Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:01
I'm going to show you how I created an
0:03
AI course for marketers. All of
0:05
it done using AI. I'm going to go
0:07
behind the scenes. I'm going to show you each
0:09
and every lesson. I'm going to give you
0:11
some advanced prompts. I'm going to give you some
0:14
really cool AI tips, and I'm going to give
0:16
you the summary of how I felt AI
0:18
performed in creating this course for me so you
0:20
can use AI to create things like this
0:22
for you. All of this and
0:24
much more on this episode of Marketing Against the Grain.
0:27
I'm Kieran Flanagan, co-host of Marketing Against the
0:29
Grain. Let's get into today's episode. We
0:33
write back to today's show, but first here's a
0:35
quick word from HubSpot. Teams everywhere are dealing with
0:37
a serious problem, a productivity problem.
0:41
Don't worry, HubSpot's customer platform can't
0:43
help. No more time-consuming
0:45
tasks. No more chasing down prospect
0:47
info. No more one system for
0:50
this and another system for that. HubSpot
0:52
helps you find leads, reach prospects, and
0:54
deliver the insights to convert them into
0:57
customers all in one place. Using
0:59
your productivity problem, not a problem at
1:02
all. Visit hubspot.com to
1:04
learn how they can help you grow
1:06
better. It's
1:20
a little bit of a special episode. It is
1:22
a Kieran Solo episode, and what I'm going
1:24
to do is cover a short email course
1:27
I put out. It doesn't sound
1:29
that interesting. Stay with me. I
1:31
know it's just a short email course. Everyone who listens
1:33
to this has probably released some form of
1:35
an email course over the past decade. The
1:38
reason we're going to cover this in
1:41
a short Solo episode is because it
1:43
was, for the most part, created by
1:45
me and AI, and
1:47
it was all about how to bring
1:49
the timeless craft of marketing together with
1:51
AI and have AI instrument that timeless
1:53
craft for your brand. Now, let me
1:55
give you a bit of the backstory.
1:57
Why did I decide to create this?
2:00
The A discourse. First of all, You.
2:02
Should always have little aside things that you do.
2:04
It is the best way to learn. I don't
2:06
care where you are in your career if you
2:08
don't really like doing the stuff do when the
2:10
word do in the craft like I shipped. This.
2:13
Little course yesterday. put a link and post
2:15
op of money than one that that a
2:17
promo about somewhere to go this week and
2:19
like this. but four hundred and fifty five
2:21
hundred people already signed up. That's cool. I love
2:23
that like one linked in post, people seem to
2:25
enjoy the course and you just watch the little
2:27
conversions come in on the hotspot platform and you
2:30
can see people I see some of your course
2:32
you know you create. Something that is the craft
2:34
and then people. Actually find invaluable. That's why I
2:36
love doing what I love and I think at
2:38
times you get really removed from that right? You
2:41
actually end up become and a people manager and
2:43
that's all you really do as boxes around and
2:45
move budget of nine different places. I actually enjoy
2:47
doing the work I want to do the craft
2:49
and so this is really cool. from the I
2:51
was up at twelve o'clock last night to put
2:54
the linked in post up at midnight. Like refreshing
2:56
the former: How many people have signed up and
2:58
I've got a bunch of promo? Go and Live!
3:00
So my goal was to get two thousand people
3:02
signed up for that. So little bit about why
3:04
I think. These gonna little side things are
3:06
really important. It's really how I learned I
3:08
have to actually do the thing. Aside from
3:10
that, there is a couple of reasons I
3:12
decided to do that. and so recently I
3:15
put out a post on Linked In that
3:17
blew up I've seen occurred lots of times
3:19
since then. Would says a quote attributed to
3:21
Sam Altman said that he thinks that a
3:23
I in the future will be up to
3:25
automate ninety five percent of a marketers role.
3:27
Fun! Episode Com: Enough Franciscans thinks
3:30
the opposite is true. A I will
3:32
only be able to automate maybe five
3:34
percent of a marketers role and we're
3:36
going to debate topics I think it
3:38
can be much closer to maybe ninety
3:40
five percent, not ninety five percent. With
3:42
the closer than random assortment debates up
3:44
for this quote was attributed some outman
3:46
a blue Optimizers concede. Are tons
3:48
and tons of engagement. Like I think
3:50
it was like seven hundred different commons.
