Curious about how AI is transforming the landscape of marketing? Join us as we host Jess Gonzalez, a seasoned marketer with a decade of experience at the intersection of AI and marketing. You'll uncover the secrets behind Jess's innovative approach to content production and curation. From scraping trending topics on Reddit to leveraging a human-in-the-loop system, Jess and her team have mastered the art of creating SEO-optimized content that truly converts.
Discover how to streamline your ABM outreach and content creation with cutting-edge AI tools like HubSpot, Claude, Perplexity, and ChatGPT. Jess shares invaluable insights on automating processes to save time and cut costs, all while maintaining the quality and relevance of your marketing efforts. We also delve into the practical applications of AI in affiliate marketing—think automated image generation, review analysis, and ad creation—all integrated seamlessly into your campaign management. Tune in to learn why marketing fundamentals still matter and how AI can amplify rather than replace your existing strategies.
Jess Gonzalez on LinkedIn
Send us a Text Message, give feedback on the episode, suggest a guest or topic
Visit the Remarkable Marketing Podcast website to see all our episodes.
Visit the Remarkable Marketing Podcast on YouTube
Remarkable Marketing Podcast Highlights on Instagram
Eric Eden on LinkedIn
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Eric Eden (00:00):
Today we are talking about marketing and AI what is
the future of marketing and AIand we have a great guest to
help us talk through that.
Jess, welcome to the show.
Jess Gonzalez (00:10):
Thanks for having me.
Eric Eden (00:12):
Why don't we start off by you taking a minute to
share a little bit about who youare, what you do, what you've
done?
Jess Gonzalez (00:20):
Yeah, hi, I'm Jess.
I love marketing.
I've been a quota carryingmarketer for almost my entire
career, so I come to marketingwith really a revenue and
process focused lens.
I've spent my career in I wouldsay, really fortunate once of a
career roles where I've workedwith some of the largest media
companies, publishers, some ofthe fastest growing SaaS brands
(00:41):
to really grow their companymore quickly by growing
marketing quickly, with a focuson driving revenue.
And lately I've been doing alot with AI.
I have been just a caveat.
I've been using AI since beforeit was a hot thing and before
ChatGPT was launched, so I'vebeen in AI for about a decade,
but I've been finding somereally interesting ways to use
(01:03):
it that I think not everyone'sleveraging it for.
Eric Eden (01:07):
Awesome, and you're also a podcaster.
You have your own podcast.
Jess Gonzalez (01:12):
Yes, marketers Talking and Marketers Talking AI
.
Where we bring on marketerstalking, we bring on marketers
of all walks of life anddifferent seniority and
experience levels to share theirstory and what they have found
that works for them in themarketing world.
And then marketers talking AI.
My co-founder, pavel, and Icover the recent AI news and we
are talking to founders of AIMartech specifically, founders
(01:35):
who are actually marketers, notjust developers, who are
building AI tools to sell tomarketers.
So that's an exciting serieslaunching soon.
Eric Eden (01:44):
All right, we're ready to be inspired.
Tell us a story about some ofthe best things you've done with
marketing and AI so far.
Jess Gonzalez (01:52):
Yeah.
So my favorite use case and I'mgoing to start with content,
which for a lot of people,they're probably saying yeah,
content.
I know there's 50 tools thatwill make me a million blog
posts a day.
It's everywhere.
We took a little bit differentapproach, but we've been using
it heavily for contentproduction and curation.
So, specifically, we had aclient that was starting off
(02:13):
fresh, brand new website, nodomain authority at all, and
they wanted to be able to growand really engage in the startup
community.
Their ICP was founders, ideallyfounders who are going through
incubators or some other earlystage funding.
So they want to get in front ofthis specific audience and they
have nothing on their websiteto really make them compelling
(02:34):
as an authority on the topic yet.
So we launched a programmaticSEO effort, which means that we
produce a ton of SEO optimizedcontent, and we did it all
really quickly using somedifferent workflows, processes,
tools, however you want to do it.
So, instead of manuallycreating 200 pages, we had a
(02:55):
system to be able to do it usingAirtable, makecom, retool,
zapier a couple tools similar tothat.
So we started off with askingourselves what does this target
audience want to engage with?
And one of the spots that I'vegone to for almost my entire
career for inspiration is Reddit.
So we actually set up a processthat it would scrape some
(03:17):
different subreddits every dayand it would see what are the
three or four most interestingarticles that have been posted
in the last 24 hours, that aretrending, like what's getting
the most traction.
And so we use Reddit to curatethe topic pool.
And then we took those topics,put them into an engine to get
some different headlines out ofit.
(03:38):
So we would take, let's say, 10topics that are doing really
well, let's distill that downinto three that we care about,
and then for each of thosetopics let's generate 10
different titles, and then wewould have a human in the loop.
Human in the loop is a termthat not a lot of people may be
familiar with, but it means thatyou have an actual person who's
(03:59):
taking your AI output andthey're looking at it to refine
it before putting it back intoAI to get to that next step.
So our human in the loop wouldgo in and they would look and
see those different titles andthey would say, okay, these
eight, I think, are reallycompelling.
I like these two, but they'renot quite great.
Let me go regenerate them ormake some changes, and then we
(04:20):
would take those titles and wewould use those to produce
content, but here's where itgets different.
You can't just put that titleinto ChatGPT.
You can.
You're going to get something,though, that really sounds AI
generated, and so what we did iswe really thought about what is
the unique point of view thatour client has, what is their
(04:43):
opinion and ideas.
We want to make sure is included, and are there different areas
that we know that they playreally well.
So, for example, if you were acompany that does, let's say,
software development, you'd wantto make sure that topic tied
into software developmentsomewhere, and it might not
always be super obvious, but AIis really good at making stuff
up, and so we would tell it.
(05:04):
Here's our topic.
We want to make sure that thetone is optimistic, and here's
our super long explanation ofthe brand voice so we can make
sure the content actually soundslike it was written by a
specific person, and we want tomake sure that it ties into the
future of software development,and these are tech stars funded
companies.
We want to mention tech starsand maybe find a couple
(05:26):
references to these otherpublications, so we're giving it
a really robust prompt thatliterally took weeks to develop
and months of fine tuning, andthen it puts out a piece of
content that then goes to aperson to look at Quick glance.
Does it look like it's in theright direction, yes or no?
And if it is, it'll go back intoa system where we actually have
(05:47):
different AI models talking toeach other to provide feedback
on the content to each other,and so it's like having a team
of three or four copy editorsthat are coming in reviewing and
providing feedback, but it allhappens in like 20 seconds and
it goes back and forth.
And then it would spit out thisarticle that sounds pretty good
(06:08):
.
It's 95% of the way there, andwe actually eventually added in
a step where it would alsoprovide feedback of what it
thinks should be tweaked in thearticle.
And then that goes to anotherhuman in the loop who's an
actual copy editor, actualwriter, who looks through, makes
some changes, make sure that itfeels human.
It sounds human Also, that allthe data is correct in there.
(06:29):
Ai does hallucinate.
We try and get rid of that asmuch as possible, but sometimes
it will make up its own factsand even put in a citation for
it that doesn't exist.
So they give it a once lookover and then it's ready for
publishing.
And so by doing that, we'reable to make a system where it's
automatically finding trendingtopics on Reddit and producing
(06:52):
10 to 20 blog posts in one ortwo days, which means in a month
you can put out hundreds ofpieces of content that are all
unique, all designed to speak tospecific keywords and following
topics that are resonating withyour audience.
So it's not just using it forthat first content production
piece, but it's thinking abouthow can AI give me leverage
(07:14):
across what should be an entirecontent production process?
And we've taken that.
We built it for a client, wetweaked it, repeated it, built
it for another client, tweakedit, repeated it, built it for
ourselves too, tweaked it,repeated it, built it for
ourselves too, so we could putout more content.
And we found that by reallyputting out this content, it's
helping with domain authority.
So one of our clients went fromzero to two in a day and then
(07:36):
they've gone up to about 12within a week and now they're
close to 30.
