Vendor reportedLive since Feb 2026Added Oct 2026

Travelers

Travelers uses a voice AI agent to take auto damage claim calls

Travelers built an AI voice agent on OpenAI technology that takes first notice of loss calls for auto damage claims. After launching in eight states it went countrywide within two months. OpenAI reports that 85-90% of customers using it now complete their claim filing through the AI.

Key figures

Copied exactly from the sources. Copy any figure with its citation.

85

–90%

Customers using the AI Assistant who now complete their claim filing through AI1

within two months

Time from the eight-state launch to countrywide availability1

more than 1.5 million

Claims Travelers handled in the previous year (context for surge demand)1

more than 100,000

Claims that a catastrophe event can generate in just days (context)1

What it took

CostNot disclosed
TeamNot disclosed
TimelineNot disclosed

Cost, team size and timeline were not disclosed in the sources.

The problem

When a car accident happens, customers want help right away. Catastrophe events can generate more than 100,000 claims in a few days, which strains any call center. Travelers handled more than 1.5 million claims last year.

What they built

The AI Claim Assistant, a fully autonomous voice service. Customers call, speak naturally, and the assistant walks them through the first report of a loss on a car damage claim: it answers policy questions, collects details and submits the claim. Callers can then continue in a digital experience to upload photos, start appraisals, schedule repairs and reserve a rental car.

How they did it

The service is built on OpenAI's Realtime API, a voice interface to its models, and OpenAI's frontier models. Travelers connected the models to its claims systems, workflow tools and internal tools. It launched in eight states, then expanded countrywide within two months. Travelers' claims executive said the real-time model stood out for its ability to perform in this environment. Travelers plans to extend the assistant to other lines of business.

Results

According to OpenAI's story, 85-90% of customers using the assistant complete their claim filing through it. Customers get support around the clock without waiting, including during catastrophes, while claim professionals handle the complex cases. The figures are reported by the vendor and not independently confirmed.

Lessons for your project

  1. Start with one narrow, high-volume step. Travelers began with first notice of loss for auto damage claims, then planned other lines of business.

  2. Size the tool for your peak, not your average. Catastrophe events create claim surges that strain a call center, and the agent has no wait time.

  3. Keep humans for the complex cases. Travelers positions the assistant to free claim professionals for work that needs human expertise.

  4. Roll out in stages. Eight states first, then national, let Travelers learn before scaling.

Caveats

  • The completion figures come from OpenAI's customer story, a vendor source. The sources include no Travelers-published measurement, and no independent source confirms them.
  • The OpenAI page was read through an archived copy because the live page blocked automated access.
  • The OpenAI story headlines 90% and the body says 85–90%. The range is used here. It does not say how completion is defined or how many calls were handled.
  • The deployment date is the February 2026 launch announcement, reported by trade press dated 18 February 2026. The sources do not give a go-live date.
  • Trade coverage of the launch repeats Travelers' own announcement, so it does not independently confirm the results.

Sources

  1. Travelers deploys AI-powered claims countrywide with OpenAI

    OpenAI · 2026-06-02 · Vendor case study · openai.com

  2. AI Claim Assistant Now Taking Auto Damage Claims Calls at Travelers

    ClickClaims · 2026-02-18 · News · clickclaims.com

  3. Travelers launches agentic AI claim assistant developed with OpenAI

    Coverager · Undated · News · coverager.com

Summarized in our own words from the sources above. Spotted an error? See corrections.