3:52
And what I noticed in the comments
3:54
was more Martyrs really fell into three
3:56
categories they sell into. A I
3:58
as only heights. I'm on even
4:00
touch net A I is like really good
4:03
for. The. In an assistant to do
4:05
some blog posts and then there was a
4:07
small number of who really got at like
4:09
oh yeah this is really gonna change the
4:11
way I do work as the Martin and
4:13
against the grain. The reason we created that
4:15
podcast is because really that many side of
4:18
five cent quote whether you agree with it
4:20
or not that it really speaks to why
4:22
wanted to this podcast my we enjoy doing
4:24
this podcast is because. We. Want the
4:26
people who listen to this podcast to be in
4:28
the five percent right? We want those marketers to
4:30
be the risk takers, the opportunistic, the people who
4:33
do things first, the present and future leaders. And
4:35
that's really the only reason I want to this
4:37
podcast when we originally agreed to start recording this
4:39
podcast. One thing, keep an eye. it's really clear
4:41
on as if we just regurgitate. Best practice for
4:43
marketing were at we own a do It Right
4:46
We actually would prefer to do something where we
4:48
can help people the in that five percent on.
4:50
I think a I as one of the things
4:52
that you can adopt much faster than anyone else
4:54
and be in that five. Percent. And then
4:56
there was the other category of marketers
4:59
You think? hey, Ai is really cool
5:01
for a blog post like isolate. Really
5:03
use it for sure. It's fine as
5:05
good a system for blog posts. And.
5:07
It's not that they are. Not.
5:09
Working hard or want to do new things.
5:12
it's just that they don't really have the
5:14
inspiration to figure out how to use a
5:16
I for more. Thanks! So I was like
5:18
ah cool wall You know what I'll do
5:20
is I will create a short course and
5:23
I will use a I to create the
5:25
course and the courses about how to. Use.
5:27
A I to implement the craft of market
5:30
in into your business and so what am
5:32
I do here is go through a lot
5:34
of bed of the course because there's some
5:37
prompts and think that that and some learn
5:39
that I've had that I can pass on
5:41
to you all here and maybe that will
5:43
be helpful seat to figure out how you
5:46
want to use A I see your marketing
5:48
going forward so this is the killer landing
5:50
page that I put together. In
5:52
a very short amount of time. You.
5:55
Probably have heard me talk about my design
5:57
skills are I'm very fortunate that I get
5:59
to use. Spot I have I have soda
6:01
cans think it's like a pro or and advice can't.
6:04
And so the you the first thing to say
6:06
and this will give you an old description of
6:08
the course and and then I'm going to get
6:10
into like how I used ai and given to
6:12
a lot of the prom said you can kind
6:14
of steel or go sign up to get the
6:16
exact taxed. This is for the most part much
6:18
better if you look at you tube does not
6:20
mean I'm going up and a make it really
6:22
good for all of you Rss subscribers we love
6:24
you all but just in terms of the visuals
6:26
when you look at the prom through the prompt
6:28
or just big blocks of text some nothing that
6:30
the we the my blog give you a synopsis
6:32
of them so you can go use. It go
6:34
scraps of you tube and check out our You Tube
6:36
channel is doing really well. You'll get all of the
6:38
visuals if you're an Rss. I'm gonna give you that
6:41
that we like. Where you can sign up for the
6:43
course, you can just get a dump of all of
6:45
that prompted. that's helpful or it says. The reason I'm
6:47
kind of starting with this is Lana Pay so I
6:49
have about four hundred. again I didn't linked in post.
6:51
I think I did a tweet to the tweeted i'm
6:53
really have not Died Stock and to figure out how
6:55
to get my audience by Parker Twitter Ever since a
6:58
lot of the algorithm changes that is on my list
7:00
of things to do this year everly just kind of
7:02
focus on like them because of time constraints and. So
7:04
I think it has like four hundred and fifty
7:06
to five hundred psi ops which is kind of
7:08
cool as the email my usual arrested of much
7:10
more linked in promo met I bought some other
7:12
from things i think I'll do trip promo that
7:14
just for fun but the reason I'm saying that
7:17
is because the landing page was created using a
7:19
i write so I give a I and I
7:21
went off that are quickly show you that we
7:23
did together credit and initial outline for the. Course.
7:26
And then within the problems that I
7:28
show people how to do there is
7:30
like this prompt from John Cables and
7:33
get into some of the lessons and
7:35
the market responders lessons and a basic
7:37
takes this incredible copywriters. Core.
7:40
Tactics to create incredible ad campaigns and then you
7:42
can credible and pace in their spouse. And a
7:44
word I thought pretty well now I would love
7:46
for you all the common sight in the you
7:49
to the tell me whether this is a good
7:51
conversion rate of knots and maybe it has and
7:53
deters the conversion rate to date on about five
7:55
hundred sign ups is about forty three percent, right?
7:58
and again. One. Of the things I. Internet
8:00
court and I will he stay true to
8:02
the purpose of this course which was to
8:04
will allow Ai to ship the things. And.
