Support & AI Agents
AI agents are created using the Claude API and configured with a system prompt that acts as their knowledge base and instructions.
How are AI agents created using the Claude API with a knowledge base?
AI agents are created using the Claude API and configured with a system prompt that acts as their knowledge base and instructions.
How do agents gather product information from across the system to answer questions?
Agents understand the catalogue by gathering data from across the system — product details, availability, pricing, and order information — so their answers reflect live data rather than a…
What is the pre-sales agent's goal and what can it do (place orders, send payment links)?
The pre-sales agent's goal is to sell.
What does the after-sales agent handle?
The after-sales agent handles post-purchase support — a different agent with different responsibilities from the pre-sales one.
Can multiple agents with different roles run at once? How are they configured?
Yes — multiple agents with different roles can run simultaneously, each configured independently with its own name, role (pre-sales or after-sales), system prompt/knowledge, model, and…
How can staff monitor any ongoing conversation in real time?
From the admin panel, staff can monitor any ongoing conversation in real time — viewing the live message thread of an AI (or human) conversation as it happens, across web and WhatsApp channels.
How does a human join and take over a conversation from the AI?
A human can join a conversation and take over from the AI — at which point the AI pauses and the human continues with the full context preserved.
How does escalation follow the company's configured algorithm when a customer asks for a human?
When a customer asks for a human (or the AI decides to escalate), escalation follows the company's configured algorithm — routing the chat to an available human according to the rules the…
How does a human pass a chat to another human while staying synchronized?
A human can transfer a chat to another human internally, and the whole system stays synchronized — the receiving agent gets the full context, and the conversation continues without loss.
What happens when no human is available on escalation?
If no human is available on escalation, the system handles it gracefully per policy — for example queuing the conversation, offering a callback, or having the AI continue within…
What does the AI do when it lacks knowledge for a query (graceful fallback, no fabrication)?
When the AI lacks the knowledge to answer a query, it is designed to escalate rather than fabricate — a graceful fallback to a human (or a safe "let me get someone who can help") instead…
What channels do support conversations run on (Web, WhatsApp)?
Support conversations run on two channels: the website chat (embedded via a copy-paste JavaScript snippet) and WhatsApp (via webhook integration).