Bevaya's Insurance-Trained AI Outperforms General-Purpose Models in Loss Run Accuracy
Bevaya has demonstrated that its insurance-specific AI model, InsurGPT™, outperformed leading general-purpose models on loss run accuracy by a significant margin. In tests involving 346 real-world documents, Bevaya's model achieved 93.1% field accuracy compared to the best general-purpose model at 85.3%. The results highlight that domain-specific training is more effective than simply using larger or newer models. The benchmark was conducted under identical conditions and showed consistent performance across all tested general AI models, which scored between 78% and 85%, with no significant improvement from higher-priced configurations. Bevaya's model also processed documents faster—less than half the time of its competitors—which is crucial for operational efficiency in insurance. The company has deployed over 120 production systems at major U.S. insurers, including three top property and casualty carriers. InsurGPT’s accuracy on critical fields like claim numbers and policy IDs was significantly higher, reducing errors that could lead to misfiled claims or incorrect submissions.
Topics: Artificial Intelligence Insurance Technology Loss Run Analysis