8:07
Me to really only do light at. It's best
8:09
of the most part I just directed A. I
8:11
had to do the edits because I really wanted
8:13
to open people's minds into the quality of the
8:15
worked at A I can do so for the
8:17
most part everything you see here. Is
8:19
created and shipped via a I and I
8:21
am in the background like the editor in
8:23
chief like hundred directed at were to provide
8:26
some added some things that that so this
8:28
entire lot of pages created by a I
8:30
went through a bunch of different variation. One
8:32
of the things that people really undervalue any
8:34
I as it doesn't get bored and it
8:37
doesn't mind the grind and the grind. The
8:39
kippen I talk about all the time as
8:41
just like those incremental ways to improve the
8:43
work did you do so for example this
8:46
kind of headline here on this planet page
8:48
which says again that poor. Value that
8:50
I thought about this course was
8:52
he's a timeless market in decades
8:54
old lessons. Implemented Three I
8:56
should bring in that old and and you
8:58
together. The headline here is Unlocked The Markman
9:01
Secrets of Legends with a I Free sixty
9:03
you know course and they're like. This sub
9:05
description was pretty cool. From the golden age
9:07
of Marks into the A. Today I discover
9:09
how to build brands stand the test the
9:11
time and then I put in the includes
9:14
as I say or prompts to make your
9:16
mark and stand out. Now the copy Again,
9:18
pretty good seat and see what's in the
9:20
course. We have the likes of David Over
9:22
The Edge on Cable Guys, Talbert and many
9:24
more. Kind of incredible. Legendary Marketers and add
9:27
copywriters and I can paint. People submerges to
9:29
view all the docks. Again to the Rss
9:31
listeners, thank you for tuning in. I will
9:33
call things I the further you to people
9:35
you can see it. So this is the
9:38
first thing we can I worked on which
9:40
is like the outline right? So I came
9:42
up with the initial idea of blended together
9:44
the craft. This is the two things I'm
9:46
really passionate about. the craft of Market that
9:49
not the science. The craft. And an
9:51
instrument than that? through the science which is air
9:53
so we came up with the know martin through
9:55
time seven lessons from the legends know i i
9:57
decide to six bucks loan them wasn't very good
10:00
And the thing I really worked with AI
10:02
on was a format for each lesson, right?
10:04
And I didn't want that just to be
10:06
a, you know, tips and tricks, best practice.
10:08
I wanted to do something much more creative.
10:11
And so if I showed you my prompt
10:13
for creating this course with AI,
10:15
and it was created for the most part
10:17
through ChatTbT Pro, and I'll come on to
10:20
a little bit about the challenges I had
10:22
recently with ChatTbT Pro, but also Claude. Over
10:24
time, I was using Claude much more. But
10:26
like my ChatTbT Pro prompt
10:29
created in this course is probably like 20 pages
10:32
long, right? It took a couple of weeks
10:34
to build, but way easier to build
10:36
than using humans. I will be really
10:38
honest because, again, AI does
10:40
not mind the grind. It doesn't mind me asking
10:42
it to create many edits, many variations, and that's
10:45
what I loved about it, and really quickly iterate
10:47
through things. And so we landed on this
10:49
structure. Everything that
10:51
I selected to do was
10:54
suggested by AI from me prompting it,
10:56
right? So I didn't say, okay, well,
10:58
like, that idea is good from AI,
11:00
and I will decide, oh, I'll use some of that,
11:02
but I'm just going to change it because I believe
11:04
the structure is better. I really wanted to stray true
11:06
to the reason I did this course, which is, hey,
11:08
like, AI can do much more than you think, right?
11:10
It's not all hype. It's not just good for blog
11:12
posts. And so this is a
11:14
structure we landed on after much kind of
11:17
iteration and variation, which is each day would
11:19
have a singular lesson, right? And that lesson
11:21
was a foundational part of marketing.
