Auto insurance companies deployed AI and photo-based estimate systems during the pandemic to replace in-person appraisers, but these systems frequently produce inaccurate estimates that undervalue repair costs, causing financial disputes and potentially incomplete vehicle repairs.
During the COVID-19 pandemic, auto insurance companies accelerated their adoption of AI and photo-based estimation systems to replace in-person appraisers. Before the pandemic, about 15 percent of US auto claims were settled using photos, but this increased to 60 percent, with expectations to reach 80 percent by 2025. Companies like CCC Information Services report that half of all claims now involve some AI tools, with their Smart Estimate product boosting productivity by 30 percent. Tractable, which uses computer vision and machine learning, says 25 percent of its estimates require no human intervention, with plans to reach 75 percent by year-end. However, auto repair shop owners report that 99.9 percent of AI-generated estimates are incorrect, according to Jeff McDowell of Leslie's Auto Body. The systems struggle to identify issues like suspension damage, bent wheels, or frame misalignment from photographs alone. This has led to increased disputes between repair shops and insurers, with vehicles sometimes stuck in shops longer than usual. Some repair shops report seeing more incomplete or shoddy work during the pandemic, attributed to inadequate photo-based estimates. The AI systems are trained on millions of photos of damaged cars and use local parts and labor data to estimate costs, but cannot detect hidden damage visible only through physical inspection.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
No population impact data reported.