But more importantly, it'sattracting the right audience to
them.
That's turning into inboundrequests, that's turning into
deals and opportunities, andthat's ultimately what we care
about is building marketing thatdrives revenue.
Eric Eden (07:55):
It's really almost fueling your inbound engine that
wasn't existing before thewebsite was built was built and
is it working that it'sgenerating this volume of
content with this machine andit's getting indexed, it's
getting ranked?
Is that piece of it working?
Jess Gonzalez (08:13):
Yeah, I will say the bottleneck on all of it.
There's a couple spots.
One is we insist on humaneditors in the process because
we want to make sure it's goodquality content.
That can bog it down a littlebit depending on the length of
the content, and Google doesn'tjust auto magically index every
single thing you give to it.
So we've had a couple instanceswhere it's taken a couple weeks
(08:35):
to get all of the contentindexed.
Typically, we'll see a goodportion of it indexed up front
at least three or four new pagesa day.
At the low end we had oneclient that we were surprised.
Day one we put it out.
It was a pretty large piecethat was really more of a pillar
page of content, and the nextday they had an opportunity to
come in from it, which isastonishing, because sometimes
(08:57):
Google does take a bit topublish.
What I will say is we haven'tseen any instances where
Google's penalizing using AIgenerated content, because I
honestly think Google can't tellthat it's AI generated, because
it's not bad content.
Eric Eden (09:15):
Yeah, I think we'll get to a point one day when
there'll be really gooddetectors all around that'll be
able to tell which content wasgenerated with AI and perhaps
what percentage of it wasgenerated with AI, but I think
the only one who perhaps hasthat tool?
There was this scandalous storyabout how OpenAI has had this
(09:35):
detector internally but hasn'treleased it.
But because they haven'treleased it, no one really knows
.
They say that they had thisinternal tool.
They could tell for 99%, butprobably Google doesn't know,
like you're saying, and it doesprobably take still just like if
you were just creating it allmanually.
It does take a little bit oftime Weeks sometimes can be 30,
(09:57):
60 days, even if you'reproducing a good volume of
content for it to get indexed.
That's what I've seen acrossmany projects over time.
It's never instantaneous, sothat probably is still there,
like you're saying.
But I think that's fascinatingbecause I worked for one company
just to give the manual analogynext to that, where we did
generate a massive amount ofbusiness this is five years ago
(10:18):
from creating a thousand blogposts that were each like 2000
words long, written byfreelancers, and I can't tell
you how much brain damage camefrom that process of humans
writing 2000 word blog poststhat didn't weren't keyword
stuff, that were actually decent, and then trying to get three
(10:41):
or five of them a week andeventually, over time, over a
period of two years, we got toenough where we were getting
like hundreds of thousands ofvisitors a month to our website
that were going into oursoftware free trial from the
blog.
But we spent years of manuallabor and a lot of money paying
freelancers to do that and itwas good, but this is a lot less
(11:03):
painful way to do somethinglike that.
Jess Gonzalez (11:05):
When you think about the cost.
So, related to this project, wehad a client that was doing
programmatic landing pagesspecifically for ABM.
So they had 40 accounts thatthey wanted to do landing pages
for and the landing pages werepretty uniform, except we're
just going to change the companylisted on the top, we're going
to change the URL and maybe fiveor six pieces of content we're
(11:29):
going to slightly tweak, butwe're not trying to index them.
So we don't care if it'sduplicative content because
they're really designed for ABMoutreach.
They had a quote.
If it's duplicative contentbecause they're really designed
for ABM outreach, they had aquote.
It was $30,000 for 50 pages andI was like, are you kidding me?
For someone to go and buildthem all and that's had them
getting like 80% of the content.
I was like that's crazy.
And so they were on HubSpot.
(11:50):
We rolled out HubSpot DB, madea spreadsheet and we did have to
set up the landing pagetemplate and the variables.
But it took an hour and a halfto do to build all the pages
when the quote was $30,000before, which just seems
ridiculous.
And if you look at how muchsometimes a blog post especially
on the B2B side costs, you'respending $1,000, $1,500 per blog
(12:11):
post coming out, so I wouldimagine that's a massive cost
savings for a lot of companiesto be able to build these
programmatically too.
Eric Eden (12:20):
Yeah, and this belt tightening era, just a lot of
people.
They just are like we can't doit.
That's where a lot of companieshave gone with it.
So I think personalization isthe use case that you're
mentioning, that a lot of peopleare starting to experiment with
, in the sense of creating alanding page for each target
account.
To do it manually is, like,really expensive.
(12:43):
As you laid out To do itautomated, you probably still
want to check over the work, butit's a lot faster than trying
to do it all manually and that'san awesome process.
Jess Gonzalez (12:53):
I think what AI is really great at that is
underutilized by people isgiving you options, giving you
something to react to.
So if super boring example youneed to figure out meals for a
week, it's a lot easier for usas humans to look at our options
and say, okay, monday night youcan have one of these four
things what do you want?
Easier to pick than to say,okay, what do you want.
(13:16):
Monday night it's so mucheasier to make a decision and so
using AI or even some automatedsystems to help you figure out
what your options are and thenbe able to make that decision
more rapidly.
I think that's one of the mostunderutilized areas for AI right
now for marketing.
Eric Eden (13:31):
I agree.
One of my favorite use casesalong those lines is in using AI
as a co-pilot.
When I'm doing messagingarchitecture projects, I have it
take like the about us for acompany.
I'm like, take this and rewritethe about us for the company in
10 different styles and I willcome back with a conversational
(13:52):
one, a formal one, aninspirational one in 10
different ways.
And then it's very interestingto share those options with
executives and see which oneresonates with them the most,
based on their company culture,because it's the same fact set
as like their core about us thatthey had.
But people are like I didn'tthink about it that way, and so
I think it is really great, as aco-pilot, of giving options of
(14:13):
different routes to go.
Like I said, I think that's agreat tip of giving options of
different routes to go Like Isaid.
I think that's a great tip.
Jess Gonzalez (14:17):
My co-founder, pavel, and I.
We use AI as a co-pilot inalmost every meeting.
We have every brainstormingsession and sometimes what we
find is that it will in thenotes, it will pick out things
that we didn't realize we talkedabout, or it'll put down action
items where it'll say based onthese other action items, you
might want to also considerthese couple things and it's
(14:39):
been really helpful for that,where it's almost like having
another brain in the room.
And on the feedback side,similar to what you're talking
about with the messaging, takethat messaging, put it in AI and
ask AI to tell you who thetarget audience is.
I found that I've written somestuff that I thought was like
really fantastic and it did notconvert and I put it in.
(15:00):
I was like what's going on here?
And AI told me that my copy wasfor a totally different
audience and I was like, oh shit, I guess I missed a note here
somewhere.
But I wouldn't have realized itwithout AI being that kind of
totally I would say objectiveperson that's able to come in
and give you feedback, notknowing what's in your head for
the context behind it.
Eric Eden (15:21):
So you're doing some really advanced things here in
August 2024.
What's your marketing?
Ai tech stack.
Jess Gonzalez (15:30):
Honestly, I love Claude and Perplexity.
Chatgpt, of course, is in their4.0.
But I think Claude andPerplexity ChatGPT, of course,
is in their 4.0, but I thinkClaude and Perplexity are two of
the best tools that people canuse.
Personally, I'm a fan of Claudemore for writing.
I think that it produces morehuman feeling content than
ChatGPT, unless you have.
(15:51):
If you take the time to make asuper robust prompt, chatgpt can
get you there too, but I feellike Claude, out of the box,
will give you a better writingproduct than ChatGPT will.
So I'm a fan of Claude forwriting and then sending that
Claude output over to ChatGPT togive it feedback on.
I love perplexity for search.
(16:11):
I feel like it's giving meanswers to things that I asked
the most, like I asked it themost random question that's not
super specific and it somehowknows exactly what I want, which
feels magical.
