User Guide: Setting Up Knowledge Sources for AI Chatbots on GTG CRM

What you can do after this guide

Open the Knowledge Source section → upload your documents → wait for the system to finish processing → enable the capabilities you want the bot to perform → try asking the bot to test its answers based on the documents. Then your bot will answer customers correctly and perform real tasks.

Preparation

#	Condition	Why it's needed
1	Log in to GTG CRM, with AI Chatbot enabled for at least one channel.	Knowledge Sources are the configuration of the AI chatbot.
2	Have business documents ready (policies, price lists, FAQs, guides).	This is the content the bot will rely on to answer.
3	(Optional) Have order, product, or calendar data for the bot to look up.	To enable order lookup/stock checking/appointment scheduling capabilities.

Step 1 — Open Knowledge Sources and upload documents

Go to the AI Chatbot configuration section, open the Knowledge Sources tab.
Upload your documents (policies, price lists, FAQs, product guides).

✅ Expected Result: Documents appear in the knowledge source list and change to a processing status.

Screen for uploading documents as knowledge sources

Step 2 — Wait for the system to finish processing

Wait for the document status to change to ready.

✅ Expected Result: Each document shows its processing status; when finished, the bot can search its content to answer.

List of knowledge sources with processing status for each document

Step 3 — Enable the capabilities you want the bot to perform

Open the bot's capability toggle section.
Enable the capability groups you want: answering from documents, looking up orders, checking stock, creating support tickets, scheduling appointments, etc.

✅ Expected Result: Only the capabilities you enable will function; lookup actions respect your data scope.

Screen for toggling individual chatbot capabilities

Step 4 — Try asking the bot to test

Open the test chat window and ask a few questions found in the documents (e.g., ask about the return policy).
Try an action query (e.g., ask about the status of an order).

✅ Expected Result: The bot answers based on the documents you uploaded (accurate to business logic, not making things up); for action queries, the bot looks up and provides real data or performs permitted actions.

Quick Reference Table of Expected Results

Action	What you will see
Upload document	Document appears in the list, processing status
Finished processing	Ready status, bot can search
Enable capabilities	Only enabled capabilities will function
Test query	Bot answers from documents, performs real tasks within limits

Notes

The bot answers based on the documents you provide — the more complete the documents, the more accurate the answers. Supplement documents when the bot's answers are insufficient.
Bot actions (order lookup, stock check, ticket creation, appointment scheduling) respect your data scope and permissions and are designed for privacy protection.
Monitor the handoff rate to humans — a high handoff rate is a sign that documents need to be supplemented.