How to set up Gong Automation? AI reads transcript and summarizes it

Hi! I am trying to setup this automation that was featured by Make a month back that does the following:

  1. Gong watches for new calls
  2. Gong fetches the audio transcript
  3. Iterate and Aggregate the text into Audio
  4. AI Summarizes it
  5. That summary is posted in Slack

My question is how am I supposed to grab the call media and have it aggregated into audio?

The example that is highlighted in post that Make made can be found here: Make on LinkedIn: Uncover market intelligence with Gong

I would love some help on setting up the Gong, Iterator, and Tools modules! I think I can figure out the AI and Slack modules once those are setup.

Any help is appreciated! Thank you!

When asking your question, please include:

:footprints: The steps you have taken
:camera_flash: Relevant screenshots
:link: Any links you have
[ Code { "and": "JSON", "in" : "code block"} ]
:x: Exclude Personal Information.

1 Like

Hey Ian, thanks for the question!

A detailed, step-by-step tutorial on this was published in our use-case repository - this should tell you everything you need to know.

If you’re looking at wider use of AI, you might also want to explore using Eden AI as a front-end to your preferred AI solution. They provide a single unified API for multiple AI tools, and a single point of billing. It makes it really easy to compare the results from different engines. There’s a Make app too!


Hey @DavidGurr_Make !

With the current published tutorial - I am having a hard time aggregating call transcripts.

I have used the template and the given JSON, however, it appears that only the first sentence of the call is making it through.

I checked the call that Make is referencing and it is 22 minutes long - so I know it isn’t the call that is the problem.

Can you let me know if the tutorial or my automation has an error that is causing this to happen? I am sharing the blueprint.json but also a few images of the settings for each module.

Any help is greatly appreciated!

Screen Shot 2023-08-15 at 10.32.44 AM
Screen Shot 2023-08-15 at 10.32.32 AM

blueprint.json (25.8 KB)

I can’t see any obvious error, but I don’t have any Gong call records in our sandbox to test with.

If you run the scenario, you can then click on the “speech bubbles” to the top right of each module to see exactly what it returned. Take a look at the results of the Make an API Call module, the Iterator, and the Text Aggregator.

If nothing jumps out to you, copy and paste screenshots from each of those and post here.

Thanks David! Nothing stands out so I added some screenshots of the bubbles.

It looks like the Iterator is only picking up one sentence from the Make an API call module.

What do you think?
Screen Shot 2023-08-15 at 11.29.14 AM
Screen Shot 2023-08-15 at 11.28.57 AM

Actually, when I open up the Call Transcripts Array in Bundle 1, I can see all the bundles from the transcript - yet when I go to the Iterator - I only have the option to select sentences, which is returning only one sentence - not the rest of the Array.
Do you think there is a way around that?

Yes, that’s what it looks like.

For the Make an API Call module, expand the callTranscripts element all the way down.

We should see a transcript array within that, containing an array of sentences.

Yep! I can see that.

Screen Shot 2023-08-15 at 12.50.16 PM

But when I select sentences for the iterator, see attached, it is only returning 1 bundle in the array.

Screen Shot 2023-08-15 at 12.52.06 PM

I think it has to do with the transcript from the API call not making it to the iterator for some reason.

Screen Shot 2023-08-15 at 12.52.14 PM

I am not super sure if there is anything else that you are able to do without maybe jumping on a call and showing you.

Unless you can think of something else, would you mind letting me organize a call with you?

Thank you so much.

Hi Ian.

It looks like there may be an error in the original use case.

The Gong API documentation shows (under Response, if you expand out each “…”) that the callTranscripts array is one entry per call, each containing a transcript array of one per monologue, which in turn contains the sentences from that monologue.

The word “monologue” is important here - there’s a new transcript every time the speaker changes. The example in the use case only picks up sentences from the first monologue, which in your case is just “Can you hear me?”.

What you’ll need is an additional Iterator inserted between the Make an API Call and existing Iterator modules.

The new Iterator will process the transcript array:

The old Iterator now processes the sentences from the new Iterator:

And finally, the Text Aggregator needs to know to aggregate everything from the new Iterator:

If that works, do let me know and we’ll get the published use-case updated.



Thank you so much David - this is so helpful for our team!

So glad we were able to get this working.

I appreciate all of your help!