And then, when we look betweenthe tools, where I think they
really become powerful is howwe're automating between them
and so, using Zapier, retool,makecom to try and find ways to
(16:35):
pull it in.
I'll give an example For mypodcast.
I have an entire workflowbetween ClickUp and HubSpot and
then I'll actually take thetopic and I'll take the
transcript from the podcast andI'll send it into Summarize AI,
which is my favorite tool forvideo transcriptions and
summaries, and it'll pop out acouple of different title
(16:56):
options.
I can take those and send thoseback to Claude to give me other
options if I don't like it andfinding ways to really get those
tools to work together.
So Claude Can't RecommendEnough Perplexity.
Summarize are my kind of threefavorite, I would say.
Eric Eden (17:12):
Three go-to.
And what about for images orsoon-to-be video?
Any favorites in those areas?
Jess Gonzalez (17:18):
I'm going to tell you I so I'm going to age
myself a little bit here I don'tuse Photoshop.
I use Fireworks still, whichAdobe sunset in like 2012.
But I never took the time touse Photoshop.
But I've actually been usingthe generative AI in Photoshop
recently because it's so easy touse, so that's been my go-to.
(17:41):
I've played around with acouple different tools.
There's one that I've gone towhen I need to generate
something from scratch, whereI'm starting from absolutely
nothing, but I'm typicallystarting from an image that I'm
trying to edit, and so I'll gointo Photoshop for it, or I'll
actually use Canva, especiallylike headshots.
Sometimes clients have photosthat are a bit older and you
(18:01):
need to sharpen them up.
I've been really impressed withCanva's AI options for editing
photos.
Eric Eden (18:07):
Yeah, so is the Adobe tool.
Is that Fireflies?
Jess Gonzalez (18:13):
Yes, I think so yeah.
Eric Eden (18:16):
Have you found that to be decent?
I haven't tested that one.
Jess Gonzalez (18:18):
Oh, so I actually made my wedding invitations
with it.
I had a Call of Duty themedwedding, which, for those who
don't know which is probablyeveryone I met my husband on
Call of Duty, like all singlepeople in our 30s, and we wanted
a Call of Duty themed weddingbecause a lot of our gaming
friends were coming.
If you search Call of Dutythemed party stuff, it's all for
(18:40):
, like little kids and it's allarmy themed, and we wanted
something that I would describeas a little more up level, a
little more upscale, and I hadsome different ideas but I
couldn't quite get it to workand I spent four or five hours
trying to blend images togetherand just going in and zooming in
and erasing pixels.
Then I was like, let me justgive this a try, and I think the
video is actually still up onmy LinkedIn.
(19:01):
It took me less than 10 minutesto make our wedding invitations
and they ended up being totallycustom.
I wanted to blend a coupleimages it's two people
parachuting in and it saysdropping in together and so I
wanted to alter them so theylook a little more like the
operators that we play in thegame.
And it was the best processever.
It was the first time I used it.
(19:22):
It was so easy to use and, assomeone again who did not take
the time to learn Photoshopbecause I'm still using the 2010
tools I just didn't want tolearn it.
I found it easy to use, so Ithink that's saying a bit about
it.
Eric Eden (19:36):
Now I have to know wedding venue, reception venue
was it also Call of Duty themedor just the invites?
Jess Gonzalez (19:42):
Yes, so we turned into a LAN party, which is also
a common aging myself.
But yeah, we had it at ourhouse and then we had the
reception at a what do we callit?
It's a place that has a ton ofcomputers that are networked
together to have a LAN party.
It's like a nerd thing.
So, yeah, everything was reallyCall of Duty theme.
A bunch of our gaming friendscame in, so it was a good time.
Eric Eden (20:05):
Congratulations.
That's awesome.
So I actually have been testingmid-journey against Dolly to
see which ones I can get betterimages out of, and I think for a
lot of the marketing I do likeimages for blog posts right.
I don't need to spend hundredsof dollars for an image, an
abstract image, for a blog postwith one of the stock photo
(20:28):
companies, but I think peoplecan tell that it's AI in a lot
of cases.
But I've actually gottendiffering results.
But with ChatGPT Dolly, you canactually upload an image and
have it change it, which isinteresting.
Not quite Photoshop, but youcan say take this image and do
this with it, and a lot of timesit'll it, which is interesting.
Not quite Photoshop, but youcan say take this image and do
this with it, and a lot of timesit'll do some pretty cool
(20:49):
things and I've gotten somepretty good quality stuff out of
Midjourney.
I like to travel.
I've been to 63 countries andso all the places that I've been
to, I had it generate images ofall the places I've been to see
how accurate it could make them.
Like does it feel like when Iwas there, and what I did was to
(21:11):
create these really complexprompts.
For mid journey I uploaded aguide to chat GPT for oh, of
here's how to create prompts inmid journey, and then I
basically had chat GPT createthe prompts for MidJourney that
are like really detailed, withlike camera settings and all of
the details are like four orfive paragraphs long prompts,
(21:32):
and I created an Instagramaccount and put all the pictures
there and I'm like you knowwhat it looks, like these
pictures are from all thoseplaces, like it's pretty good If
you were doing a travel blog,for example, like it is pretty
good for a lot of use cases, Ithink for marketers and I think
it's going to get better.
Jess Gonzalez (21:48):
I love that use case.
Here's my question Are you asubscriber to MidJourney's print
magazine?
Eric Eden (21:55):
I'm not.
Jess Gonzalez (21:56):
It's $4 a month and it's a collection of, I
think people can submit theirimages to it and it'll have the
image and then it'll also havethe prompt that was used to
produce it.
And sometimes I will see abeautiful image and the prompt
is like half the page.
And there's other times whereI'm like that image is awesome
and the prompt is give me what ahot dog dreams of.
(22:17):
And I'm like I never getanything like that when I put
random shit in, like randomstuff, random shit in like
random stuff you have to use.
Eric Eden (22:28):
The weird thing about mid journey is that you have to
use discord to put in theprompts, which is really weird
because it's like a gamer chatapp and it's a very different
sort of experience because whenyou're putting in prompts, other
people are putting in promptstoo and you're in a stream
watching what other people aredoing.
And sometimes if I'm reallybored, I'll just go in there and
not put in prompts and watchwhat people are creating and the
(22:49):
prompts that they're using.
And it's very fascinatingbecause some people are really
sophisticated.
They put in like camerasettings and like the background
should look like this andthey're doing like product
photography and I'm seeing likerestaurants clearly creating
menu, like photos, and I'm likethat looks good enough to eat.
I'd eat the whole thing.
(23:09):
I'm like out of one to 10,that's 100.
I'm just like voyeuristicallywatching Midjourney images, so
that, and or the magazine, Ithink is a great way to get
inspired on how to do greatstuff.
Jess Gonzalez (23:23):
I think Midjourney is trying to make
more of a community around itand hoping that's the exact use
case that you see it, you getinspired.
We actually built a scraper sothat we could do an API call to
generate an image and return it.
We built a little affiliatemarketing tool, so essentially
it would take trending, it wouldtake high commission products
off of a couple of differentsites that give you affiliate
(23:43):
links and then it would useGoogle listing for that product
to pull in reviews and feedbackon it and they would use those
reviews and feedbacks withChatGPT to write the prompt that
would go to mid journey togenerate the images and the
headlines for the ads and thenit would automatically push it
into Airtable and then push itinto Facebook and run it as an
actual campaign.
And we in that process, like wecan't have someone like
(24:05):
manually sitting here in discorddoing this, and so there are
ways to make it automated.
It's not easy and you do loseaccess to seeing what other
people are doing with it, butthere are a couple ways to get
it there.
Eric Eden (24:19):
So where do you think this is all going with AI,
specifically for marketers andcompanies using AI in their
sales, marketing and theircustomer journeys?
Where do you think this isgoing?
I'm fascinated to see how thisstory goes.
Jess Gonzalez (24:36):
Yeah, so I'm going to give you the most
boring answer ever.
It goes back to marketingfundamentals.
