I am a UX designer and
researcher who is passionate
about using creativity and
empathy to empower people.
I am a UX designer and
researcher who is passionate
about using creativity and
empathy to empower people.
I worked in the Data Science and Analytics Lab (DSAL) of American Family Insurance as a UX designer. DSAL mainly works on using machine learning and artificial intelligence to build tools for people to get information about properties.
During my employment, I researched and developed AI-driven products and built both web and mobile user interface for emerging machine learning technologies.
Figma
Miro
Google Slide
Web Design
Mobile Design
UX Writing
UX Research
7 Months
While employed at American Family Insurance (AmFam), my primary focus revolved around two DSAL projects - Propview and Canary. Additionally, I also took on some side projects related to these initiatives.
PropView is an internal website currently used by employees of AmFam enterprise. It is a “one-stop-shop” for property photos, information, map view, and ai-enhanced predictions about the property or its attributes.
Currently PropView is used by a handful of Claims employees who piloted its use back in February-June 2021, it’s also occasionally been used by AmFam Insurance Agents and AmFam underwriters.
Canary is an AI project with two parts: Canary Advise and Canary Refer, and they are tools that help people understand and evaluate their property risk.
Canary Advise is customer facing, this is AI&ML predictions on what will be most useful to tell the end user based on what we know about their property. The predictions is based on Computer Vision result, and the advice will then be personalized based on a set of questions answered by the user at the time they take photos.
The customer receives their AI Advice from Canary in the form of Helpful tips listed out in a PDF. In the same PDF, the customer also views photos they took during a self-guided home inspection they completed earlier.
Canary Refer is internal and employee-facing specifically built for Underwriters. The Refer models evaluate for specific risks that the insurance company cares about.
The primary use case for these ML models is to help Insurance underwriters speed up and better evaluate self-inspections completed by customers who have a new or renewal insurance policy
We want to expand the use case of PropView by creating customized experiences for the following user groups: Agents, Claims, Public, and Commercial Farm/Ranch. We focused on Agent first as they are important stakeholders of the company.
Method: Remote 1-on-1 Interview; Showcasing PropView and Sales Portal demo
Sample: 12 Amfam agents over 6 states
Goal: Want to find out how would they use the tool, What’s the most and least favorite features, and if there is anything missing in PropView
Map view is useful for getting a look at the exterior of the home and when looking for any detached buildings or pools
😄 Pros: Many were already using Google satellite tools so agents were pleased to see it included in PropView.
Map being interactive and having normal Google functionality was appreciated
😫 Cons: Street View wasn’t available for many homes in rural areas
Property images is a useful section to have, as many agents are going to real estate websites just for the photos
😄 Pros: There is more value placed in the exterior photos than interior, as exterior is an indicator of the condition of the
interiorThe photos coming from MLS was a positive as MLS is sometimes more trusted than real estate sites.
😫 Cons: Missing the date the photo was taken / uploaded. As “sales” photos don’t help build an accurate quote.
Property attributes had a lot of valuable information with some even saying they’d go straight into this section
😄 Pros: Most of the information needed to complete the quote was available in the table
Having the source of the information is good
😫 Cons: Not necessarily not “work”, but there were curiosity around the accuracy of the data
Interest seemed to come down to the way an agent does business, if they see themselves as an advisor or a salesman. Agents who act as a risk advisor seeing the most value.
😄 Pros: Agents who see themselves as risk advisor saw it as a conversation to be had; others were only looking for input necessary for the quote.
The most useful sections seems to be “electrical” and “roof”
😫 Cons: There were a lot of accuracy related question. They are unsure about if there was actually something wrong or if it was just a caution.
Based on the user research results, I built a few personas for us to better recognize user needs and identify design opportunities.
Top Priority: what tool will set me apart from other agents in the city?
Can this tool help me get more new customers.
Can this tool save me more time? So I can increase the amount of quote completed in a day.
Design Opportunities: Allow user to see houses pending on sale/ want for insured/any way indicate potential customer for potential customers.
A map view for search. see more than the current house of pending sale.
The ability to search a home profile page and quickly find the relevant info to save more time.
Top Priority: How can this tool help to keep my customers or policy holders insured.
