Liability Analysis Tool
Designing a claims-ready liability experience for auto insurance agents — turning scattered accident evidence into a clear, confident liability decision.
Determining liability meant piecing together evidence scattered across disconnected tools.
When an auto accident claim came in, adjusters had to manually gather and cross-reference evidence — scene photos, vehicle damage, speed and distance data, police reports, and driver statements — before they could confidently determine liability. That evidence lived in different systems, in different formats, with no single place to evaluate it together. The result was slower claims, inconsistent liability decisions, and adjusters spending more time hunting for information than analyzing it.
What I owned
- End-to-end UX from research to high-fidelity design
- Stakeholder interviews with claims adjusters and engineering teams
- Information architecture for the liability decision workflow
- Key screen design — scene overview, internal/external factors, damage comparison, transcripts
- Design rationale documentation for engineering handoff
- Usability testing and iteration
Adjusters didn't need more data. They needed the data they already had to live in one place, organized around the decision they were trying to make — not around the systems it happened to come from. Every screen we designed had to answer one implicit question: does this evidence point toward or away from liability? If a screen couldn't answer that, it didn't belong in the primary workflow.
Three decisions that defined the tool
Scene overview as the anchor screen, not a supporting one
Early concepts treated scene photos as supplementary evidence accessed from a sidebar. Adjuster interviews revealed the scene was actually their starting point for every liability judgment. We redesigned the scene overview as the primary landing screen — full context first, supporting detail second.
Separate internal and external factors, side by side
Liability depends on both what happened inside the vehicles and what happened in the environment around them — speed, distance, road conditions, vehicle damage. We designed a dedicated internal/external factors screen that put both data sets side by side, so adjusters could correlate them directly instead of holding two separate mental models.
Transcripts as searchable evidence, not static documents
Driver statements were stored as flat transcripts adjusters had to read top to bottom to find relevant details. We redesigned transcripts as structured, searchable evidence — surfacing key statements relevant to liability rather than requiring adjusters to read the entire document every time.
What shipped and what it changed
The Liability Analysis Tool gave Allstate's claims adjusters a single, structured workflow for evaluating liability — replacing a fragmented process of cross-referencing disconnected evidence sources. By organizing every screen around the liability decision itself rather than the system the data came from, adjusters could move from evidence to determination with more consistency and confidence.
If we had another quarter
I would have pushed for a guided liability recommendation — using the same evidence adjusters were already reviewing to surface a preliminary liability assessment they could confirm or override. We built the tool to organize evidence for human judgment. The natural next step is using that same structure to make the first pass at the judgment itself, with the adjuster always in control of the final call.