What I'm noticing is there'stwo things.
There's a lot of developersthat are building AI tools for
marketers who have never donemarketing, who don't know what
marketers need.
They're the ones you see thatsay, you can make a hundred blog
posts a day, but they're notgiving you anything to make sure
(24:57):
that it's the right topic, orthey're not giving you tools to
figure out your content pillars.
They're just giving you volume.
They're like marketers loveblog posts, let me give them
more blog posts.
There's no strategy behind it.
And then, on the other side,you have people that are just
wanting to consume everything AI, but they don't actually have a
system.
Let's say that you decide tobring on a tool to help you do
(25:18):
outbounding so you can do morecold outbounding with AI.
What happens when someoneresponds?
Does it go into an actualprocess where your rep is
alerted?
Does it do any scoring?
Are you qualifying anyone?
If you're missing marketingfundamentals, all you're doing
is amplifying tactics andthey're only going to get you so
far.
The people who really win inthe age of AI are the marketers
(25:41):
who understand how to domarketing well and use AI to
help amplify what they'recurrently doing, not the
marketers who just use it as away to get more stuff done and
just put sometimes completetrash out there with it.
So that's my take Got to knowthe fundamentals.
Eric Eden (26:00):
I agree a hundred percent.
I think that AI is a great tool, but it can't just replace
everything you're doing.
You have to give it at thisstage the direction.
It can be a really greatresearch buddy or a co-pilot,
but it may not be ready to be afull agent.
And I think of the movie theMatrix and the agents just
(26:20):
popping in and they take care ofthe task completely without
direction, just popping in andthey take care of the task
completely without direction.
I don't think we're definitelynot there right now.
The question is how long willit take for us to go from a
stage of AI is all co-pilot tothere's AI agents that can do
certain use cases and they cando it all the way through or
(26:42):
almost all the way through.
It just gets a human approvallike book a trip according to
your specifications, book ameeting according to your
specifications.
It seems like we'll get to thispoint where there'll be agents
and there'll be some pain ofgetting there where it doesn't
right, but eventually there'llbe some agent processes.
But there'll be some processesto your point that almost have
(27:02):
to be all co-pilot your pointthat almost have to be all
co-pilot.
But I think the main thing thatI've seen is that a lot of
people are misunderstanding thatprobably 90% of things will be
co-pilot and then 10% of thingswill be agents and then 0% of
things have no impact of AI.
I think a lot of people arethinking, oh, ai only impacts
some things.
(27:22):
But I think it's going toimpact everything in different
sort of adoption levels and then, as time goes on, it'll
probably just get better andbetter.
Jess Gonzalez (27:30):
Yeah, this might be a little bit controversial.
I'm sure someone's going tohear this and absolutely hate it
.
I think of AI like I think ofinterns.
You have an intern coming inwho, let's say, they're a
sophomore in college.
They've taken some marketingclasses, they taught themselves
how to use Canva, but they'venever built a full marketing
campaign, they've never done outof home and airport and digital
(27:51):
and had to have it cohesive.
They haven't done this stuff,and so you have to spend time
with your intern, showing themthe ropes, sharing your
experience with them, workingwith them so they can learn how
to do it.
It's the same thing with AI,and if you're willing to put in
the time and energy to trainyour models to understand how
you're thinking, they're goingto be able to do work similar to
(28:16):
what an intern can do, but Ithink that they have more of a
ceiling.
That's limited because they'renot going to be able to
necessarily think creatively thesame way the human does with it
.
Eric Eden (28:24):
Interestingly, I met the CEO of the software company
Descript and his team at thePodcast Movement Conference in
DC this week and they have oneof my favorite AI software tools
that allows you to edit audioand video, and they came out
(28:46):
with their latest release of AIcapabilities and they dubbed it
this is your AI intern.
Jess Gonzalez (28:54):
Oh, I love it.
Eric Eden (28:56):
Named it the perfect name Underlord.
Oh my God, not Overlord,because it doesn't rule you yeah
.
You rule it and it's a funnyjoke of it's not quite ready for
prime time to just do things onits own.
That's a joke they're gettingat and they just had this very
playful culture with theirproduct videos where they lean
(29:16):
into.
This is where we are in thedevelopment phase of this, which
makes people like me intoraving fans, right.
Jess Gonzalez (29:23):
Yeah, I love that .
I think a lot of people thinkAI will somehow magically give
you the perfect output, and Ithink back to 10 years ago,
working with writers which, eric, I know that you've done too.
There's shitty human writers,there's bad human writers who
cannot write well, who cannotadd emotion and thought to
content that they're producing,who don't add POVs, and so
(29:45):
people think AI is going toproduce this perfect thing, and
then it doesn't, and thenthey're like AI is useless and
they throw it out.
But there is a sweet middleground between those, and if you
can figure that out and you canstill work on honing your craft
, it'll make you an unstoppablemarketer.
Eric Eden (30:01):
So what is the use case that you're going to work
on in the near future, thatyou're excited about with
marketing and AI?
Jess Gonzalez (30:10):
So I think AI is really fantastic at ingesting
unstructured data and being ableto turn it into insights and, I
think, a huge challenge andI've advised a handful of
attribution companies during mytime.
I have a master's in it.
I love data, which is probablynot a shocker based on my
personality so far.
I think AI is really good attaking it, structuring it in
(30:33):
ways that maybe you didn't thinkabout, organizing it, doing
core analysis, and so what I'mworking on now is trying to
build a system to better helptake in all this data that our
clients have in secure, notgetting shared with chat to UPT.
But how does it take in thatdata and how does it structure
it differently?
(30:53):
Or help us think about it justfrom a different perspective to
be able to get differentinsights that are going to help
us figure out what we can do toreally drive the needle, to
drive more revenue withoutspending more money so much data
(31:18):
for marketers to absorb in yourbrain across all these
different systems and silos, andtry to manually coagulate it
together is mind-numbingly hardsometimes.
Eric Eden (31:24):
I think that's awesome.
Jess Gonzalez (31:26):
I never want to write SQL again.
That's my life goal Me either,yeah.
(31:47):
So what are your top threepoints of advice for marketers
related to AI?
Don't need to change everythingyou're doing to fit into what
AI can do, because AI in four orfive months is going to be able
to do different things than itcan do today.
Ai a year from now is going tolook totally different, so think
about how you can start pullingAI into your existing systems
(32:08):
to really amplify the good thatyou're already doing.
That's like my number one pieceof advice.
I feel like that reallyencompasses everything.
Number two would be to test Trydifferent tools, because a lot
of tools do different things.
I personally would say ifyou're thinking about a tool,
look at who built it, becausesometimes the tools will change
your behavior and how you thinkabout things.
(32:30):
Use tools built by marketersfor marketers.
It's going to get you a lotfurther than these tools that
are built by developers justbeing sold to marketers.
And number three is I wouldencourage anyone in leadership
to encourage AI in your teams sothat everyone at your company
and your team feels comfortabletesting and embracing this new
technology.
That can really help them bebetter at their job, not with a
(32:53):
focus on using AI to replacepeople or AI to replace systems
or other tools and technology.
Eric Eden (33:01):
Yeah, it feels like we're in a really interesting
testing point where I wouldn'trecommend pros and the cons of
it are and then finding out howto get efficiencies, Because
(33:24):
there'll probably bebreakthroughs in the near future
Not in years, but probably inmonths.
Oh yeah, that if you're notdoing this sort of testing and
understanding how the tools workand understanding what the pros
and cons are today, you'll bebehind when these things come in
the very near future.
Right?
Jess Gonzalez (33:41):
Yeah, ai today is totally different than AI was
this time last year.
Eric Eden (33:47):
That's bonkers to think about, isn't it?
Jess Gonzalez (33:50):
It's moving fast.
Eric Eden (33:51):
And any final thoughts.
Jess Gonzalez (33:54):
What I would say is what I will offer is if
anyone ever wants to talk AI orhas questions, I'm always happy
to talk about this stuff.
