This article contains info about using ChatGPT or the Azure OpenAI suite as a way to answer questions from a specific data set, e.g. company documents. 

This model could be thought of as a pseudo DB where the user could "query" the db by asking the model questions about the data it got trained on.

> "What are the top 10 sales agents for North America for December?"

> "Please list all conversations where someone mentioned that the coffee was cold."

One way to do this is by creating a [[https://openai.com/blog/introducing-gpts|custom GPT]] which will answer questions based on data you add yourself.

Yes, Azure OpenAI can be used in this manner as can be seen in the link above.

It’s unclear whether you //should// use it this way. There are still sources that suggest ChatGPT might not be accurate even when external data is uploaded to a custom GPT.

> [[https://zapier.com/blog/custom-chatgpt/|It’s worth mentioning that my chatbot didn’t consistently refer to the guide I uploaded to teach it how to write alternative text.]]

If you’re interested in trying out OpenAI in this manner, the article from Zapier above has detailed steps on how to do it. In summary, you would:

  - Use “Create a GPT” option in the paid version of ChatGPT.
  - Instruct it that you’re creating a company reference GPT based on company documents.
  - Upload documents using the upload Knowledge function in the custom GPT configuration page.

It’s worth noting that you cannot exclude all other data that chatgpt uses, because it is used to help chatgpt understand input and write answers ([[How ChatGPT Works]]).