploaded. The Entity, or Persona as we call it in this iteration, affects where the data is pulled and will not change throughout the document. The user can map this form faster and with a decreased chance of error by removing this task from this already tedious process.
We restructured the items in the Data panel into a data tree, allowing the user to easily process the information and follow a clear path to their desired data point.
Final Designs & Results
The final product was implemented and usable by the deadline. Shortly thereafter, 700 doctors would be enrolled with an insurance company using this automated system in minutes, 33% faster than competing products.
Reflection
Communication and Collaboration
While group consensus is always important, it was especially crucial to be on the same page as the Product Owner and Development team for this project. The novelty of the product and the tight deadline meant that we needed to ensure all of our proposed features were important enough to be developed. I attribute the success of this product to consistent informal meetings about concerns, ideas, and questions across teams.
Persistence
As a UX Designer, it is important to be comfortable with ambiguity. This project pushed the limits of my comfort several times. I was worried that it’d be impossible to take this project from a loose concept to a full-fledged product, but time and again, I surprised myself by coming up with effective solutions.
Next Case Study
Medical Vendor App
INTELLICENTRICS, SEPTEMBER 2022 - JANUARY 2023
Improving the usability of the SEC3URE mobile app by refining the navigation, integrating Bluetooth pairing to e-badge, and modernizing the interface.