N°
N°001
Name
WhatsApp enquiries, answered automatically
Client
Turbo Diesel Engine Parts
Year
2024–2026
Type
Conversational automation
Stack
Twilio WhatsApp API, FastAPI, Redis, PostgreSQL, Anthropic Claude, Google Sheets
Outcome
119 of 119 live transitions walked by a test; 31 templates held in git with a rebuild command; 270 of 274 recent messages read
About
Turbo Diesel Engine Parts sells marine and industrial engine parts to trade customers across Europe, and its customers were already on WhatsApp. We built what answers them now: a menu on the number they already use (Sales, Track an order, Admin), an assistant that captures the enquiry from a fixed facts sheet and hands it to a person rather than guessing, a reference number allocated from the client's own price sheet, and a console where the representatives answer. Every entry point — the site button, a link, a QR if one is ever printed — goes through one redirect on our own domain, so a number change never invalidates any of them. The 31 approved message templates live in git beside the code, held equal to the account by a test, with a rebuild command for the day Meta or Twilio has to be started again. 28 states, 127 transitions, 119 of them walked by a test; 426 customer journeys mapped.
One redirect on our own domain: a number change never invalidates a button, a card or a QR code
31 approved templates, held equal to the git copy by a test, with a rebuild command for a new sender
A thread in the representatives' console, on sample data: the assistant's capture, then the reply



