Intent Recognition
The NLU process of identifying what action a caller wants to take based on their words.
Intent recognition is the process of determining what a caller actually wants to accomplish from what they say. It is a core function of natural language understanding (NLU): the caller says "I need to move my appointment to next week," and intent recognition classifies that as a reschedule_appointment intent, then extracts the relevant entities (the new date).
Intents vs. entities
- Intent = the goal or action ("book", "cancel", "check status").
- Entities = the parameters that fill it in (date, name, order number, party size).
Together they turn an open-ended sentence into a structured, actionable request the system can execute.
Classic NLU vs. LLM-based recognition
Traditional NLU engines classified intent against a fixed, pre-trained list — accurate but brittle, failing on phrasings they never saw. Modern AI voice agents use large language models, which generalize to novel phrasings and handle multi-intent utterances ("cancel Friday and rebook for Monday") in a single turn. Strong intent recognition is what makes intent-based call routing possible — and it directly drives first-call resolution by routing or resolving correctly the first time.