Can this tool keep me more up-to-date on my customers’ property.
If I do take pictures onsite and how easily is it to transfer, upload, and download the file.
Design Opportunities: Allow user to complete and access PropView on their mobile phone.
Being able to add, replace, upload from camera roll, download images, and give feedback would be helpful.
Top Priority: Need data for new county but not sure where to get.
Is there a way for me to request property data?
Design Opportunities: View home similar to this for customer she’s familiar with.
Able to filter for images in selected locations, eg. neighborhoods/other county.
Able to request any new info she want. Give feedback for us to add more info.
Top Priority: How can I teach new agents to ensure the continuity of knowledge regarding same data of policy holder.
How can agent adapt the style to be more modern.
How can I better understand my customers thus help them by using PropView
Design Opportunities: Allow user to take notes and give comment or suggestions on this property that are viewable for the whole team.
The ability to see who else viewed this property, assign agents to different properties, and communicate with other agents with overlapping duty.
The ability to see team members and what properties they are working on.
After identifying user needs and design opportunities, I build a information architecture for PropView agent to better help me develop my concept.
Below shows the final design work for mobile version of PropView for agent based on the design opportunities I identified from the user research study.
Able to see what properties are team members working on, View team members and add new member
Able to check team member’s comments and reply to their comments
Agents are able to view all the images corresponding to this property
Agents are able to like and save the property to review later
Can upload and download images for the specific property
Agents are able to add comment for the property and view team members' comments
.png)
Able to view the risk for electrical elements, roof, plumbing, and exterior of the property
Show predicted risk level of a specific part of the property
The collection of all saved property of the logged in agent
.png)
Show All the saved information including photos and attributes of a specific property
Canary Advise: As we want to measure the impact of our advise, we created feedback loops in the designs and scoped user research study to evaluate customer insights.
Based on the results we received, I helped to turn those insights into design questions, and conducted brainstorm session to come up with solutions with a few members in the team.
We evaluate the top voted solutions based on Feasibility and Importance
Information for better home care and maintenance is more welcomed while some others are less important.
How can we present/prioritize the content to customers that better appeals to their needs and preferences?
.png)
Some customers are concerned about data transparency, storage, and privacy.
How can we reassure users about data related concerns?
.png)
Customers want less generic but more personalized information. Some want access to more topics and in-depth coverage of topics specific to their needs.
How can we present wide and in-depth coverage of topics to users and bring a more customized experience?
.png)
Top preferences of receiving the report is through insurance company App and Email.
How can we improve the channel of sending reports to users to provide a better/seamless experience?
.png)
Customers appreciated links and tips but raised questions around generalized content and assumptions. (generalizations of extent of damage and the assumptions of prevalence of smart devices)
How can we provide more assistance and guidance to customers to make them feel supported?
.png)
To provide underwriters access to Canary Refer, I need to incorporate result of those AI&ML Refer models into Propview. I worked on designing data visualization with regard to how we make the AI predictions more understandable to the user.
I also worked on how we can measure the impact of AI models on underwriting decision making when they determine insurance coverage, and incorporate in my designs.
.png)
I choose to display override factors at the top because these are factors required to be reviewed.
On the left there is the overall referral score calculated based on the weight of referral factors, and below shows a summary of the referral data.
On the right shows a list of all referral factors and user can click to expand to view the image, value, and description of each.
User can also quickly review whether our provided information to the referral is helpful so we can get the feedback
User can view all the Ai referrals corresponding to a property after they click on any image corresponding to this property.
They can also quickly review for the referrals
My work embodied the intersection of machine learning technologies and human-centered design, ensuring the tools we developed served both the users and the business needs effectively. I am glad that my role as a UX designer impacted the organization's AI-driven products as my work been deployed by software engineer in the team.
Moreover, I am proud of the solutions I contributed to these projects and am gratified by the tangible impact they had on the insurance processes at American Family Insurance.
In the future, I would also like to see my team bringing more state-of-art technologies such as mixed reality and augmented reality into home inspection tools that we've worked on, creating more success for stakeholders and the company.
As I recognize the iterative nature of UX design and the importance of continuous learning and adaptation, I am excited to take these lessons and experiences forward in my career and to continue creating user-centered, data-informed solutions.