Glossary · AI in customer service
What is RAG for customer support?
In short
RAG (retrieval-augmented generation) is how an AI answers from your own documents. It first searches your knowledge base for the passages that match the customer’s question, then writes an answer from those passages, ideally with a link to the article it used.
Reviewed 9 November 2026. How it works in Axonave: Axonave AI.
Search first, then answer
A language model on its own answers from what it learned in training, which knows nothing about your delivery times or your warranty. RAG adds a search step: before answering, the system finds the parts of your articles that match the question and gives only those to the model.
That has two effects. Answers follow your policies, and each one can point to its source, so a customer or an agent can check it. It also means the AI is only as good as your articles. An outdated article gives an outdated answer, and a missing one should mean no answer and a hand-over to a person.
Good support RAG searches by meaning as well as by keyword, so “bila sofa saya sampai?” finds the delivery article written in English.
- Retrieve: find the matching passages in your articles
- Generate: answer from those passages only
- Cite: link the article, or hand over when nothing matches
How it works in Axonave
Answers only from your published articles, with the source shown.
Axonave’s AI answers only from your published, customer-facing knowledge base, and shows the article it used. Search combines keywords and meaning, in English, Malay and Chinese. When it isn’t sure, it hands over to the right team, and the questions it couldn’t answer are grouped into drafts for a person to publish.
- Limit an AI answer step to the knowledge bases you choose
- Answers streamed, so the customer sees them start quickly
- Unanswered questions turned into draft articles
In our demo workspace, the AI answers from the published help-centre article, names its source and offers to take a message.
RAG (retrieval-augmented generation): common questions
Does RAG stop an AI from making things up?
It makes it much less likely, because the AI answers from your articles instead of its training. It doesn’t replace good articles or a way out: in Axonave, with no published article to answer from, the AI hands over to a person.
Do we need to train our own model?
No. You publish articles; the AI reads the matching ones at the moment it answers. Update an article and the next answer uses the new version.
Can the AI answer from our internal documents?
Not to customers. Axonave’s AI answers customers only from published, customer-facing articles. Agent Assist, which only your team sees, can draft from internal knowledge bases too.
See it working with your own customers.
We reply within one business day by email or WhatsApp.