Understand the request
Normalize the incoming message and attach the context that makes the question meaningful: organization, conversation, contact, and channel.
Wazapin turns a WhatsApp message into a grounded resolution—checking context, choosing the right path, and knowing when a human should take over.
Illustrative live trace · CX-0842
Scroll to drive one run. The visual trace stays in place while each decision enters the viewport.
sample runCX-0842
Normalize the incoming message and attach the context that makes the question meaningful: organization, conversation, contact, and channel.
The engine classifies intent and skill, then moves the request to the right next step automatically.
Search the tenant’s knowledge with semantic and lexical signals, fuse the candidates, then rerank the evidence for this exact question.
The model writes a concise response using the selected snippets, the agent brief, and soft SOP guidance. Missing evidence means no invention.
Before auto-send, the answer passes a final gate. If it is empty, weak, unsafe, or ungrounded, the engine stops and escalates honestly.
One interface, different decisions. Tap a scenario to see how the final branch changes without changing the core pipeline.
“Refund biasanya berapa lama?”
Retrieve the refund policy, compose a concise answer, validate the source, then send.