February 13, 2025 • 36 mins
Join us for an enlightening conversation with Guillaume Dumortier of Maestrix.AI as he unveils his journey from traditional marketing to the forefront of AI-driven strategies. Guillaume shares how he transformed his marketing approach from the early days of Google AdWords to today's sophisticated, data-driven methodologies. Discover the growth marketing fit framework, which emphasizes the seamless integration of marketing with product development, audience engagement, positioning, and conversion channels. Learn how creativity, data analytics, and behavioral psychology are essential in navigating the modern marketing landscape.
Listen to Guillaume's fascinating story behind the creation of Matrix, an AI tool inspired by the fear of obsolescence, designed to streamline marketing processes. He provides an insider's view on overcoming AI output challenges and the evolution of Maestro X, a push-button marketing solution that delivers professional-grade, structured outputs. This episode underscores the significance of comprehensive solutions, moving beyond simplistic prompt libraries to achieve impactful marketing results.
We also delve into the critical aspects of AI privacy and security, discussing how Maestro X ensures user input protection and intellectual property safeguarding. Guillaume offers insights into transitioning from a side project to a product with market fit, catering to diverse users, from agencies to solopreneurs. Explore the shift from Web 2.0 to Web 3.0, emphasizing the importance of continuous learning and adaptation in the rapidly evolving digital media landscape. Stay tuned for more insights on how AI is revolutionizing marketing and redefining the future of the digital world.
This podcast is proudly sponsored by USC Annenberg’s Master of Science in Digital Media Management (MSDMM) program. An online master’s designed to prepare practitioners to understand the evolving media landscape, make data-driven and ethical decisions, and build a more equitable future by leading diverse teams with the technical, artistic, analytical, and production skills needed to create engaging content and technologies for the global marketplace. Learn more or apply today at https://dmm.usc.edu.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
Welcome to Mediascape insights from digital
changemakers, a speaker seriesand podcast brought to you by
USC Annenberg's Digital MediaManagement Program.
Join us as we unlock thesecrets to success in an
increasingly digital world.
Speaker 2 (00:22):
Welcome to 2025 Mediascape.
I am thrilled to have somebodywho I've known for almost a year
, whose tool I use in my classesin the digital media management
program to introduce some ofour students who have less
knowledge about branding andexecuting AdWords concepts for
(00:44):
brand integration with yourmarketing and your advertising.
So, guillaume Demortier ofMaastricht AI, thank you so much
for being here today.
Speaker 3 (00:54):
Thank you, Annika, for having me on.
It's a great pleasure.
Speaker 2 (00:57):
Yeah.
So your background.
You have had over 20 years inmarketing, working with big
firms, big brands that we allknow and love and probably use
every day.
Can you talk about yourexperience transitioning from
working in agencies, workingwith these big brands, to then
realizing that there is a holein the market that you needed to
(01:19):
fill, and how you've been ableto do that with artificial
intelligence?
Speaker 3 (01:24):
All right, that sounds about right.
So I'll try to keep it shortfor that first question.
But yeah, so marketingbackground from you know,
academic and practice obviously,but it was back.
I graduated back in 2005.
So that's 20 years ago thisyear.
So the marketing landscape wastotally different.
I mean, if we're talkingdigital marketing, it was just
(01:47):
Google and AdWords and periodright, so nothing too exciting.
Still, the keyword foundation,we'll talk later, but this time
the marketing experience wasreally about CPG, traditional
product marketing forsupermarkets, and was promotion
and things like that.
So not too much the scale ofdigitization at the time.
(02:07):
But then the social networksappear, like the Facebook, the
Twitters, then the technicalmarketing with the pixel
tracking and all thisdata-driven approach to
marketing that has kind of takenover the entire way.
We're kind of deciding andmaking arbitrage on where to
spend the marketing dollar thatwe thought so hard to get.
(02:29):
And so first, and being based inSilicon Valley for the past 17
years, it was very important toformalize the approach to
marketing as much as equal asthe product.
So the product is the star,right, it's mainly
product-driven and the notion ofproduct-market fit kind of
epitomizes this view thateverything revolves around the
(02:51):
product right.
But it would be a big mistaketo forget marketing as kind of
the vehicle, or at least part ofthe tooling of the vehicle,
that brings growth right,whether it's into the way you
present the solution or the toolthat you're building, or it's
really how you want toexperiment against the market
(03:13):
and see how the market resonates.
And so there was a need to feel, which is putting marketing
back in the equation of, youknow, building a product,
building a brand and whatnot.
And, on that note, all theyears starting 2012 where growth
hacking was a very kind of verymuch buzzwords.
