What a RAG can do

A RAG system makes knowledge that is often spread across large storage areas easily accessible. This makes it ideal for internal knowledge management. Instead of searching through handouts, emails, or notes, internal organizational data can be specifically integrated and made available in natural language.

This allows team members to ask questions and receive answers that come directly from the organization's knowledge base. It creates clarity, saves time, brings existing knowledge to life, and makes it usable without creating new structures or data silos.

What is a RAG?

A RAG (Retrieval Augmented Generation) is an AI (artificial intelligence) application in which a language model (LLM) is linked to a knowledge base—for example, a specific data collection from an organization.

Before this link is established, the language model only has the knowledge with which it was originally trained but does not yet know any specific information from the organization. By connecting to its own (vector) database, the language model can search there for relevant information (retrieval), combine it with its own knowledge (augmented), and then generate a response (generation).

Our handbook will help social-impact organizations set up their own RAG system in compliance with data protection regulations. It includes a brief introduction and step-by-step instructions for your convenience.

The system's GDPR compliance in the described structure is achieved by the deliberately chosen components: There is no need to transfer data externally; everything can be hosted in-house, so only the server needs to be located in an appropriate place.

The documentation is five pages long and available in German below.

Enjoy setting up your own RAG!

We created our own RAG and its documentation with the support of zukunft zwei and N3XTCODER as part of the Civic Coding initiative.

Contact

Do you want to join the discussion about AI in social-impact organizations?

Tobias Gerber
tobias.gerber@iac-berlin.org