Skip to content
← Tools / Skills
ToolTractable/10 trustVerified
Tractable logo

Tractable

Computer vision for auto and property damage assessment, from photo to repair estimate.

Visit official site →

About

Tractable is a London applied-AI company founded in 2014 whose computer vision models assess vehicle and property damage from photos. Policyholders or repairers upload images from a phone, and the AI identifies damage, classifies it, and generates repair estimates, cutting auto appraisal from days to minutes and, through its Verisk Xactimate integration, property claim settlement from months to about a day. Its models train on billions of annotated images across millions of historical accidents.

The company's AI processes over $2 billion a year in vehicle repairs and purchases and has been deployed by more than 20 of the world's top insurers, including Tokio Marine, Covea, Admiral, MS&AD, and The Hartford, alongside partnerships with Verisk, Duck Creek, and Guidewire. It also piloted AI-verified climate insurance claims with the UN Capital Development Fund in Fiji. Pricing is enterprise and quote-based.

What it can do

  • Vehicle damage detection and classification
  • Automated repair cost estimation
  • Property damage photo assessment
  • Verisk Xactimate integration
  • Guidewire and Duck Creek partnerships
  • Salvage and recycling optimization

Setup

Enterprise integration with claims systems; mobile photo capture for policyholders and repairers.

Prerequisites

Auto or property claims operation; claims system for integration.

Use cases

auto damage photo appraisalproperty damage estimationsalvage and repair decisionsdisaster claims triage

Works with

insurersclaims-teamsfounders

Trust score — 7/10

/10 trustVerified

Scored 8/10 (TIER verified, metrics-v1). Evidence: partner announcements (Verisk, Duck Creek, Guidewire) and carrier deployments (Tokio Marine, MS&AD, Hartford) are confirmed across multiple independent outlets. Strengths: deep insurer adoption, ecosystem integrations, billions of training images. Docked for: (1) closed models, no public benchmarks; (2) volume and accuracy claims are vendor-reported. License: proprietary.

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
9.0/10
Maintenance · recency, releases, issue load
8.0/10
Openness · license + docs
4.0/10
Manual review · sanity read + red-flag check
7.0/10
  • · Partner announcements (Verisk, Duck Creek, Guidewire) and carrier deployments (Tokio Marine, MS&AD, Hartford) confirmed across outlets
  • · Deep insurer adoption, billions of training images; volume claims vendor-reported

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.