11:23
So we have, you know, day one is
11:25
introduction to conversational copyright, and then we pick
11:27
a legend, which was Quasi Hopkins, because that
11:29
was a big part of how he created
11:31
the book of scientific advertising. And he has
11:33
that famous quote that you should write in
11:35
the same language that your customers use. Then
11:38
each lesson, I wanted to have a video
11:41
summary for me, again, created by AI, and
11:43
I'll get into that. Then I wanted to
11:45
have real actionable steps, and the actionable steps
11:47
are actually an advanced prompt to actually instrument
11:50
that lesson. And then I wanted
11:52
to have a homework. And so I really thought
11:54
about this, like, what would make the course different,
11:56
and the ability to have like a homework assignment,
11:58
and actually instrument this for your own brands. So
12:00
that's the way I learned is by doing something it's the reason I
12:02
did this course and so I was like how can I use a
12:04
custom GPT to be able to do that and I'll come on to
12:06
that a little bit later because it worked a little differently than I
12:09
thought. So I was pretty happy with that right so I was pretty
12:11
happy with the way that I structured
12:13
each lesson. I think people really want to
12:15
know the prompts so I'm gonna actually go
12:17
through maybe three lessons here. If
12:20
you want the entirety of the course you
12:22
can go to the follow-on link bit.ly
12:25
and then just dash timeless dash
12:27
marketing dash AI so it's just
12:29
like bit.ly slash timeless dash
12:31
marketing dash AI. I'm sure the team will
12:34
put it in the show
12:36
notes and so let's get into
12:38
some cool lessons and I will
12:41
open up the actual emails
12:43
and show you some of the videos. I think
12:45
we're gonna do I think some of
12:47
the cool ones in terms of the prompts you'll
12:50
be interested in. We're gonna do conversational copywriting because
12:52
there's some cool things in there. I think
12:54
we'll do the Gary Halbert one because it's
12:56
really had you create a short sales deal
12:59
for your product and then the one I
13:01
really want to get into is the psychology
13:03
of persuasion by Dr. Robert Caudini because I
13:05
try to create an entire marketing campaign using
13:07
his principles and something else like a neat
13:09
little trick that I'll show you as well.
13:12
Alright we are in HubSpot the very first
13:14
email that I think some people are starting
13:16
to get today after the introduction I did.
13:19
So again all of
13:21
this created by AI all
13:23
of the copy done by AI
13:26
me as the editor not creating
13:29
any copy tiny little words here and there
13:31
maybe but again staying through to the purpose.
13:33
So each email has this quick introduction which
13:35
I think it's done really well. I'm gonna
13:38
do a minute of the first video which
13:40
was for Kossi Hopkins a quick summary of
13:42
that lesson and the thing to
13:44
know here is that AI actually
13:46
created a script and then I instrumented the
13:49
video through Heijin so I did nothing right.
13:51
So like basically I did really nothing other
13:53
than like again continue to prompt AI where
13:55
I thought it could do better or I
13:58
thought it has to be well. was not
14:00
that good and say, hey, like, you should make
14:02
these edits for this video script, do a little more of this, do
14:04
a little more of that. And that did take a bunch of work,
14:06
the edit, and really did take a bunch of work, but again, I
14:08
did not change really any words or
14:10
anything like that for the video script. So here's
14:13
what it sounds like, and I'll give you my
14:15
thoughts and video in general for, as
14:17
part of a course like this, after about a
14:20
minute of this video. Hey there.
14:22
Today we're going to cover a lesson
14:24
from Claude Hopkins. He is considered
14:27
one of the pioneers of modern advertising.
14:29
In the early 90s for the people on
14:31
YouTube, you see this like cool little image of
14:33
Claude Hopkins come up in the top
14:36
right hand panel. All my
14:38
own quick time editing skills. I have
14:40
real skills, right? Pretty
14:42
awesome. And strategies that are still
14:44
used in advertising today, such
14:46
as the use of scientific approaches. Other things
14:49
for people on YouTube and even on RSS,
14:51
what do you notice about the video? I
14:54
think why it's better quality and ones that
14:56
I've done previously on HeyGen because I'm a
14:58
hands wavy person. I kept my
15:00
hand very, very still. And so this is why this
15:02
video looks much more realistic than other ones that I've
15:04
done previously. That's the video. So every
15:06
single lesson comes with a cool video
15:08
overview with me with a very Americanized
15:10
accent. Now I will say
15:12
all of my friends think I sound
15:14
American anyway, because I've worked with American
15:16
companies so long, but HeyGen really like
15:19
gives me that real American accent. Now
15:26
what do I think about these videos? They're good. You
15:29
know, they're fine. I actually would use them as part
15:31
of a course because they're so easy to do, right?
15:33
I'm sitting downstairs at 10 p.m. and
15:36
I'm doing the video script via AI. And then
15:38
I don't have to go upstairs and record it
15:40
myself because I've already trained HeyGen to have an
15:42
avatar for me. So then I can just put
15:45
in the script and actually have the AI generated
15:47
video without me having to go back and forth
15:49
and actually record. Now if you really want the
15:51
video to be something that sells people on something
15:54
or converts them or makes them feel a certain
15:56
way, then yeah, these videos are not great,
15:58
right? Because they don't have much emotion. don't have
16:00
much randomness of tone. All of these things,
16:02
they don't have any of that. It's very
16:05
formulaic. I think it's good for an educational
16:07
video. I also think, really, they
16:09
can be a little boring. So the first one was
16:11
two minute long. And I watched it and was like,
16:13
oh, people probably want to think this is a little
16:15
bit boring. And so then I made the other ones
16:17
a little bit shorter. But I wanted to include a
16:20
video for every single email, again, to say true to
16:22
the purpose of this, to say, hey, what would it
16:24
look like if AI created this course with someone directing
16:26
it? So I think, again, good for educational things. Not
16:29
great if you really want to use video to sell someone, make
16:31
someone click on something, make someone feel a certain way.
