OpenAI for Email Parsing - Extract Unstructured Data

Coming over from Zapier I have been very happy but one feature I’m struggling with is how to nicely work with OpenAI and I don’t mean for ChatGPT functions.

The most useful (and cost effective) I have found is pulling unstructured data using OpenAI. In my few months of using it I have not had a single case of error even when the input format changed.

Has anyone tried to make this work with Make? I’m sure it can be done since you can do it with code but I don’t quite have those skills.

I sent this over to Make support to show them some of the crazy things that can be done safely with OpenAI since by default data sent this way isn’t used for training.

Appreciate any feedback or suggestions.

J

Welcome to the Make community!

Using OpenAI is definitely possible, but may not be the best use-case for your scenario.

Could you provide examples of “Unstructured Data” and the output you desire to achieve?

2 Likes

Hey there, for sure! I’m just parsing an inbound email to get data from a legacy system that can email out updates.

So I am pulling the account number, shipment status and some other information.

Basically I feed the inbound email body to the api then list what I want. I suspect mine works perfectly (literally never had it not get a status wrong) because the emails are machine generated.

The example in the link I sent is for a lead email, this is an amazing use case that companies like Salesforce are using to get data into a CRM or other structured data from emails a customer or prospect might send.

The API request in that case looks like:

Parse the contact details from the lead.

Return as fully structured JSON object like this:
{
"NameFirst": "",
"NameLast": "",
"Email": "",
"PhoneNumber": "",
"Company": "",
"Website": "",
"Comments": ""
}

Content: {{CONTENT}}

If you want to go the ChatGPT route, you can either use watch emails (scheduled) or setup a mailhook (instant):

2 Likes

Thanks for the response - I wasn’t clear on what to fill in where to make this work.

See attached, but there is a bunch of data I used here, I assume this is just about formatting it. I tried to find other examples online but struggled.

Make doesn’t have such a module like Zapier’s “Extract Structured Data”, but you can do a little bit of prompt engineering to get the Make “Create a Completion” module to work similarly.

Give it a try with the above and let us know if you get stuck!

You can also suggest for it to be made in the Idea exchange. Don’t forget to search for it first, in case someone already suggested it!

2 Likes