Hotel concierge chatbot
Front desks answer the same forty questions every day. This assistant takes those, in plain language, at any hour — so the staff can spend their attention on the guest standing in front of them.
Understands the question
Natural language in, useful answer out — no menu trees, no keyword matching, no asking the guest to rephrase.
Answers in the moment
Replies arrive fast enough to be worth using at 2am, which is when most of these questions actually get asked.
Knows its limits
Anything unusual gets handed to a person, with the context already collected rather than a guest repeating themselves.
How it’s built
Python and Flask on the server, the OpenAI API for language understanding, and a prompt layer that keeps the assistant inside what the hotel actually offers. The interesting work wasn’t the model call — it was deciding what the bot must never answer on its own.