I know it can be scary also tomeet a new tool that can be hard
to understand until you getused to it.
So I'm on LinkedIn.
I'm pretty responsive.
Feel free to hit me up.
If anyone does want to talk AIat all, I am happy.
(34:17):
I'm always happy to chat shop.
Eric Eden (34:21):
Awesome.
Thank you for that offer.
I'm going to link to yourLinkedIn and your website in the
show notes so people can easilyget there.
We really appreciate yourstories, your insights and your
advice today.
Thanks so much for being withus.
Jess Gonzalez (34:32):
Thanks for having me.
Today we are talking about marketing and AI what is
the future of marketing and AIand we have a great guest to
help us talk through that.
Jess, welcome to the show.
Jess Gonzalez (00:10):
Thanks for having me.
Eric Eden (00:12):
Why don't we start off by you taking a minute to
share a little bit about who youare, what you do, what you've
done?
Jess Gonzalez (00:20):
Yeah, hi, I'm Jess.
I love marketing.
I've been a quota carryingmarketer for almost my entire
career, so I come to marketingwith really a revenue and
process focused lens.
I've spent my career in I wouldsay, really fortunate once of a
career roles where I've workedwith some of the largest media
companies, publishers, some ofthe fastest growing SaaS brands
(00:41):
to really grow their companymore quickly by growing
marketing quickly, with a focuson driving revenue.
And lately I've been doing alot with AI.
I have been just a caveat.
I've been using AI since beforeit was a hot thing and before
ChatGPT was launched, so I'vebeen in AI for about a decade,
but I've been finding somereally interesting ways to use
(01:03):
it that I think not everyone'sleveraging it for.
Eric Eden (01:07):
Awesome, and you're also a podcaster.
You have your own podcast.
Jess Gonzalez (01:12):
Yes, marketers Talking and Marketers Talking AI
.
Where we bring on marketerstalking, we bring on marketers
of all walks of life anddifferent seniority and
experience levels to share theirstory and what they have found
that works for them in themarketing world.
And then marketers talking AI.
My co-founder, pavel, and Icover the recent AI news and we
are talking to founders of AIMartech specifically, founders
(01:35):
who are actually marketers, notjust developers, who are
building AI tools to sell tomarketers.
So that's an exciting serieslaunching soon.
Eric Eden (01:44):
All right, we're ready to be inspired.
Tell us a story about some ofthe best things you've done with
marketing and AI so far.
Jess Gonzalez (01:52):
Yeah.
So my favorite use case and I'mgoing to start with content,
which for a lot of people,they're probably saying yeah,
content.
I know there's 50 tools thatwill make me a million blog
posts a day.
It's everywhere.
We took a little bit differentapproach, but we've been using
it heavily for contentproduction and curation.
So, specifically, we had aclient that was starting off
(02:13):
fresh, brand new website, nodomain authority at all, and
they wanted to be able to growand really engage in the startup
community.
Their ICP was founders, ideallyfounders who are going through
incubators or some other earlystage funding.
So they want to get in front ofthis specific audience and they
have nothing on their websiteto really make them compelling
(02:34):
as an authority on the topic yet.
So we launched a programmaticSEO effort, which means that we
produce a ton of SEO optimizedcontent, and we did it all
really quickly using somedifferent workflows, processes,
tools, however you want to do it.
So, instead of manuallycreating 200 pages, we had a
(02:55):
system to be able to do it usingAirtable, makecom, retool,
zapier a couple tools similar tothat.
So we started off with askingourselves what does this target
audience want to engage with?
And one of the spots that I'vegone to for almost my entire
career for inspiration is Reddit.
So we actually set up a processthat it would scrape some
(03:17):
different subreddits every dayand it would see what are the
three or four most interestingarticles that have been posted
in the last 24 hours, that aretrending, like what's getting
the most traction.
And so we use Reddit to curatethe topic pool.
And then we took those topics,put them into an engine to get
some different headlines out ofit.
(03:38):
So we would take, let's say, 10topics that are doing really
well, let's distill that downinto three that we care about,
and then for each of thosetopics let's generate 10
different titles, and then wewould have a human in the loop.
Human in the loop is a termthat not a lot of people may be
familiar with, but it means thatyou have an actual person who's
(03:59):
taking your AI output andthey're looking at it to refine
it before putting it back intoAI to get to that next step.
So our human in the loop wouldgo in and they would look and
see those different titles andthey would say, okay, these
eight, I think, are reallycompelling.
I like these two, but they'renot quite great.
Let me go regenerate them ormake some changes, and then we
(04:20):
would take those titles and wewould use those to produce
content, but here's where itgets different.
You can't just put that titleinto ChatGPT.
You can.
You're going to get something,though, that really sounds AI
generated, and so what we did iswe really thought about what is
the unique point of view thatour client has, what is their
(04:43):
opinion and ideas.
We want to make sure is included, and are there different areas
that we know that they playreally well.
So, for example, if you were acompany that does, let's say,
software development, you'd wantto make sure that topic tied
into software developmentsomewhere, and it might not
always be super obvious, but AIis really good at making stuff
up, and so we would tell it.
(05:04):
Here's our topic.
We want to make sure that thetone is optimistic, and here's
our super long explanation ofthe brand voice so we can make
sure the content actually soundslike it was written by a
specific person, and we want tomake sure that it ties into the
future of software development,and these are tech stars funded
companies.
We want to mention tech starsand maybe find a couple
(05:26):
references to these otherpublications, so we're giving it
a really robust prompt thatliterally took weeks to develop
and months of fine tuning, andthen it puts out a piece of
content that then goes to aperson to look at Quick glance.
Does it look like it's in theright direction, yes or no?
And if it is, it'll go back intoa system where we actually have
(05:47):
different AI models talking toeach other to provide feedback
on the content to each other,and so it's like having a team
of three or four copy editorsthat are coming in reviewing and
providing feedback, but it allhappens in like 20 seconds and
it goes back and forth.
And then it would spit out thisarticle that sounds pretty good
(06:08):
.
It's 95% of the way there, andwe actually eventually added in
a step where it would alsoprovide feedback of what it
thinks should be tweaked in thearticle.
And then that goes to anotherhuman in the loop who's an
actual copy editor, actualwriter, who looks through, makes
some changes, make sure that itfeels human.
It sounds human Also, that allthe data is correct in there.
(06:29):
Ai does hallucinate.
We try and get rid of that asmuch as possible, but sometimes
it will make up its own factsand even put in a citation for
it that doesn't exist.
So they give it a once lookover and then it's ready for
publishing.
And so by doing that, we'reable to make a system where it's
automatically finding trendingtopics on Reddit and producing
(06:52):
10 to 20 blog posts in one ortwo days, which means in a month
you can put out hundreds ofpieces of content that are all
unique, all designed to speak tospecific keywords and following
topics that are resonating withyour audience.
So it's not just using it forthat first content production
piece, but it's thinking abouthow can AI give me leverage
(07:14):
across what should be an entirecontent production process?
And we've taken that.
We built it for a client, wetweaked it, repeated it, built
it for another client, tweakedit, repeated it, built it for
ourselves too, tweaked it,repeated it, built it for
ourselves too, so we could putout more content.
And we found that by reallyputting out this content, it's
helping with domain authority.
So one of our clients went fromzero to two in a day and then
(07:36):
they've gone up to about 12within a week and now they're
close to 30.
But more importantly, it'sattracting the right audience to
them.
That's turning into inboundrequests, that's turning into
deals and opportunities, andthat's ultimately what we care
about is building marketing thatdrives revenue.
Eric Eden (07:55):
It's really almost fueling your inbound engine that
wasn't existing before thewebsite was built was built and
is it working that it'sgenerating this volume of
content with this machine andit's getting indexed, it's
getting ranked?
Is that piece of it working?
Jess Gonzalez (08:13):
Yeah, I will say the bottleneck on all of it.
There's a couple spots.
One is we insist on humaneditors in the process because
we want to make sure it's goodquality content.