16:33
Now, I want to get into the thing people really
16:35
will want to know is, OK, well, tell me about
16:37
the action plan. How do we have AI create the
16:40
craft for each of these lessons? So the lesson here
16:42
was really to talk to them in the language they
16:44
use every day, the language in which they think. Marketing,
16:47
this stuff still is the
16:49
key. Like, we've maybe
16:51
forgotten that, in some ways,
16:53
marketing is a simplistic thing.
16:56
Create value for customers, sell customers on
16:58
the benefits of your product, and the language
17:00
that they understand. I don't know how brands,
17:03
and particularly B2B, just
17:05
don't get this. You go to their website,
17:07
and it's so jargony. It's such bullsh**. It's
17:10
such, bleh, words, vomit
17:12
of words. Do
17:15
you talk to customers? Do you actually listen to
17:17
the way they talk about your product? Do you
17:19
actually hear how they describe the benefits of your
17:21
product? Why not just put that on your website?
17:23
It's much, much cleaner. So how
17:25
does AI help you to do this in
17:27
a much more scalable way, right? So today,
17:29
you have to be pretty good at copy.
17:31
You have to be a good copywriter, understand
17:33
content, understand your customers. Here's what
17:35
I think you can do in the future.
17:37
And there's a detailed, step-by-step way to do
17:39
this in this very first email, which is
17:41
really build a gramily of customer language. I
17:43
use gramily all the time. Got an email
17:45
notification from them. Whereas one of the only
17:47
ones I find valuable, all brands do this.
17:49
Now they think it's really cool. They can
17:51
a real kind of way to get people
17:53
to reconvert, which is here's all of your
17:55
data, how you're using the product, like in
17:57
really interesting ways. Gramily is pretty interesting. and
17:59
then I'm a power user and it showed
18:01
me how I do better than all of
18:03
the other Grammarly users on like vocabulary and
18:05
all these different things, right? But what about
18:07
a Grammarly for your customer language that actually
18:09
when you put in content and you create
18:11
that content, whether it's a web page
18:14
or landing page or an email, it starts
18:16
to correct it to be much more like
18:18
how your customers speak. And AI will be
18:20
able to do this, right? So you'll be
18:22
able to take your sales calls and transcribe
18:24
all those sales calls using something like otter.ai.
18:26
You'll be able to compile those transcriptions and
18:28
then you'll be able to create a chat
18:30
activity prompt to create categories of themes, right?
18:32
And I put in here and again for
18:34
the people in RSS, what I'm showing is
18:36
a detailed prompt where you've when you've got
18:38
the transcriptions, all you need to do is
18:40
transcribe through any one of these AI tools.
18:43
I kind of said otter.ai, but there's a bunch
18:45
of them and I think even Gong might have
18:47
its own features that you can just export the
18:49
transcriptions. And then I provide this
18:51
kind of advanced prompt that when you just
18:53
have it dumped into a dock, we'll actually
18:55
be able to convert those things into themes,
18:57
right? Convert all of those things into themes.
19:00
And that is the beauty of AI that
19:02
it's able to distill so much
19:04
data into like core themes and categories.
19:06
And I think that's one of the
19:08
main advantages of it. And then I
19:11
go through this where you can basically
19:13
get all the way to the end and
19:15
then create a custom GBT. So think about
19:17
the custom GBT, right? I would pre-program it
19:20
to basically look for errors in the content
19:22
that I upload that don't fit the
19:24
themes and categories in the way my customers speak,
19:26
right? So example, simplistic prompt here
19:28
is highlight how this copy aligns with
19:30
our customer language themes and provide suggestions
19:32
for making it resonate more effectively with
19:34
our target audience. So I'd actually have
19:37
a custom GBT pre-programmed to
19:39
look for these errors where my copy deviates
19:41
from the way the customers actually speak about
19:43
it. And I would have already had an
19:45
AI go through these transcriptions and create summaries
19:47
and themes and things that are really important
19:50
to make sure that they're incorporated into our
19:52
copy. I would have uploaded that for the
19:54
custom GBT and the custom GBT will then
19:56
actually kind of act like agromal in a
19:58
way whenever I have... upload a document,
20:00
it will check against the pre-programmed
20:02
doc that I've uploaded to see if
20:04
that language resonates with our customers. The
20:06
cool thing here is you can actually
20:08
have a custom GBT for sales, close
20:10
one deals and close loss deals. What
20:13
are the commonalities in language for
20:15
when people close versus when they don't? And
20:17
can I check my copy against either
20:20
of those things because I want to be
20:22
much more like the language you use when
20:24
people actually close and buy my product? And
20:27
then you can actually continue to expand this.
