Breaking the Mold: Designing the First AI-Powered Claims Agent in US Auto Insurance

Background

In September 2024, I took over the design responsibility for Fairmatic’s claims experience — a longstanding pain point for our customers. The existing process was subpar, heavily reliant on third-party tools, and offered little user value. While a redesign had already been scoped and partially executed by the previous designer, core issues still remained.

The Problem

Redesigning the existing flow was necessary, but it didn’t solve our biggest challenge:

Most drivers submitted claims by calling or emailing their fleet managers — resulting in many claims being handled offline and never entering our system.

Due to poor past experiences, drivers rarely used our app to report incidents.

In fact, many drivers hadn’t even installed the app — which meant we had to consider alternate channels to reach and serve them effectively.

From a product and operational standpoint, we faced additional constraints:

The MVP scope was limited strictly to the app redesign.
There was no pathway to propose or add net-new solutions to the roadmap.

The Solution

Around the same time, Eleven Labs released their voice agent API — a fortunate moment that gave me the technical foundation to explore voice-based claims reporting. Seizing this opportunity, I proposed and prototyped three tech interventions:

AI Voice Agent – Let drivers report incidents via phone or in-app through an AI agent thoughtfully crafted to sound natural, human, and approachable — making a typically stressful experience feel simple and conversational.
Automated Email Parser – Extracting claim data from incoming emails to auto-generate claim reports.
Proactive Outreach – Triggering automated messages and emails to drivers upon crash detection, prompting immediate claim reporting.

To navigate roadmap constraints and gain buy-in:

I circulated the voice agent prototype with leadership and stakeholders to generate awareness and demonstrate value.
I pushed the idea as a standalone hackathon project, where it won — catalyzing support and roadmap inclusion.

A prototype of the AI voice agent that helped gain stakeholder buy-in.

QR code for prototype — Try out our very first prototype (subject to tokens - may not work due to excessive usage)

Impact

Within a month of launch, the impact was clear:

Claims submitted directly through our system rose from 33% to over 60%.

Average lag time from incident to report dropped from 5 days to 2 days and 22 hours.

Driver satisfaction scored 4.5/5.

Process

Due to NDA restrictions, reach out to adithkvn@gmail.com for a detailed walkthrough.

Broadly, my approach included:

Evaluated emerging tools like Eleven Labs' newly launched voice agent to determine feasibility for integration.
Analyzing user behavior and identifying patterns (e.g., preference for calling).
Aligning with stakeholders on core problems.
Ideating and prototyping multiple solutions.
Internally testing with stakeholders, then iterating with drivers.
Advocating for roadmap inclusion.
Launching and refining based on real-world feedback.

What Could’ve Been Better

More pre-launch testing: Some edge cases weren’t captured initially. Drivers’ varied ways of interacting with the agent posed unexpected challenges.

Learnings

The roadmap isn’t static — user needs and well-placed ideas can create room for impact.
Trust and visibility matter — strong stakeholder relationships can amplify the reach and adoption of your ideas.
Be ready to act on opportunity – The timely release of Eleven Labs' voice agent became a catalyst. Good ideas often need timing, and being prepared helped turn that moment into impact.

THE END