Free YouTube Transcribe

Video transcript

How I Built An AI Agent To Analyse My Cold Emails (Super Easy)

Max Mitcham · 1,241 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's going on guys today we're going

0:01to build a new agent this time I want to

0:05understand Trends and Analysis on my

0:08Smiley campaigns but this can be

0:10applicable for instantly reply kind of

0:12like Outreach based platform as long as

0:13you can export the data and the

0:15responses um via CSV so what I've gone

0:19ahead and done is we've gone to SM lead

0:21we've created a campaign we've done the

0:23campaign and then we've gone to the

0:25Inbox and exported the inbox with all of

0:27the campaign sequences and and statuses

0:29and things like that from there we then

0:32uploaded this into here so we can see

0:34all of the uh data basically coming back

0:38the reply time reply message things like

0:39that so now what we want to do is we

0:41want to create an agent so um as usual

0:44we're going to go to here and we've got

0:46a few different options right we could

0:50use the tools agent we could use the

0:54plan and uh execute agents which could

0:58be probably the one that we might go for

1:00or the conversational one for these

1:02these different ones for this one I

1:03think we're just going to keep to the

1:05tools in fact we're going to use the

1:07conversation we're going to test the

1:08conversation on we can maybe even do a

1:10little bit of a test to see how they

1:12differ actually uh for the purpose of

1:14this so we're going to use 40 mini for

1:17the purpose of this but for the test um

1:20but I would probably use 40 because

1:21you're going to get better outputs or

1:23maybe even

1:25clawed we're going to use a memory uh

1:28here and the memory we going to take

1:29change it to a context context length

1:32window of 10 so now I created that

1:35Google sheet I think it's called

1:36influence campaign report so we're going

1:39to do this as

1:41well now one thing that I thought could

1:44be cool is we're going to get the email

1:46address so we could also take this a

1:50little bit further by understanding who

1:52are the types of people that have

1:54replied but for now let's go ahead and

1:59chat to it a previous session um let's

2:02change that can you spot any

2:07Trends on this data based on the

2:14individuals that have replied to

2:20me anything I could use going forward to

2:25improve my response rate

2:31probably conversation agent oh I use

2:34conversation agent oops didn't mean to

2:36do that I meant to use plan execute

2:39agent okay redo

2:42that the tools are not supported what

2:46tool is not

2:47supported uh that should be fine

2:51um okay there okay that makes our answer

2:55I never knew that so we got to use a

2:56tools agent for this purpose for Google

2:58Sheets clearly

3:01let's going test that there we go okay

3:03so that is now going to start to to run

3:06um this might take a while because it's

3:08got to Crunch a lot of data so we may

3:11skip ahead to the results coming back

3:13okay we now have a response based on the

3:18data you provided regarding individuals

3:20who engaged with your AR messages here

3:22are some Trends General Trends Laris

3:24bonray mostly message show no replies

3:26delivery issues interesting balce rate

3:30blah blah blah you have SRU to emails

3:32consider personalization so it's really

3:34ripping apart my campaign anyway I want

3:37to take this further because this is all

3:39quite generic we haven't done AB testing

3:40obviously this would be a lot better if

3:42I did AB testing things like that I want

3:44to really understand based on those

3:46individuals that have replied what were

3:49the themes right because if the copy is

3:51relatively the same and this would be

3:53even better if the copy varied based if

3:55I was using like AI copy personalized

3:56for everything we could then like really

3:58like hone in on it but here I've got two

4:00other nodes that I just copied and uh

4:03pasted from another flow so we don't

4:06need to read it every time and we're

4:08going to hook these up so this is going

4:09to get the information of the

4:11individuals from social media and this

4:12is going to get the information it's

4:13going to allow us to search and so now

4:15what we want to do is I want to Define

4:18The Prompt

4:19um uh give it assistant message you're a

4:22helpful assistant helpful uh research

4:30assistant that is going

4:33to crunch numbers for me to improve my

4:38cold email campaigns going

4:44forward you have the ability to search

4:50Google and social media

4:55to

4:57understand if there are any Trends

5:01emerging based on who the people are

5:06that are replying or

5:10not and can check people's backgrounds

5:14via social media

5:18Etc which is good and so now we're going

5:20to say Can you spot any

5:24Trends on the type of people that have

5:28replied to me

5:30in this

5:33campaign you can use the Google search

5:37or social media info to help

5:42inform your

5:46decision is there anything the people

5:51have the people that have replied have

5:55in

5:57common okay we're going to let this run

5:59can analyze their background com

6:02trades okay do I need to update this set

6:06manually read to update to the Google

6:09Sheets to

6:11find the called email campaign

6:16performance use the Google Sheets tool

6:20to get the campaign

6:25stats okay use now because I've used the

6:28window buffer we can actually

6:30have a conversation with it we don't

6:31need to restart the pomp use the

6:33Google Sheets tool to

6:39collect all the uh information you

6:45require okay everyone right we've

6:47finally replied to us which is good so

6:50you can see here spoke to it grabbed all

6:52the information it's checked different

6:55profiles so it should spit us out quite

6:58a nice yeah quite a nice response let's

7:00see what we've got so common

7:02characteristics May respondents were

7:05senior individuals industry exposure the

7:08majority come from technology

7:09backgrounds okay we've got the country

7:13educational we got a few of the notable

7:16highlights so here we go we've got four

7:18actionable insights focus on decision

7:20makers fine interesting so highlight how

7:22you can your product can assist and

7:23optimize marketing efforts given the

7:25interest expressed by many enhancing

7:26their digital presence and efficiencies

7:29consider case study development

7:31testimonials to resonate with

7:32high-profile

7:33audience

7:36interesting we could then ask it based

7:39on this data you now have is there any

7:45Trends with we could even do it like

7:47with the post that they do or something

7:48like that that could be cool because

7:49we've hooked this up to triggery now but

7:52uh we could say basic on this data is

7:53there any Trends uh with the work

7:58experience Maybe previous roles or

8:01anything like

8:03that okay so we need to go get it again

8:07so that is an agent right which now is

8:10taking all of our campaign data it's

8:15we've armed it with the ability to sech

8:16Google we've armed it the ability to

8:18pull a social data via

8:20triggery and we can just keep asking it

8:22questions right and and we can keep

8:24tinking with this to make this better

8:25but that is another use case of a

8:28individual agent that can get so much

8:31value and help you spend hours and hours

8:33of time just by improving your campaign

8:36performance

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.