20:29
You can export customer support tickets, you can
20:31
export website copy, you can export social media
20:33
mentions and have a real actually customer comms
20:36
library pre-programmed into a 12 GBT to act
20:38
like a good primary for your content. I
20:41
really enjoy this part. It's a mean AI
20:43
kind of cycle through different examples from these
20:45
marketers. It was a really famous one from
20:47
Kwasi Hopkins for Schultz Brewery. There's
20:49
this amazing piece of copy that
20:51
he wrote to really change the
20:54
fortunes of that brand. You
20:56
can check this from the course if you sign up.
20:58
And then it kind of goes through, again, AI did
21:00
all this, why that campaign was so successful and I
21:02
think it was pretty good, right? But I want to
21:04
get into a couple more lessons and then I'll get
21:07
into my thoughts in the custom GBT because there's some
21:09
things there that happened that are pretty interesting. All
21:11
right, so this is a lesson from Dr.
21:14
Robert Caudine, who's an American psychologist and wrote
21:16
this book about the art of persuasion. And
21:19
the lesson basically shows you the
21:21
different persuasion techniques that he feels you can
21:23
incorporate into your marketing to get people to
21:25
buy your product. And then
21:28
I give this advanced AI prompt to
21:30
actually create an entire marketing campaign that
21:32
incorporates those techniques into the campaign itself.
21:34
Now, I think this is a good
21:36
time to show the custom GBT. This
21:39
was a big part of what I wanted to
21:41
do, which is to make courses different if you
21:43
provide a custom GBT, would that change the experience
21:45
of the course, make it a little bit more
21:47
interactive? And so each lesson, and we'll go through
21:49
this one, has this ability to go
21:51
to this homework bot, right? And so if
21:54
I do about this bot, and
21:56
you'll see on the screen for the people
21:58
listening, it basically gives you guidance. and
22:00
how to use the bot and so it will
22:02
say, hey
22:04
choose the homework assignment that you want to do. I've been trying to
22:06
figure out like courses, a lot of
22:08
them aren't that interactive unless you have a community and this is
22:10
a way to make it a little more interactive without having to
22:12
set up a community. So we're going to
22:15
do day six, we're going to do the psychology of persuasion
22:17
and copywriting. So we're going to
22:19
choose day six and this is not going to
22:21
work really well and there's an interesting reason why
22:23
and I want to tell you about
22:26
that. And I'm going to jump between this and the
22:28
prompt to show you what's going to happen. So
22:30
we're going to do the first step. And
22:32
I'm going to do the prompt to show you what
22:34
is happening to chat GPT right now because
22:36
it's pretty interesting. I'm so glad
22:39
it's doing this because it's actually wild that
22:41
it's doing this. So
22:44
here is the incredible thing that is happening
22:46
to chat GPT. Pro,
22:48
I'm on pro. And
22:50
I noticed this because I have some people going through
22:52
the course. Prior to me actually promoting it, I had
22:54
some people go through the course and
22:56
they started to tell me that the prompts that I was
22:58
providing were not working in the same way. And
23:01
why are they not working in the same way?
23:03
Because chat GPT is getting really lazy and the
23:06
answer is getting much worse than when they were
23:09
like two to three weeks ago, this prompt worked perfectly.
23:11
And now I've noticed it's the exact same prompt and
23:13
it works very differently. So this is a great example
23:16
of what it started to do. So this is
23:18
the result from chat GPT. This is pre-programmed
23:20
with a prompt. So the way
23:22
you should do this is the
23:24
custom GPT has an 8,000 character limit.
23:26
So you can only actually incorporate 8,000
23:28
words of instructions. And so what I do
23:30
is I put the instructions in PDF if
23:33
you want it to be a complex custom GPT. And
23:36
then really what I put into the custom GPT
23:38
itself in terms of the instructions is just a
23:40
big if then else loop. So if the user
23:42
selects this, execute this PDF that I've uploaded, which
23:44
has all the instructions for that option. If
23:47
the user selects this, execute all these instructions
23:49
that I've just uploaded. And
23:51
so what I'm going to show you
23:53
is the prompt to get
23:56
the marketing campaign created by Dr.
