I love projects where code stops being just code and becomes part of people’s everyday lives. This is one of them.
Coffee Point is a coffee shop inside a business center. I designed and built an AI agent for it that works as a consultant and takes orders. It has been running for 1.5 years now, and every day it helps real people get their coffee faster.
Here is the situation. Depending on the day and the season, between 80 and 560 people come through: office employees, tenants and visitors. Imagine being at your desk in the middle of a busy day. You don’t want to stand in a queue, you just want your coffee. So you open the agent, ask it what to choose if you can’t decide, place your order, and go down to the cafe exactly when your drink is ready or almost ready. No waiting and no crowd at the counter.
The system runs on a local web server and is available only to people who work in the business center. Visitors from the street can’t pre-order, but they are served in the cafe as usual.
I built it with a local web server, an AI agent on Ollama, PHP, JavaScript, Python and MySQL.
What changed
We measured the effectiveness only after three quarters of analysis, because I wanted honest conclusions, not first impressions. Financial data is confidential, so I will describe the results in words:
- There are fewer mistakes in orders.
- Sales of extra items went up: pastries, and salads at lunchtime.
- Orders from the cake and pastry suppliers became more accurate, and there are significantly fewer leftovers.
- The average check grew.
- Customers leave many positive reviews, and they mostly thank us for the fast service.
For me, the best part is the team. The cafe staff work more freely and without stress, and that matters to me more than any number.
I have no separate data on customer loyalty, so I won’t make claims about it. Reviews speak for themselves. 🙂