/answers/rag-vs-fine-tuning/index.md · 42 lines
---
title: "RAG vs fine-tuning: which does your AI project need?"
description: "Retrieval-augmented generation (RAG) looks up the relevant passages in your own documents each time a question is asked and gives them to the model, so answers can cite their sources and stay current when the documents change. Fine-tuning adjusts a model with many labeled examples so it follows a format, tone or narrow task more consistently, but it is a poor way to teach a model facts that change. Most business assistants that answer from policies, manuals or records start with RAG; fine-tuning is worth considering later, when there are enough good examples and a clear behavior that prompting and retrieval cannot get right."
url: https://lidvizion.ai/answers/rag-vs-fine-tuning/
site: Lid Vizion
index: https://lidvizion.ai/llms.txt
---
[Direct answers](https://lidvizion.ai/answers/)
# Should a business connect an AI model to its own documents with retrieval (RAG), or fine-tune a model on its data?
Retrieval-augmented generation (RAG) looks up the relevant passages in your own documents each time a question is asked and gives them to the model, so answers can cite their sources and stay current when the documents change. Fine-tuning adjusts a model with many labeled examples so it follows a format, tone or narrow task more consistently, but it is a poor way to teach a model facts that change. Most business assistants that answer from policies, manuals or records start with RAG; fine-tuning is worth considering later, when there are enough good examples and a clear behavior that prompting and retrieval cannot get right.
## What this means in practice
- Use RAG when answers must come from your own, changing documents.
- Use RAG when people need to see the source behind each answer.
- Consider fine-tuning for a consistent format, tone or narrow task.
- Collect enough good labeled examples before fine-tuning.
- Start with retrieval and prompting, and test before adding fine-tuning.
## How Lid Vizion approaches it
We start with the business decision and operating context, define ownership and measurable milestones, then use configuration, integrations, automation, or custom software only where each earns its place.
## More on Automation & custom AI
- [When should a business automate a workflow with a no-code automation platform, and when does it need custom code?](https://lidvizion.ai/answers/no-code-automation-vs-custom-code/)
- [How can a support team use AI to find the questions its help center does not answer well and fix them first?](https://lidvizion.ai/answers/how-to-find-help-center-gaps-with-ai/)
- [How can a subscription or services business use AI to notice customers who may be about to leave while there is still time to help?](https://lidvizion.ai/answers/how-to-spot-at-risk-customers-with-ai/)
[Compare services](https://lidvizion.ai/services/) · [Use a planning tool](https://lidvizion.ai/tools/) · [Explore workflow use cases](https://lidvizion.ai/use-cases/) · [More direct answers](https://lidvizion.ai/answers/)
Talk it through
## Have a version of this question inside your company?
Bring the decision, the constraints, and what a good outcome looks like.
[Talk through the decision](https://lidvizion.ai/contact/?need=ai-workflow&from=%2Fanswers%2Frag-vs-fine-tuning%2F#write) · [Book a call](https://calendar.app.google/bB26LdA9e2gmGFp69)