Auto-Owners Insurance is a Fortune 500 company that provides automotive, home, life and commercial insurance to nearly three million policyholders in 26 states. Headquartered in Lansing, Michigan, the company is recognized for exceptional financial strength and customer service.

At Auto Owners, employees utilize a variety of AI models, with various strengths and weaknesses each. Currently users must pick-and-choose the best models for specific tasks, leading to unnecessary confusion and inefficiencies.

AO Harmony mitigates this by tasking multiple large language models with the same prompt and returning only the most relevant answer. This helps Auto-Owners developers communicate more effectively with large language models and to obtain higher quality responses.

Users first sign in and insert their question, task, or prompt. Our software takes that input and forwards it to several large language models. Another model then ranks each response and returns all responses in a ranked order of relevance. Through a simple and intuitive UI, users can better leverage the different strengths of modern large language models.

AO Harmony saves chat history so that users can return to past conversations without having to regenerate responses. Benchmark data is stored so administrators can better analyze large language model selection and performance for future proofing.

Our system simplifies access to different large language model capabilities and helps people aggregate and use different skills and features as needed.

AO Harmony uses ReactJS with a Python Flask back end. Persistent storage is handled with MongoDB. Our software uses Docker for containerization. Ollama runs the large language models locally without depending on external model providers.