Ragic announced that its AI Agents now automatically use Jev, a specialized decision model from TypeSafe AI, instead of a general-purpose LLM for qualifying classification tasks, helping businesses reduce the cost of repetitive AI automation without requiring users to select models, write code, or configure additional integrations. Ragic checks whether an AI Agent can use Jev when users save it and automatically selects the model when appropriate, notifying users in the interface. Jev is designed to make decisions from predefined options rather than generate free-form text. Ragic's demo-follow-up Agent, which helps its sales team track customer inquiries, update categories, and generate follow-up tasks, was initially expected to cost less than $1 per month but ran more than 10,000 times over several weeks; at approximately $0.02 per run using Gemini Flash, the total cost reached about $200. Based on model pricing, Ragic estimates that switching this type of classification task to Jev could reduce model costs by approximately 97–98%. CEO and Founder Jeff Kuo said an agent that only needs to choose between a few options doesn't always need a large language model, and automatic selection of Jev helps businesses reduce unnecessary AI costs without changing how they work. Actual savings will depend on the task, model usage, and pricing.