Nova: A Real-Time AI Sales Call Assistant with Live Speech Transcription, Knowledge-Grounded Coaching and Simulated Sales Training

Sales representatives must respond to customers in real time while recalling product details, recognizing customer needs and buying signals, and deciding on suitable follow-up actions. This paper presents Nova, a full-stack Real-Time AI Sales Call Assistant that supports representatives during customer conversations. The browser captures live speech through the Web Speech API and converts it into a running transcript, and the React frontend sends the relevant conversation data to a Flask backend through REST APIs. An AI coaching service analyzes the conversation context and returns live suggestions, namely a recommended response, a follow-up question, objection-handling guidance, detected buying signals and a next action. Lead management, call history, call notes, follow-ups, a dashboard and follow-up email workflows are supported using SQLite and SQLAlchemy with JWT-based authentication. A training module simulates customer scenarios and scores the practice session. Completed calls store the transcript, call duration, speaker information, sentiment, summary and follow-up recommendations. The implemented system supports live microphone capture, speech transcription, real-time coaching suggestions, lead management and call-history storage. The current coaching logic is built around rules, knowledge and application logic, and no formal predictive-model accuracy is reported. The system integrates live transcription, coaching, record keeping and training within a single web application.