Pattern: pre-filter
Stop paying AI to classify ๐
Run a free regex pre-filter before the LLM. Skip noise, save tokens on every message.
Pick a message
Agent analyzing...
agent_run.output
completed
pre-filter: skipped1.0 confidence
pre-filter: skipped1.0 confidence
routed โ LLM0.93 confidence
Suggested reply
โ no reply sent
now
Suggested reply
โ no reply sent
now
Suggested reply
Extra towels are on their way to room 304!
now
{"intent": "noise","confidence": 1.0,"filter": "regex_prefilter","llm_cost": 0,"actions": [{"type": "system.noop.skip"}],"suggested_reply": null,"language": "en"}
{"intent": "acknowledgement","confidence": 1.0,"filter": "regex_prefilter","llm_cost": 0,"actions": [{"type": "system.noop.skip"}],"suggested_reply": null,"language": "en"}
{"intent": "room_service","confidence": 0.93,"filter": "passed","llm_cost": "~0.001ยข","actions": [{"type": "operations.task.create","payload": {"room": "304","request": "towels","priority": "normal","department": "housekeeping"}}],"suggested_reply": "Extra towels are on their way to room 304!","language": "en"}
This is what WhatsRB does. Every message. Automatically.
Free during beta. No credit card required.
How it works
Four steps. Zero complexity.
1
Message in
A guest sends a WhatsApp message to your number.
2
AI analyzes
The agent extracts intent, entities, and sentiment.
3
Structured output
You get a clean JSON payload with typed actions.
4
Action fired
Your backend receives a webhook and executes the action.
Ready to automate your WhatsApp?
Join the beta and start building structured WhatsApp workflows in minutes.
Free during beta. No credit card.