Personal RAG: chat with your own data

Ask a general chatbot "where are we on the Dupont project?" and it has no idea — or worse, it makes something up. Personal RAG solves that. It lets an AI search your information first, then answer while citing what it found.

RAG in a nutshell

RAG stands for *Retrieval-Augmented Generation*. When you ask a question, the system first retrieves the relevant passages from a set of documents, then asks the language model to write the answer from those passages.

A personal RAG applies that principle to your own data: your emails, calendar, documents and notes. The AI no longer relies on its general memory but on what you actually wrote and received.

The result: precise answers about your work, and far fewer inventions, since the answer has to rest on retrieved passages.

Why sources change everything

A good RAG answer comes with its sources: the email, document or note the information came from. You can check in one click instead of taking the AI at its word.

That's also what sets it apart from a plain search engine. Search gives you a list of results to read; RAG reads for you, summarises, and shows you where it read.

For a professional, it's the difference between "the AI thinks that" and "the client wrote it on 12 March, here's the message".

The privacy questions

A personal RAG indexes sensitive data. Before adopting one, ask: where is the index of your documents stored? Is access to your inbox read-only? Is there a GDPR-compliant data processing agreement (DPA)?

Also check what you control: being able to remove a document, choosing what gets indexed, and knowing which subprocessors handle your data.

For a European company, hosting in the EU makes those answers much simpler.

Personal RAG in Virtus Lever

Virtus Lever's AI chat is a personal RAG. You ask your question in plain language; it searches your emails, calendar, documents and knowledge base, then answers and shows its sources.

You build the knowledge base yourself: PDF, Word, Excel, PowerPoint, text files or notes. If an answer falls short, you can suggest a correction, which is saved. The connection to your inbox is read-only, and data is hosted in the EU.

It's the foundation of the AI personal assistant Lever aims to be: an assistant that knows your files, rather than a chatbot that guesses.