23:58
Robert Haldini. You can see here
24:00
what it's done is I've said day six and it
24:02
said, cool, here's basically what you should do. And it's
24:04
told me to do all the work, right? It's told
24:06
me to provide a product name. It's
24:09
told me to provide a campaign strategy. It's
24:11
just given me like what I should do,
24:13
incorporate psychological principles. So then it says I
24:15
should do what Dr. Robert Caudini does. Then
24:18
it says campaign strategy refinement. It is basically telling
24:20
me to execute the prompt. It is treating me
24:22
like I am the AI because I'm going to
24:24
show you the prompt that I've
24:26
been refinement and work really
24:28
well. This was actually my favorite
24:30
prompt. So basically you give it a task,
24:33
right? So the task is to
24:35
create a compelling marketing campaign for the product's
24:38
target audience. And then I say, you know,
24:40
the campaign should offer something of incredible value.
24:42
I provide some suggestions. The marketing campaign should
24:44
include the key principles. And I
24:46
give the principles that I want it to include
24:48
from things taught by Dr. Robert Caudini. The only
24:50
input the user should have is product name. So
24:53
the user should input the product name because I
24:55
think that they have to input the product name
24:57
to get a campaign for that product. Then it
24:59
says do each of the steps below for that
25:01
product name. And you can see it says, hey,
25:03
create a campaign strategy. The thing that
25:05
was working really well, for most people, they just
25:07
take the first output from AI and it's really
25:09
generic and it's not that useful. And
25:12
this here is a loop. And that loop should
25:14
be done five times. And at each time it
25:16
goes around, it should get edits
25:19
to make the campaign better on the next iteration.
25:21
And I tell it the loop is considered over when
25:23
you deliver the fifth campaign. I tell it here at
25:25
the start, you must do this five times or else
25:27
we have not served the user. I've tried
25:30
multiple ways to get it to do this five times. This
25:32
was working. It was working really well. I did one for
25:34
Airbnb. It would create a campaign.
25:36
Then it would say, cool, here are the edits
25:38
to make this even better and more compelling. And
25:40
it would do a second version. And then it
25:42
would say, these are the edits we can use
25:44
to make this better, more compelling, more interesting. And
25:46
it would create a third version. Now what
25:49
it's doing, and you can see it in
25:51
the homework bot, you'll
25:53
basically see that it's
25:55
telling me to do all the work. It's like,
25:58
no, screw you. you do
26:00
all the work, right? So we
26:02
can see how this works here. Because I think if
26:04
you copy and paste it in, sometimes you get different
26:07
results. So let's just see. We can copy and paste
26:09
this exact prompt. So we will do
26:12
this, Airbnb. And then I
26:14
will also do one in Claude and show you
26:16
the results for that, because I think Claude is
26:18
starting to become much more of my favorite bot
26:20
to use. Now, while this is happening,
26:22
I'm going to come back to that at the end.
26:24
But I want to wrap up with some thoughts. I'm
26:26
going to just share my exact doc with my thoughts
26:28
in this course. All right,
26:31
so if you want to go
26:33
get this course, here is the
26:35
URL bit.e timeless-marketing-ai will
26:37
be included in all of the notes.
26:39
Here's what I used it for. Show
26:41
outline, the episode format, the individual
26:43
emails. So you saw the video summary using
26:46
HeyGen. You saw the action plans. So
26:48
the advanced prompts were created, for the most part,
26:50
by Claude. And so I've been
26:52
using AI to create prompts. Now I instruct it.
26:54
I give it a format. I give it a
26:56
template. I can upload a doc for Claude. Using
26:59
a template that I want the prompts to be in. And
27:01
so Claude created most of the advanced prompts for
27:03
me. And it was really good. And
27:06
then an example from history. So show outline, I would
27:08
give it like a three out of five. Did really
27:10
well. Episode format, three out of five. I think it
27:12
did pretty well. The individual emails, probably a three out
27:14
of five. But I would say that prompts, I would
27:16
give it a four out of five. Claude was really
27:18
good giving me advanced prompts when I give it a
27:21
template. So pretty good here.
27:23
I think for most SMBs, using
27:25
AI would get you a pretty good
27:27
course. Would it be the best
27:30
course on a topic on the entire
27:32
internet? No. But would it be pretty good that people
27:34
will find valuable? Yes. Custom GPT was
27:36
good, has degraded. I'm not going to go back
27:38
to see how we're performing here. This part is
27:40
a bit of a bummer. It took me ages
27:42
to create that custom GPT. It actually
27:44
basically does all of the lessons, shows you how
27:46
it does the prompt, and does them for any
27:48
product or service you put in. And it's just
27:50
degraded over time. Created the land page. Conversion rate
27:53
is about 500 now. Sign
27:55
up so you can get an estimation of how many
27:57
people have gone there. And the conversion rate is
27:59
about 40. What would I
28:01
do next time? So I would do a more advanced version of this.
28:03
I probably am going to release another version which is going to be
28:05
much more advanced. I would use delphi.ai to
28:07
create a version of me as part of the course.
