Free YouTube Transcribe

Video transcript

How To Build A Fully Automated Customer Support Agent with n8n

Max Mitcham · 1,211 words · 6 min read

Want to search this transcript, jump the video from any line, or download it as TXT, SRT, or VTT?

Open in the transcript tool

Full transcript

0:00What is going on everyone? I wanted to

0:02go over one of the agents that I have

0:03just finished building and is currently

0:06whirling away for me right now running

0:09every single day, nearly every minute of

0:11the day at the moment, which is quite

0:12cool. And this is what I call my low

0:16usage agent for customer success. So

0:20before I dive into the complexities of

0:22this uh kind of monster looking agent, I

0:25want to go over high level what this

0:27thing is doing. So first of all we have

0:31segment and postto. So for those of you

0:34that don't know segment is like a way of

0:36um transferring data between systems and

0:39we have all of our data coming in from

0:42segment through to atio and our data

0:45being anyone who signs up on the

0:47platform. So it's any users that we have

0:50inside of the kind of triggery e

0:52ecoscope is coming in through segment

0:55from our database and from postgress

0:57through into

0:59also then using postog which is our

1:01product analytics um

1:04uh kind of uh system to then push

1:07product analytics into

1:09as well. Super super seamless super

1:12super easy to do. From there, we then

1:15are using Atio basically just to store

1:17all of the data telling us the features

1:19that they've used, when they last logged

1:21in, all of that fun stuff which we're

1:23getting through from segment and post

1:25hog.

1:27So once we have that, we're then using

1:29NA10 to start to query ATIO essentially.

1:34And what NA10 is going to do is it's

1:36going to pull through individuals from

1:39Atio that haven't logged in in the last

1:4210 days. Then what it's going to do is

1:46it's going to go and look up these users

1:48inside of Postto to get more of a

1:50comprehensive overview than what we

1:53currently have inside of Atio. From

1:55there, we're going to use Claude to

1:57basically start to compile information

1:59based on how they've used the platform

2:01and use cases that they could use on the

2:03platform before we send them through to

2:05Loops, which is our marketing uh

2:07automation tool that we use and then

2:09we'll update Atio. So, let's dive into

2:12this uh system going through it. So, the

2:16first thing that we have is we basically

2:19query ATIO. So this is using Atio's API

2:22and what we're doing is every day we

2:24basically go and look and we say hey go

2:27find me any users that haven't logged in

2:30in the last 10 days. And so in this

2:32instance there's 77 users uh who are

2:35using triggery that haven't logged into

2:38the triggery platform in the last 10

2:41days. So what that basically means is

2:42like they haven't uh triggered what we

2:44call like a page view event um and uh

2:48therefore they've come up here. So then

2:50what we're doing is we're going into

2:52these users and getting more details. So

2:55we go get the their initial information

2:56via post hog and then what we start to

2:59look at is we look at the different

3:01events inside of post hog that mean

3:04something to us. So these are

3:05effectively for for for those of you who

3:07don't know what events are, these are

3:09effectively the different features

3:10inside of triggery that someone could or

3:12couldn't use as well as looking at more

3:15of like holistic overview with these

3:16page views. So we can see everything

3:18that they have or haven't done.

3:20What then happens is if we dive into

3:23this one a little bit deeper, you can

3:24see, okay, well this person so far has

3:26used this event 162 times. They've used

3:30the uh sync event six times. have not

3:32used social signals and they've done 15

3:35active days since they've logged in or

3:38since they created their triggery

3:39account. And then we can see the most

3:41visited pages, all of that fun stuff. So

3:44what then happens is we go through to a

3:46series of events uh or agents I should

3:50say. And the agents are doing a few

3:53different things. So the first agent is

3:55effectively our user and company

3:57research agent. So this is going to use

4:00uh funny enough the 4.5 preview search

4:03or sorry the four preview search from

4:05GPT to search the web and understand all

4:08the information that it can find on that

4:10person and on that company. It's then

4:13going to feed that through to the usage

4:16agent which you can see is now running.

4:18And this usage agent is basically going

4:20to take all of this data from Postto and

4:23start to understand what they have or

4:25haven't done on the platform. And it's

4:27going to chat to this triggery knowledge

4:30hub. This is another agent that I've

4:32created which is a trained agent on all

4:34of the features and their use cases and

4:37how they use them inside of Triggery.

4:40And what it's going to do is it's going

4:43to create a use case for that individual

4:47that they haven't already done yet based

4:49on reading everything that they've done

4:50inside of the platform. So an example

4:53would be like it's like hey you've not

4:54used this feature here's how I would

4:56think about using it and here's how I

4:58would use it. Because what then it's

5:01going to do is it's going to feed it

5:02through to this email agent which is

5:04then going to start to construct the

5:06actual uh uh kind of uh email itself.

5:10And so for example, you can see here

5:11it's like one monitor competitor

5:13engagement to find high intent

5:14prospects. Go to the engagement tab. Add

5:17a competitor URLs like Jasper AI and

5:19copy AI. Export engaged prospects to

5:22your CRM. Track buying intent through

5:24social signals. Click on the social

5:26signal setup in the left navigation.

5:28Upload your target prospect list.

5:30Monitor marketing workflow AI automation

5:32signals. And then lastly, identify pain

5:35points via our topic search. Um add

5:37keywords for marketing automation

5:39challenges. And then track discussions

5:41around content production bottlenecks.

5:43So it's super super specific to that

5:46user in terms of an actual use case on

5:49how they should be getting started or

5:51re-engaging with their Triggery

5:52platform. And then what it's doing is

5:54it's firing this over to the email

5:56marketing system that we use which is

5:58loops. And then it's also updating

6:01uh atio to say hey we just sent them an

6:04email and it adds a time stamp because

6:07one of the things that we're doing back

6:08here is when we query ato we are also

6:13making sure um that we're not emailing

6:16anyone that's been emailed in the last

6:1810 days. I.e. if this usage email goes

6:20out to them they still don't log in. we

6:21don't want to hit them the next day and

6:23the next day and the next day. So, there

6:24is at least a buffer between that. And

6:27this is uh what I think an awesome

6:31awesome agent to help me when I've got a

6:33a small team to stay on top of our

6:36customer base and continue to re-engage

6:38and engage with our customer user base

6:41with a highly personalized element. Um

6:44looking at their current like usage and

6:47data uh perspective.

This transcript was generated from the captions YouTube publishes for this video. Get the transcript of any YouTube video atfreeyoutubetranscribe.com: free, unlimited, no sign-up.