RAG Search
Search a customer's whole conversation history and get answers that cite the original source.
What it is for
RAG Search (Retrieval-Augmented Generation) lets you ask questions in plain language about a customer's conversations, emails, notes and tickets. The answer cites its source (email X from 3 March, a WhatsApp from Ana the CSM, meeting note Y) so you can verify it and open the original context.
Unlike keyword search, RAG understands meaning: asking "what are Juan's objections to the price?" returns the relevant moments even if Juan never used the word "objection".
When to use it
- Research before a meeting: "what did we discuss last time about integrations?"
- Prepare a renewal: "what has this customer asked for that we have not delivered?"
- CSM handover: "what are the sensitive points with this customer?"
How to reach it
- From the customer record: Ask Luminance tab, then type your question.
- From a specific contact: contact record, Ask about this contact (useful when a company has several contacts with different opinions).
- Main menu: Intelligence → RAG Search to search inside the context of an open customer.
Example
Joaquín has a meeting with Acme Corp in 20 minutes and has never handled the account. He opens the record, Ask Luminance tab, and types: "what are the open topics from the last 60 days?". RAG returns 4 moments: an unresolved bug reported by Ana (the champion) on 2 March, a HubSpot integration request from the CTO, a complaint about support response time, and praise for the new module roll-out. Each point links to the original email or message. Joaquín walks into the meeting prepared in 90 seconds.
What it does not do
- It does not draft replies to the customer. Copy the context where you need it, or use Ask this customer if you want a structured answer.
- It does not search across your whole book (that is Luminance Chat).
- It does not invent information that is not in the data. If there is nothing to answer with, it says so.
Known limitations
- Data is indexed roughly every 5 minutes; very recent conversations may not show yet.
- It only indexes emails, WhatsApp, notes and meetings inside LealUp. External data (Jira, Slack without the integration) does not appear.
- Citations keep the original message's language; answers are generated in the language of your question.
Last updated: 2026-04-21.