The AI agent now automatically analyzes chats and identifies gaps in information and topics with low data quality.
The Yandex team has presented an update to its proprietary "Neurosupport" service, adding tools for creating and evaluating a knowledge base for text support.
Now companies can form a knowledge base based on real dialogues with customers, receive recommendations for improvement, and automatically analyze the quality of responses. According to the developers, the solution is especially useful for organizations with large volumes of requests, complex products, or rapid business growth.
The service allows you to create a structured knowledge base, on the basis of which AI responses and prompts for operators are formed. In your personal account, it is now possible to check the model's responses before sending them to customers, which helps to quickly identify gaps and inaccuracies in the knowledge base.
As a reminder, the "Neurosupport" AI agent is designed to solve narrowly focused support tasks and is based on the Alice AI LLM neural network. The service is based on lightweight models retrained on instructions for operators of more than 50 Yandex services, and RAG technologies are also used.