That can bog it down a littlebit depending on the length of
the content, and Google doesn'tjust auto magically index every
single thing you give to it.
So we've had a couple instanceswhere it's taken a couple weeks
(08:35):
to get all of the contentindexed.
Typically, we'll see a goodportion of it indexed up front
at least three or four new pagesa day.
At the low end we had oneclient that we were surprised.
Day one we put it out.
It was a pretty large piecethat was really more of a pillar
page of content, and the nextday they had an opportunity to
come in from it, which isastonishing, because sometimes
(08:57):
Google does take a bit topublish.
What I will say is we haven'tseen any instances where
Google's penalizing using AIgenerated content, because I
honestly think Google can't tellthat it's AI generated, because
it's not bad content.
Eric Eden (09:15):
Yeah, I think we'll get to a point one day when
there'll be really gooddetectors all around that'll be
able to tell which content wasgenerated with AI and perhaps
what percentage of it wasgenerated with AI, but I think
the only one who perhaps hasthat tool?
There was this scandalous storyabout how OpenAI has had this
(09:35):
detector internally but hasn'treleased it.
But because they haven'treleased it, no one really knows
.
They say that they had thisinternal tool.
They could tell for 99%, butprobably Google doesn't know,
like you're saying, and it doesprobably take still just like if
you were just creating it allmanually.
It does take a little bit oftime Weeks sometimes can be 30,
(09:57):
60 days, even if you'reproducing a good volume of
content for it to get indexed.
That's what I've seen acrossmany projects over time.
It's never instantaneous, sothat probably is still there,
like you're saying.
But I think that's fascinatingbecause I worked for one company
just to give the manual analogynext to that, where we did
generate a massive amount ofbusiness this is five years ago
(10:18):
from creating a thousand blogposts that were each like 2000
words long, written byfreelancers, and I can't tell
you how much brain damage camefrom that process of humans
writing 2000 word blog poststhat didn't weren't keyword
stuff, that were actually decent, and then trying to get three
(10:41):
or five of them a week andeventually, over time, over a
period of two years, we got toenough where we were getting
like hundreds of thousands ofvisitors a month to our website
that were going into oursoftware free trial from the
blog.
But we spent years of manuallabor and a lot of money paying
freelancers to do that and itwas good, but this is a lot less
(11:03):
painful way to do somethinglike that.
Jess Gonzalez (11:05):
When you think about the cost.
So, related to this project, wehad a client that was doing
programmatic landing pagesspecifically for ABM.
So they had 40 accounts thatthey wanted to do landing pages
for and the landing pages werepretty uniform, except we're
just going to change the companylisted on the top, we're going
to change the URL and maybe fiveor six pieces of content we're
(11:29):
going to slightly tweak, butwe're not trying to index them.
So we don't care if it'sduplicative content because
they're really designed for ABMoutreach.
They had a quote.
If it's duplicative contentbecause they're really designed
for ABM outreach, they had aquote.
It was $30,000 for 50 pages andI was like, are you kidding me?
For someone to go and buildthem all and that's had them
getting like 80% of the content.
I was like that's crazy.
And so they were on HubSpot.
(11:50):
We rolled out HubSpot DB, madea spreadsheet and we did have to
set up the landing pagetemplate and the variables.
But it took an hour and a halfto do to build all the pages
when the quote was $30,000before, which just seems
ridiculous.
And if you look at how muchsometimes a blog post especially
on the B2B side costs, you'respending $1,000, $1,500 per blog
(12:11):
post coming out, so I wouldimagine that's a massive cost
savings for a lot of companiesto be able to build these
programmatically too.
Eric Eden (12:20):
Yeah, and this belt tightening era, just a lot of
people.
They just are like we can't doit.
That's where a lot of companieshave gone with it.
So I think personalization isthe use case that you're
mentioning, that a lot of peopleare starting to experiment with
, in the sense of creating alanding page for each target
account.
To do it manually is, like,really expensive.
(12:43):
As you laid out To do itautomated, you probably still
want to check over the work, butit's a lot faster than trying
to do it all manually and that'san awesome process.
Jess Gonzalez (12:53):
I think what AI is really great at that is
underutilized by people isgiving you options, giving you
something to react to.
So if super boring example youneed to figure out meals for a
week, it's a lot easier for usas humans to look at our options
and say, okay, monday night youcan have one of these four
things what do you want?
Easier to pick than to say,okay, what do you want.
(13:16):
Monday night it's so mucheasier to make a decision and so
using AI or even some automatedsystems to help you figure out
what your options are and thenbe able to make that decision
more rapidly.
I think that's one of the mostunderutilized areas for AI right
now for marketing.
Eric Eden (13:31):
I agree.
One of my favorite use casesalong those lines is in using AI
as a co-pilot.
When I'm doing messagingarchitecture projects, I have it
take like the about us for acompany.
I'm like, take this and rewritethe about us for the company in
10 different styles and I willcome back with a conversational
(13:52):
one, a formal one, aninspirational one in 10
different ways.
And then it's very interestingto share those options with
executives and see which oneresonates with them the most,
based on their company culture,because it's the same fact set
as like their core about us thatthey had.
But people are like I didn'tthink about it that way, and so
I think it is really great, as aco-pilot, of giving options of
(14:13):
different routes to go.
Like I said, I think that's agreat tip of giving options of
different routes to go Like Isaid.
I think that's a great tip.
Jess Gonzalez (14:17):
My co-founder, pavel, and I.
We use AI as a co-pilot inalmost every meeting.
We have every brainstormingsession and sometimes what we
find is that it will in thenotes, it will pick out things
that we didn't realize we talkedabout, or it'll put down action
items where it'll say based onthese other action items, you
might want to also considerthese couple things and it's
(14:39):
been really helpful for that,where it's almost like having
another brain in the room.
And on the feedback side,similar to what you're talking
about with the messaging, takethat messaging, put it in AI and
ask AI to tell you who thetarget audience is.
I found that I've written somestuff that I thought was like
really fantastic and it did notconvert and I put it in.
(15:00):
I was like what's going on here?
And AI told me that my copy wasfor a totally different
audience and I was like, oh shit, I guess I missed a note here
somewhere.
But I wouldn't have realized itwithout AI being that kind of
totally I would say objectiveperson that's able to come in
and give you feedback, notknowing what's in your head for
the context behind it.
Eric Eden (15:21):
So you're doing some really advanced things here in
August 2024.
What's your marketing?
Ai tech stack.
Jess Gonzalez (15:30):
Honestly, I love Claude and Perplexity.
Chatgpt, of course, is in their4.0.
But I think Claude andPerplexity ChatGPT, of course,
is in their 4.0, but I thinkClaude and Perplexity are two of
the best tools that people canuse.
Personally, I'm a fan of Claudemore for writing.
I think that it produces morehuman feeling content than
ChatGPT, unless you have.
(15:51):
If you take the time to make asuper robust prompt, chatgpt can
get you there too, but I feellike Claude, out of the box,
will give you a better writingproduct than ChatGPT will.
So I'm a fan of Claude forwriting and then sending that
Claude output over to ChatGPT togive it feedback on.
I love perplexity for search.
(16:11):
I feel like it's giving meanswers to things that I asked
the most, like I asked it themost random question that's not
super specific and it somehowknows exactly what I want, which
feels magical.
And then, when we look betweenthe tools, where I think they
really become powerful is howwe're automating between them
and so, using Zapier, retool,makecom to try and find ways to
(16:35):
pull it in.
I'll give an example For mypodcast.
I have an entire workflowbetween ClickUp and HubSpot and
then I'll actually take thetopic and I'll take the
transcript from the podcast andI'll send it into Summarize AI,
which is my favorite tool forvideo transcriptions and
summaries, and it'll pop out acouple of different title
(16:56):
options.
I can take those and send thoseback to Claude to give me other
options if I don't like it andfinding ways to really get those
tools to work together.
So Claude Can't RecommendEnough Perplexity.
Summarize are my kind of threefavorite, I would say.
