FIELD NOTES / 02 · APPLIED AI
Your knowledge.
Within reach.
What is RAG—and how does it use your knowledge? Think of an assistant with a reference shelf: it finds relevant pages before composing an answer.
APPLIED AI / HOW RAG WORKS
Your knowledge.
Put to work.
RAG stands for retrieval-augmented generation. It finds relevant information in a selected knowledge source and gives it to an AI model as context for answering a question.
- 01 / ASK
Your question
Start with a question about your business or SAP processes.
- 02 / RETRIEVE
Find the context
Retrieve relevant passages from selected documents and reference material.
- 03 / GENERATE
Connect the information
The AI uses your question and the retrieved context to compose a response.
- 04 / ANSWER
A grounded response
An answer informed by the reference material, with sources when configured.
I apply RAG to make specialist knowledge easier to access—from SAP reference material to business process documentation. The quality of the sources and retrieval matters; answers still need appropriate review.
Discuss your AI use caseStart with the question
RAG means retrieval-augmented generation. A search step finds relevant reference material, and an AI model receives that material alongside your question. It can then compose an answer using the supplied context. It is a way to connect a model to information outside its built-in knowledge.
This combination of retrieval and generation is described in the original RAG research paper.
A small business example
Imagine asking: “Which checks do we follow before releasing a billing change?” An assistant searches the documents you have selected, finds the relevant checklist and uses it to draft a response. If source references are configured, you can follow them to check the original passage.
This is an illustrative example, not a claim about a deployed client system. A sensible first prototype would use a small collection of approved documents and a set of questions with known answers.
What the model actually sees
The application usually sends selected passages rather than an entire document library with every question. Preparing content, retrieving useful passages and enforcing access permissions are distinct parts of the system. Updating a reference source and making it searchable can be separate steps.
Retrieval can use keyword, vector or hybrid search. RAG does not, by itself, mean retraining the model on all your documents. Microsoft’s RAG overview explains these building blocks.
Better context still needs judgement
An answer can be wrong if the source is outdated, the wrong passage is retrieved or the model misinterprets it. A citation is a starting point for checking, not proof that the conclusion is correct. Permission checks must also prevent users retrieving documents they are not allowed to see.
For SAP configuration or a business decision, verify important details against current official documentation and test in the appropriate environment. Choose model hosting and document access arrangements deliberately before using confidential material.
Where RCA applies it
RCA’s primary focus is enterprise and SAP consulting. Practical RAG prototypes are a complementary capability: a way to make specialist process knowledge and internal documentation easier to access.
Explore typical consulting engagements or email Jay about a knowledge-AI prototype.