# Custom AI Models: Prompting, RAG or Fine-Tuning? A Guide

**“We want our own AI model.” That usually means one of three things with very different costs and timelines. Which one is yours?**

> Most businesses don’t need to train a model from scratch. The right order is: better prompts and instructions first, then RAG to answer from your own documents, and fine-tuning only if you need a fixed format or behaviour. Choose by measuring accuracy on real examples, not by guessing.

Author: Saman Ghahreman · Published: 2026-10-08 · Updated: 2026-10-10 · Canonical page: https://solocodelab.com/en/blog/custom-ai-model-guide/

“We want an AI model of our own.” We hear this often, but it usually means three different things with very different costs and timelines. This guide helps you work out exactly what you need.

## Four ways to “customise” AI

| Approach | What it solves | Cost and time | When? |
| --- | --- | --- | --- |
| Better prompts and instructions | Format, tone and answer quality | Very low; days | Always start here |
| [RAG](https://solocodelab.com/en/blog/how-rag-chatbot-works/) | Answering from your documents and data | Medium; weeks | When the model must know your knowledge |
| Fine-tuning | Fixed, repeatable behaviour, format or style | Higher; needs quality examples | When prompting and RAG weren’t enough |
| Training from scratch or classical ML | Forecasting and scoring on structured data; training a language model from scratch is rare | High | Only for special needs with plenty of data |

## The recommended order: simple to heavy

OpenAI’s official guidance on improving accuracy suggests the same path: start with a prompt, add fixed examples, then a retrieval step (RAG), and finally, if needed, fine-tuning. These methods can be combined. For example, one assistant can be fine-tuned and also answer from your documents.

## RAG or fine-tuning? A simple rule

- If the problem is “the model doesn’t know something” (prices, policies, technical documents): RAG.
- If the problem is “the model doesn’t behave correctly” (output format, tone, fixed sorting): fine-tuning.
- If your knowledge changes often: RAG, because you only update the knowledge base and don’t retrain the model.
- If you have both problems: combine them.

## What data do you need?

For RAG, your existing documents are a good start, as long as they are organised and current. For fine-tuning you need correct input-and-output examples. OpenAI’s guidance suggests starting with at least about 50 high-quality examples. Quality and variety matter more than the raw number.

## How do you measure accuracy?

1. Collect 30 to 100 real questions or tasks from your daily work, including hard cases.
2. Define the correct answer or pass rule for each.
3. Test and compare every approach on this same set.
4. Sort the errors: not knowing, making things up, wrong tone.
5. Before launch, review the [responsible-AI checklist](https://solocodelab.com/en/blog/responsible-ai-2026-business-checklist/).

## What if the model must also take actions?

When a model must work with tools and move through several steps, you enter the world of [AI agents](https://solocodelab.com/en/blog/ai-agents-2026-practical-business-guide/), where guardrails and access control matter even more.

Before deciding, check [what a chatbot and AI model cost](https://solocodelab.com/en/blog/chatbot-and-ai-model-cost-guide/) so your choice fits your budget.

Reduce the risk of building a model with [a proof of concept before payment](https://solocodelab.com/en/blog/show-live-sample-before-you-pay/).

For a real example of AI running on the device itself, see [offline face-recognition attendance: how Hazeri works](https://solocodelab.com/en/blog/hazeri-offline-face-attendance-app/).

> If you’re unsure which path fits your problem, [AI model development](https://solocodelab.com/en/services/ai-model-development/) starts with a simple prototype and a measured accuracy, before you spend on a heavier route.

## Frequently asked questions

### Should we train a language model from scratch?

Almost never for an ordinary business. The cost and data needed are huge, and prompting, RAG or fine-tuning an existing model usually reaches the same result.

### Can RAG and fine-tuning be used together?

Yes, and it’s common: fine-tuning fixes behaviour and format, while RAG supplies current, company-specific knowledge.

### Can you run the model on our own server?

If data privacy matters, open-source models can run on your server. The choice between an outside model and your own depends on accuracy needs, cost and how sensitive the data is.

### How do we know the model works well?

With a test set of your own real tasks. Compare results before and after each change and sort the errors.

## Sources

- [Optimizing LLM accuracy](https://platform.openai.com/docs/guides/optimizing-llm-accuracy) — OpenAI
- [Fine-tuning LLMs: overview and guide](https://cloud.google.com/use-cases/fine-tuning-ai-models) — Google Cloud
- [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) — NIST