Eric Eden (17:12):
Three go-to.
And what about for images orsoon-to-be video?
Any favorites in those areas?
Jess Gonzalez (17:18):
I'm going to tell you I so I'm going to age
myself a little bit here I don'tuse Photoshop.
I use Fireworks still, whichAdobe sunset in like 2012.
But I never took the time touse Photoshop.
But I've actually been usingthe generative AI in Photoshop
recently because it's so easy touse, so that's been my go-to.
(17:41):
I've played around with acouple different tools.
There's one that I've gone towhen I need to generate
something from scratch, whereI'm starting from absolutely
nothing, but I'm typicallystarting from an image that I'm
trying to edit, and so I'll gointo Photoshop for it, or I'll
actually use Canva, especiallylike headshots.
Sometimes clients have photosthat are a bit older and you
(18:01):
need to sharpen them up.
I've been really impressed withCanva's AI options for editing
photos.
Eric Eden (18:07):
Yeah, so is the Adobe tool.
Is that Fireflies?
Jess Gonzalez (18:13):
Yes, I think so yeah.
Eric Eden (18:16):
Have you found that to be decent?
I haven't tested that one.
Jess Gonzalez (18:18):
Oh, so I actually made my wedding invitations
with it.
I had a Call of Duty themedwedding, which, for those who
don't know which is probablyeveryone I met my husband on
Call of Duty, like all singlepeople in our 30s, and we wanted
a Call of Duty themed weddingbecause a lot of our gaming
friends were coming.
If you search Call of Dutythemed party stuff, it's all for
(18:40):
, like little kids and it's allarmy themed, and we wanted
something that I would describeas a little more up level, a
little more upscale, and I hadsome different ideas but I
couldn't quite get it to workand I spent four or five hours
trying to blend images togetherand just going in and zooming in
and erasing pixels.
Then I was like, let me justgive this a try, and I think the
video is actually still up onmy LinkedIn.
(19:01):
It took me less than 10 minutesto make our wedding invitations
and they ended up being totallycustom.
I wanted to blend a coupleimages it's two people
parachuting in and it saysdropping in together and so I
wanted to alter them so theylook a little more like the
operators that we play in thegame.
And it was the best processever.
It was the first time I used it.
(19:22):
It was so easy to use and, assomeone again who did not take
the time to learn Photoshopbecause I'm still using the 2010
tools I just didn't want tolearn it.
I found it easy to use, so Ithink that's saying a bit about
it.
Eric Eden (19:36):
Now I have to know wedding venue, reception venue
was it also Call of Duty themedor just the invites?
Jess Gonzalez (19:42):
Yes, so we turned into a LAN party, which is also
a common aging myself.
But yeah, we had it at ourhouse and then we had the
reception at a what do we callit?
It's a place that has a ton ofcomputers that are networked
together to have a LAN party.
It's like a nerd thing.
So, yeah, everything was reallyCall of Duty theme.
A bunch of our gaming friendscame in, so it was a good time.
Eric Eden (20:05):
Congratulations.
That's awesome.
So I actually have been testingmid-journey against Dolly to
see which ones I can get betterimages out of, and I think for a
lot of the marketing I do likeimages for blog posts right.
I don't need to spend hundredsof dollars for an image, an
abstract image, for a blog postwith one of the stock photo
(20:28):
companies, but I think peoplecan tell that it's AI in a lot
of cases.
But I've actually gottendiffering results.
But with ChatGPT Dolly, you canactually upload an image and
have it change it, which isinteresting.
Not quite Photoshop, but youcan say take this image and do
this with it, and a lot of timesit'll it, which is interesting.
Not quite Photoshop, but youcan say take this image and do
this with it, and a lot of timesit'll do some pretty cool
(20:49):
things and I've gotten somepretty good quality stuff out of
Midjourney.
I like to travel.
I've been to 63 countries andso all the places that I've been
to, I had it generate images ofall the places I've been to see
how accurate it could make them.
Like does it feel like when Iwas there, and what I did was to
(21:11):
create these really complexprompts.
For mid journey I uploaded aguide to chat GPT for oh, of
here's how to create prompts inmid journey, and then I
basically had chat GPT createthe prompts for MidJourney that
are like really detailed, withlike camera settings and all of
the details are like four orfive paragraphs long prompts,
(21:32):
and I created an Instagramaccount and put all the pictures
there and I'm like you knowwhat it looks, like these
pictures are from all thoseplaces, like it's pretty good If
you were doing a travel blog,for example, like it is pretty
good for a lot of use cases, Ithink for marketers and I think
it's going to get better.
Jess Gonzalez (21:48):
I love that use case.
Here's my question Are you asubscriber to MidJourney's print
magazine?
Eric Eden (21:55):
I'm not.
Jess Gonzalez (21:56):
It's $4 a month and it's a collection of, I
think people can submit theirimages to it and it'll have the
image and then it'll also havethe prompt that was used to
produce it.
And sometimes I will see abeautiful image and the prompt
is like half the page.
And there's other times whereI'm like that image is awesome
and the prompt is give me what ahot dog dreams of.
(22:17):
And I'm like I never getanything like that when I put
random shit in, like randomstuff, random shit in like
random stuff you have to use.
Eric Eden (22:28):
The weird thing about mid journey is that you have to
use discord to put in theprompts, which is really weird
because it's like a gamer chatapp and it's a very different
sort of experience because whenyou're putting in prompts, other
people are putting in promptstoo and you're in a stream
watching what other people aredoing.
And sometimes if I'm reallybored, I'll just go in there and
not put in prompts and watchwhat people are creating and the
(22:49):
prompts that they're using.
And it's very fascinatingbecause some people are really
sophisticated.
They put in like camerasettings and like the background
should look like this andthey're doing like product
photography and I'm seeing likerestaurants clearly creating
menu, like photos, and I'm likethat looks good enough to eat.
I'd eat the whole thing.
(23:09):
I'm like out of one to 10,that's 100.
I'm just like voyeuristicallywatching Midjourney images, so
that, and or the magazine, Ithink is a great way to get
inspired on how to do greatstuff.
Jess Gonzalez (23:23):
I think Midjourney is trying to make
more of a community around itand hoping that's the exact use
case that you see it, you getinspired.
We actually built a scraper sothat we could do an API call to
generate an image and return it.
We built a little affiliatemarketing tool, so essentially
it would take trending, it wouldtake high commission products
off of a couple of differentsites that give you affiliate
(23:43):
links and then it would useGoogle listing for that product
to pull in reviews and feedbackon it and they would use those
reviews and feedbacks withChatGPT to write the prompt that
would go to mid journey togenerate the images and the
headlines for the ads and thenit would automatically push it
into Airtable and then push itinto Facebook and run it as an
actual campaign.
And we in that process, like wecan't have someone like
(24:05):
manually sitting here in discorddoing this, and so there are
ways to make it automated.
It's not easy and you do loseaccess to seeing what other
people are doing with it, butthere are a couple ways to get
it there.
Eric Eden (24:19):
So where do you think this is all going with AI,
specifically for marketers andcompanies using AI in their
sales, marketing and theircustomer journeys?
Where do you think this isgoing?
I'm fascinated to see how thisstory goes.
Jess Gonzalez (24:36):
Yeah, so I'm going to give you the most
boring answer ever.
It goes back to marketingfundamentals.
What I'm noticing is there'stwo things.
There's a lot of developersthat are building AI tools for
marketers who have never donemarketing, who don't know what
marketers need.
They're the ones you see thatsay, you can make a hundred blog
posts a day, but they're notgiving you anything to make sure
(24:57):
that it's the right topic, orthey're not giving you tools to
figure out your content pillars.
They're just giving you volume.
They're like marketers loveblog posts, let me give them
more blog posts.
There's no strategy behind it.
And then, on the other side,you have people that are just
wanting to consume everything AI, but they don't actually have a
system.
Let's say that you decide tobring on a tool to help you do
(25:18):
outbounding so you can do morecold outbounding with AI.
What happens when someoneresponds?
