Intent recognition is the process of identifying what a person wants from what they say or type. A message like "can I swap these for a bigger size" has the intent "exchange an item", whatever the exact wording.
Older chatbots matched keywords or trained classifiers on a fixed list of intents. Modern systems use large language models, which cope better with unusual phrasing, spelling mistakes and several requests in one message. Many still combine the two, using a defined list of intents to trigger set workflows and a fallback when nothing fits.
Retailers and venues use intent recognition to route messages, such as separating bookings, order status, complaints and sales enquiries, and to decide whether the AI can answer or a person should. It also shows which questions customers ask most.
Common problems are ambiguous messages, slang, multiple intents and customers who change topic. Look at how the system behaves when it is not confident: a good one asks a clarifying question instead of guessing. Review conversation logs regularly to find intents that are being missed. Update the list when you launch new products, menus or services, or customers will ask things the system has never seen.
It matters because every automated response depends on first understanding the request, and mistakes here lead to irrelevant replies and frustrated customers.