Zapier
      Back to home
      1. SalesAi Help Center
      2. Integrations
      3. Zapier

      Utilizing Chat GPT and Zapier to Perform Post-Call Transcript Analysis

      Follow this guide to extract specific information from your call transcripts to add to your CRM/App

      Please make sure you have set up the Connection between SalesAi and your CRM/app utilizing Zapier! Here is an article on that process if you have not begun:

      Outbound Webhooks (via Zapier)

      Sections:

      • Use Cases
      • Setting up the Chat GPT step

      • Adding a Balance to OpenAI

      Use Cases

      Utilizing Chat GPT to analyze the call transcript sounds pretty broad initially. This is on purpose because there are really SO MANY options with this step in Zapier. Here are a few common use cases for this that the integration team encounters often!

      • Extracting answers to questions from the transcript and putting them into custom properties
      • Determining Sentiment
      • List out all the contact's pain points
      • Extracting the meeting date and time that was scheduled
      • Determining if they wanted to receive a follow-up email with resources
      • Determining if the contact consented to SMS

      As you can see above, there are some pretty different options to utilize. There are plenty more, since we are utilizing a Chat GPT prompt, so please don't limit yourself to these!

      Setting up the Chat GPT step

      • Open up your existing Zap that has your integration between SalesAi and your App. It should look something like this:
      • Click on the plus in the middle of those two steps and look for Chat GPT (OpenAI)
      • Once Selected, we will be using the Extract Structured Data Action Event
      • Click Select in order to attach an Account. You should see a new window open. Click on the Create an API key link.
      • Make sure to Log in or Sign Up!
      • Click on Create new secret key

      • You can set the name of the key to be SalesAi. Set the project to default, or use an existing project if you have one. Click Create secret key
      • Copy your API key and go back to the open window for connecting your account:
      • Paste in your secret key, and skip the organization ID. Click Yes, continue to ChatGPT (OpenAI)

      • You can now continue to the Configuration part of the Chat GPT step
      • Click on the Plus Symbol inside of the Unstructured Text field
      • Search for the Transcript Text and click it to add to the box.
      • Leave the model as 4.0 mini
      • Think of the Description box as your prompt. This will be how the Chat GPT step knows what data to extract. Best practice is to do the following:
        • Start your prompt with "Analyze the transcript to determine..." or "Analyze the transcript to extract..."
        • Refer to the contact as the user.
        • Make sure to specify instructions for determining specific pieces of information if necessary
          • Ex. Specifying all the parameters on what makes a lead "qualified".
      • The next section is Values to Extract. You will want to have a value for EACH section you are pulling. You will need to type each one and hit Enter to move to the next. Here is an example based on the prompt I had above:
      • For each Value added, there will be more fields to fill out. Typically, you will only need to answer the Is this required? field and the Data type field. I usually always put these answers for each:
        • Should {field name} be required?: True
        • What data type is {field name}: String
      • Make sure to answer these questions for each field you created!

      If the question has limited possible answers, you will need to also answer the What are the possible values for {field name}? question. 

      Example: Did the user consent to receive a follow-up email?

      Possible values:

      • Yes
      • No
      • Once all of these steps have been filled out, you can click Continue
      • You may now click Test Step to see the results!

      If the test fails with an insufficient quota message, you will need to add a balance to your OpenAI account if you have not already! Steps are located at the bottom of this article!

      • If your answers are not correct, you may need to be more specific in your Description!

      Adding a Balance to OpenAI

      In order to run API callouts to Chat GPT, we will need to have a balance in Chat GPT. Here are the recommended values

      • Starting Balance: $10
      • When dropped to $5, return to $10 (Auto-recharge)

      Follow the steps below to add this balance:

      • Go to the OpenAI billing page for your account: https://platform.openai.com/settings/organization/billing/overview
      • Click Add payment details
      • Enter your payment information
      • Confirm the initial charge, and then you can test your Chat GPT step!