Does it go into an actualprocess where your rep is
alerted?
Does it do any scoring?
Are you qualifying anyone?
If you're missing marketingfundamentals, all you're doing
is amplifying tactics andthey're only going to get you so
far.
The people who really win inthe age of AI are the marketers
(25:41):
who understand how to domarketing well and use AI to
help amplify what they'recurrently doing, not the
marketers who just use it as away to get more stuff done and
just put sometimes completetrash out there with it.
So that's my take Got to knowthe fundamentals.
Eric Eden (26:00):
I agree a hundred percent.
I think that AI is a great tool, but it can't just replace
everything you're doing.
You have to give it at thisstage the direction.
It can be a really greatresearch buddy or a co-pilot,
but it may not be ready to be afull agent.
And I think of the movie theMatrix and the agents just
(26:20):
popping in and they take care ofthe task completely without
direction, just popping in andthey take care of the task
completely without direction.
I don't think we're definitelynot there right now.
The question is how long willit take for us to go from a
stage of AI is all co-pilot tothere's AI agents that can do
certain use cases and they cando it all the way through or
(26:42):
almost all the way through.
It just gets a human approvallike book a trip according to
your specifications, book ameeting according to your
specifications.
It seems like we'll get to thispoint where there'll be agents
and there'll be some pain ofgetting there where it doesn't
right, but eventually there'llbe some agent processes.
But there'll be some processesto your point that almost have
(27:02):
to be all co-pilot your pointthat almost have to be all
co-pilot.
But I think the main thing thatI've seen is that a lot of
people are misunderstanding thatprobably 90% of things will be
co-pilot and then 10% of thingswill be agents and then 0% of
things have no impact of AI.
I think a lot of people arethinking, oh, ai only impacts
some things.
(27:22):
But I think it's going toimpact everything in different
sort of adoption levels and then, as time goes on, it'll
probably just get better andbetter.
Jess Gonzalez (27:30):
Yeah, this might be a little bit controversial.
I'm sure someone's going tohear this and absolutely hate it
.
I think of AI like I think ofinterns.
You have an intern coming inwho, let's say, they're a
sophomore in college.
They've taken some marketingclasses, they taught themselves
how to use Canva, but they'venever built a full marketing
campaign, they've never done outof home and airport and digital
(27:51):
and had to have it cohesive.
They haven't done this stuff,and so you have to spend time
with your intern, showing themthe ropes, sharing your
experience with them, workingwith them so they can learn how
to do it.
It's the same thing with AI,and if you're willing to put in
the time and energy to trainyour models to understand how
you're thinking, they're goingto be able to do work similar to
(28:16):
what an intern can do, but Ithink that they have more of a
ceiling.
That's limited because they'renot going to be able to
necessarily think creatively thesame way the human does with it
.
Eric Eden (28:24):
Interestingly, I met the CEO of the software company
Descript and his team at thePodcast Movement Conference in
DC this week and they have oneof my favorite AI software tools
that allows you to edit audioand video, and they came out
(28:46):
with their latest release of AIcapabilities and they dubbed it
this is your AI intern.
Jess Gonzalez (28:54):
Oh, I love it.
Eric Eden (28:56):
Named it the perfect name Underlord.
Oh my God, not Overlord,because it doesn't rule you yeah
.
You rule it and it's a funnyjoke of it's not quite ready for
prime time to just do things onits own.
That's a joke they're gettingat and they just had this very
playful culture with theirproduct videos where they lean
(29:16):
into.
This is where we are in thedevelopment phase of this, which
makes people like me intoraving fans, right.
Jess Gonzalez (29:23):
Yeah, I love that .
I think a lot of people thinkAI will somehow magically give
you the perfect output, and Ithink back to 10 years ago,
working with writers which, eric, I know that you've done too.
There's shitty human writers,there's bad human writers who
cannot write well, who cannotadd emotion and thought to
content that they're producing,who don't add POVs, and so
(29:45):
people think AI is going toproduce this perfect thing, and
then it doesn't, and thenthey're like AI is useless and
they throw it out.
But there is a sweet middleground between those, and if you
can figure that out and you canstill work on honing your craft
, it'll make you an unstoppablemarketer.
Eric Eden (30:01):
So what is the use case that you're going to work
on in the near future, thatyou're excited about with
marketing and AI?
Jess Gonzalez (30:10):
So I think AI is really fantastic at ingesting
unstructured data and being ableto turn it into insights and, I
think, a huge challenge andI've advised a handful of
attribution companies during mytime.
I have a master's in it.
I love data, which is probablynot a shocker based on my
personality so far.
I think AI is really good attaking it, structuring it in
(30:33):
ways that maybe you didn't thinkabout, organizing it, doing
core analysis, and so what I'mworking on now is trying to
build a system to better helptake in all this data that our
clients have in secure, notgetting shared with chat to UPT.
But how does it take in thatdata and how does it structure
it differently?
(30:53):
Or help us think about it justfrom a different perspective to
be able to get differentinsights that are going to help
us figure out what we can do toreally drive the needle, to
drive more revenue withoutspending more money so much data
(31:18):
for marketers to absorb in yourbrain across all these
different systems and silos, andtry to manually coagulate it
together is mind-numbingly hardsometimes.
Eric Eden (31:24):
I think that's awesome.
Jess Gonzalez (31:26):
I never want to write SQL again.
That's my life goal Me either,yeah.
(31:47):
So what are your top threepoints of advice for marketers
related to AI?
Don't need to change everythingyou're doing to fit into what
AI can do, because AI in four orfive months is going to be able
to do different things than itcan do today.
Ai a year from now is going tolook totally different, so think
about how you can start pullingAI into your existing systems
(32:08):
to really amplify the good thatyou're already doing.
That's like my number one pieceof advice.
I feel like that reallyencompasses everything.
Number two would be to test Trydifferent tools, because a lot
of tools do different things.
I personally would say ifyou're thinking about a tool,
look at who built it, becausesometimes the tools will change
your behavior and how you thinkabout things.
(32:30):
Use tools built by marketersfor marketers.
It's going to get you a lotfurther than these tools that
are built by developers justbeing sold to marketers.
And number three is I wouldencourage anyone in leadership
to encourage AI in your teams sothat everyone at your company
and your team feels comfortabletesting and embracing this new
technology.
That can really help them bebetter at their job, not with a
(32:53):
focus on using AI to replacepeople or AI to replace systems
or other tools and technology.
Eric Eden (33:01):
Yeah, it feels like we're in a really interesting
testing point where I wouldn'trecommend pros and the cons of
it are and then finding out howto get efficiencies, Because
(33:24):
there'll probably bebreakthroughs in the near future
Not in years, but probably inmonths.
Oh yeah, that if you're notdoing this sort of testing and
understanding how the tools workand understanding what the pros
and cons are today, you'll bebehind when these things come in
the very near future.
Right?
Jess Gonzalez (33:41):
Yeah, ai today is totally different than AI was
this time last year.
Eric Eden (33:47):
That's bonkers to think about, isn't it?
Jess Gonzalez (33:50):
It's moving fast.
Eric Eden (33:51):
And any final thoughts.
Jess Gonzalez (33:54):
What I would say is what I will offer is if
anyone ever wants to talk AI orhas questions, I'm always happy
to talk about this stuff.
I know it can be scary also tomeet a new tool that can be hard
to understand until you getused to it.
So I'm on LinkedIn.
I'm pretty responsive.
Feel free to hit me up.
If anyone does want to talk AIat all, I am happy.
(34:17):
I'm always happy to chat shop.
Eric Eden (34:21):
Awesome.
Thank you for that offer.
I'm going to link to yourLinkedIn and your website in the
show notes so people can easilyget there.
We really appreciate yourstories, your insights and your
advice today.
Thanks so much for being withus.
Jess Gonzalez (34:32):
Thanks for having me.
Popular Podcasts
1. Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.
2. Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations.
3. Crime Junkie
If you can never get enough true crime... Congratulations, you’ve found your people.