US auto insurers are using AI to generate nearly instantaneous repair claims, using only photos of the damage, with COVID-19 spurring on adoption
Using algorithms, tech companies are helping insurers speed up the process after an accident, make it more accurate and keep estimators out of the field — a plus during a pandemic. Tweets: @tractable_ai , @sub8u , and @nytimesbusiness Tweets: Tractable / @tractable_ai : A must-read: how our #AI is changing the game for analysing and understanding auto damage, featured today in @nytimes and including an in-depth interview with @AlexDalyac https://www.nytimes.com/... Subrahmanyam Kvj / @sub8u : That's...quite high, no? “On a typical day, about 80,000 American drivers have accidents serious enough to warrant calling their insurers.” https://www.nytimes.com/... @nytimesbusiness : Using algorithms, tech companies are helping insurers speed up the process after an accident. The Tractable algorithm is “kind of magical, but it's very data hungry,” a founder of the company said. https://www.nytimes.com/...
Context & Ripple Effects
Photo-based claims estimation did not start with the pandemic: as early as 2017, insurers were already using drones and AI to inspect damage and automate decisions, and customer satisfaction with home and auto claims rose alongside that automation. Ant Financial had also shown by mid-2017 that AI could estimate car-insurance damages at consumer scale in China.
What changed by September 2020 is that US auto insurers moved to nearly instantaneous repair estimates from photos alone, keeping human estimators out of the field during COVID-19 — and vendors like Tractable turned that shift into venture-scale businesses, later reaching a $1B valuation.
First-order effects
- Field estimators and body-shop appraisal workflows lose their gatekeeping role: repair costs are priced from photos before anyone physically inspects the car.
- Tractable and similar computer-vision vendors move from pilots to core claims infrastructure at US insurers, with the pandemic serving as the adoption forcing function.
Second-order effects
- Capital follows the workflow: within nine months Tractable raised a $60M Series D at a $1B valuation, and Akur8 pulled in a $30M Series B for claims automation — competitors and investors now treat claims AI as a funded category rather than an experiment.
- Insurers that lag on photo-based estimation face a cost and speed gap against rivals whose claim cycle is measured in minutes, pressuring them to buy rather than build.
Third-order effects
- If photo-only estimation becomes the default, the same data pipeline extends toward pricing: insurers were already adjusting premiums and policies based on new forms of surveillance, and granular damage data feeds that loop.
- The pattern also carries the risk documented at One Concern, where an AI vendor exaggerated its tools' capabilities — mispriced repairs at scale would force regulators and insurers to audit algorithmic appraisals the way they once audited adjusters.
The trend: Insurance claims processing is shifting from physical inspection by human estimators to photo-based AI estimation, with COVID-19 compressing years of adoption into months.