28:09
Go check out that tool. It allows you to create
28:11
versions of yourself if you're like a business creator and
28:13
you can have that as part of the course so
28:16
they can actually interact with you at all times when
28:18
they're going through the course. So that's going to be
28:20
part of what I do. Hey Jin,
28:22
I would actually ask to feature experts as part
28:24
of the course and ask them to provide me
28:26
their avatar or a five minute video that I
28:29
can create an avatar of them and use their
28:31
image for specific lessons, teaching things that they are
28:33
renowned for whether that's a tactic or they are
28:35
part of a renowned brand and I'm doing a
28:37
lesson from that brand and I think that would really bring
28:40
the course to life. Microsoft Vazza
28:42
1 is like the next level for this. I don't
28:44
know if it's going to be released. You might not
28:46
have all seen it. It really allows you to take
28:48
an image and then create a video from that image
28:50
and it's unbelievable. It's really unbelievable. Go check out Microsoft
28:53
Vazza 1. If there was anything
28:55
like that, I could actually create incredible expert-led
28:57
content by just having the experts featured in
28:59
my course and do some sort of partnership
29:01
with them so they allow me to use
29:03
the imagery to teach a lesson from them,
29:05
to teach a tactic from them. I
29:08
would create more animated videos. It's kind
29:10
of boring. Eventually, you'd fix emails even
29:12
though you've got custom GPT, the video,
29:14
the advanced prompt, the kind of historical
29:17
example of how that lesson was done
29:19
for some known campaign. I
29:21
think animated videos really bring a course to life. There's
29:24
these companies I'm going to dig into. One of them is called Flicky.ai.
29:27
You can see the URL here. That's one of the
29:29
things I would do as well. Then I
29:31
think AI Community for Subscribers where I have a
29:33
bot, maybe delphi.ai where you can go and interact
29:35
with each other and interact with the bot. Actually,
29:38
it brings the course to life. I think
29:40
community-led web courses might be a big
29:42
thing because I think you can create them with a
29:44
bot AI version of yourself and so you don't have
29:46
to give up all of your time to be present,
29:49
but you can still provide value to that community. That's
29:52
it. Let us go back
29:54
to see the absolute nightmare that GPT
29:56
is currently doing here. We'll wrap up
29:58
with this. So like pretty
30:01
hilarious, right? So awful. So
30:04
it has come up with like for Airbnb, you
30:06
live anywhere, experience everything. That's the campaign, invite
30:09
potential guests to immerse themselves in unique, culturally
30:11
rich environments around the world. Basically,
30:13
it allows people to live in incredible places
30:16
around the world. It is something that you
30:18
could see Airbnb do and they saw that
30:20
not that creative, right? And then it talks
30:22
about how you could incorporate how the needs
30:24
principles or reciprocity, commitment and consistency, social proof,
30:26
authority, life and scarcity. And
30:28
then it says, this is the funny
30:30
part, edit for next campaign, right? And
30:33
then it doesn't do anything. For the
30:35
next iteration, I'll focus on making the
30:37
marketing campaign more interactive and personalized, yes,
30:39
because it's pretty generic right now. Adding
30:41
gamification element could enhance engagement and incorporate
30:43
in more detail local insights might provide deeper value
30:45
to potential customers. So it gets it, right? It
30:47
gets how to provide value the next
30:49
time around, the iteration the next time around. Now,
30:51
if I just said, cool, please do that, it
30:55
will do it, right? But it
30:57
was looping around without me
30:59
having to do that. And let's just go
31:01
check. Lastly, on our
31:04
good old friend, Claude,
31:07
Claude has done it, right? It's a pretty unstructured, I'd have
31:09
to dig into this, but Claude has done it. So let's
31:11
see what the fifth one is. The
31:13
ultimate iteration of Airbnb Explorers Club, we could
31:15
evolve into a full travel lifestyle subscription bundling
31:17
together all the previous components. So it's actually
31:19
built on each other. I have to go
31:21
back and read each and every
31:23
campaign. But this is an example of what I
31:25
mean where Claude does a much better job than
31:28
chat GPT right now. It actually has iterated
31:30
on five campaigns. It's built on each one
31:32
to figure out how we can make things
31:34
better over time. And I suspect if
31:36
I actually went in here, it would be really good.
31:39
So my tip here is have
31:42
the AI iterate on its own work
31:45
four or five times before you take the
31:47
output to iterate on it yourself. All
31:50
right, this was a kind of manic episode.
31:52
I hope it's been valuable. If you liked
31:54
the course, again, the URL
31:56
is dip.ly-timeless. So
32:00
basically Bitly and then
32:03
the custom link is timeless-marketing-ai. Hopefully
32:06
you learned some things. Hopefully you found this valuable.
32:08
Until next time, I will be back with Kip
32:10
as always on the next episode of Marketing Against
32:30
the World.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More