Progressive Insurance
Estimate Automation
Machine Learning / Predictive AI / Web App
Role
UX Research Lead
UI Design Lead
Work
Product research and discovery
Research synthesis
UI design and prototype
Process and flow design
Key Product Takeaways
Custom AI/ML model training using completed claims images
Machine vision identifies all relevant information and pre-fills claim for human approval
VIN, make, model, repair or replacement, part options, labor hours, etc.
Model improvement loop retrains models and measure accuracy
Human agents become more available to focus on claims process and approvals
AI does NOT handle any approvals - focus is on identifying information and ideal course of action
Project Summary
UX Considerations
Must accurately match a wide variety of vehicles to correct parts
System must account for non-visible damage like sensors and cameras
Must maintain a final human review step for accuracy and trust
Enable agents to easily train the system using previous image data
The Problem
Claims processing consumes 75¢ of every $1 earned, making it costly and inefficient
Lack of standardized estimating practices causes inconsistent customer experiences
Delays in processing increase time for customer repairs and vehicle return
Outcomes
Claims are processed faster, reducing cycle time and improving efficiency
Vehicle attributes and damage are accurately identified
Customers receive consistent service and quicker repair turnaround
Agents have more time to focus on customer support and communication
Some UI screenshots!
Ways we learned
1. Generative Research
2. Evaluative Research
3. Evaluative Proof of Concept
