Communication on auto pilot
I just rebuilt my entire STR guest communication bot from scratch, and the difference is night and day. For the past few weeks I’ve been trying to get an AI auto-responder working using Zapier + Claude. The concept was right but the execution kept failing. The system prompt had all properties hardcoded into one massive block of text. Rate limit errors. Wrong answers. Tokens through the roof. Every test run either timed out, hallucinated, or fired when it wasn’t supposed to. So I scrapped it and rebuilt it the right way. Here’s the new setup: → Google Apps Script runs every 5 minutes → Pulls new guest messages directly from Hospitable’s API → Loads ONLY the relevant property’s knowledge base from a simple text file in Google Drive → Sends it to Claude to generate the response → Sends the reply back through Hospitable automatically The knowledge base is just text files in a Drive folder. One file per property. Shared rules in a separate file. Need to update something? Open the doc, change the text, save. Done. ✅ Live on the next run. No Zapier. No rebuilding prompts. No digging through hundreds of lines of code. Adding a new property? Fill out an onboarding form, hand it to Claude, it generates the file in minutes. The cost difference is the best part: ❌ Old way → loading all properties every single run = 30,000+ tokens, constant rate limit errors ✅ New way → loads only 1 property per message = ~2,000 tokens 💰 Cost per message: ~$0.009 💰 100 messages/month: under $1 It runs 24/7, handles inquiries and reservations, screens for red flags, escalates when it doesn’t know the answer, and texts me when it needs me. My guests can’t tell it’s not me. If you’re running STRs and still copy-pasting the same answers to the same 15 questions every week, there’s a better way. Drop a comment if you want to know more. P.S. Here is an inquiry test run. I had Carissa send me messages. With the exception of one message I had an auto WiFi response turned on in Hospitable. 🤦♂️