What is Natural language understanding (NLU)?

Natural language understanding (NLU) is the branch of AI concerned with working out the meaning of human language, including intent, entities and context, rather than only transcribing the words. In a voice pipeline, speech recognition produces text and NLU interprets it. Older systems used separate NLU engines made of intent classifiers and entity taggers; most modern voice agents fold this work into a large language model. The hard cases in India are code-switching, indirect refusals, honorifics and the transcription errors that NLU must be robust to, such as a misheard locality name.

A perfect transcript is useless if the system misreads it. NLU quality decides whether the agent responds to what the caller meant rather than to the literal words, and whether the outcome recorded after the call matches what actually happened. It is often the difference between a demo that impresses and a deployment that holds up on real customers.