(03:34):
My explanation and personalview on that is that those are
all the product people andengineers that were actually
doing marketing, building it inthe product, without wanting to
say that they were doingmarketing right, so kind of a
product-nested marketing that wecannot call marketing.
And so that's kind of buildingthe growth marketing fit
(03:56):
framework.
So that's the marketingframework I've designed,
revolving around audiences,around positioning, around the
content, the user, the buyer'sjourney, the conversion channel.
So all these ingredients andbuilding blocks, if you wish,
for the growth marketing fitapproach.
So that's great.
I've run that for 10 years as anagency with big groups, et
(04:19):
cetera.
So, to start to answer yourquestion, working with big
branches, everything they don'tknow, whether they have a kind
of a scouting department orsomething that is hooked to the
pulse of what's happening, whichis almost the case via agencies
.
Well, they don't know what.
They don't know right, and theway I've been trained
(04:41):
academically as a marketer hasnever reflected in a bigger
group or bigger company.
As kind of a practitioner, Iwas more of a brief drafter for
agencies to execute the jobright.
Today, the marketers in bigcompanies they're expanding
their resources but they'redoing briefs all the time and
chances are that a brief fromone vendor to another or from
(05:04):
one project to another slightlychanges along the way, so it's
never ending unsatisfactory kindof ecosystem, or I would say
dynamic, if you will.
So that's one remark.
But I think framing marketingas a discipline that has kind of
a predictable workflow andwhere you accept that it's less
(05:24):
about being creative, as peoplethink it is, but having the
notion of psychology andbehavioral psychology, having
the notion of data analytics,everything revolving around
numbers and also gut feeling,being creative yourself, meaning
having taste, I think that's agood transition in the world of
(05:45):
AI, because the next chapter ofthe story is taking that
framework in the area of AI, andthat is giving you maestrics,
which is having all those tasksand building blocks and
ingredients that help you domarketing on a day-to-day basis,
whether it's from strategicthinking to the tactical
(06:07):
execution of things.
Well, it's about translatingevery of those elements into
AI-driven or AI-enabled promptsthat put you in the position of
having taste as to whatresonates the most, as to what
the AI has helped you formulate,or at least accelerate, some
(06:27):
very structured way of lookingat marketing.
Speaker 2 (06:30):
A couple of things that you've mentioned.
I also used to live in the BayArea doing launch marketing for
publications and coming up withinteresting concepts that would
attract both consumers andadvertisers to the publications.
It was more on the creativeside.
(06:51):
I didn't work with a lot ofdata and have that construct.
Of course, unless you thinkabout in terms of how much ad
revenue were we bringing inright, those were the kinds of
things that were our markers.
More than are we reaching theright audience?
You know, this is what we sayis our demographic and
psychographic, but is this, youknow, really matching?
(07:12):
Is our audience really matchingwith these advertisers?
We didn't really get into thatas much.
It was really more aboutshowing the differences between
products, for instance, if itwas Xbox versus PlayStation,
because I worked on a lot ofvideo game publications and at
those times, to your point, weused to think of data in one
bucket right data analytics andthen you think of the
(07:35):
storytelling and creativity inthe other bucket.
But the truth is they reallyhave to be blended because
otherwise you don't get the goodresults, and that's what I
really love about Maestrics.
I tried a lot of AI tools.
I tested a lot of tools andwhen we first met and I tested
out Maestrix and I know it's youcontinue to iterate and ideate
(07:55):
and add to it I had not seen aproduct where I could literally
put in a website or put in adescriptor for a brand and then
have various personas based onthat category, based on
competitive analysis.
See the competitors andcompetitors that you wouldn't
even think about necessarilyRight, get the psychographics,
(08:18):
think about getting the keymessaging down to even receiving
.
Here are some campaigns youcould run for your advertising
and here are the keywords andthemes and here are some things
you could do on your socialmedia, your blog posts, your
website, along with KPIs.
So you really created this veryrobust ecosystem for any
(08:38):
practitioner, whether they haveno experience or lots of
experience to test out.
So I think that's just a reallyinteresting construct, because
not a lot of people have putsomething together as robust as
you have.
So I'd love to hear what did ittake to get from the very
beginning?
How long were you working onthe product before you took it
to market?
And then how did you come upwith each iteration?
(09:00):
Was it customer feedback?
Was it that you saw somechanging trends in the market?
Because we all know AIabsolutely something as
marketers we have to know.
We have to know how to use it.
We have to know how to use iteffectively, and there's so many
things coming out, so manydifferent tools you can use, as
well as all of the trends thatwe see with AI personalization,
(09:21):
how it's helping with ad cycles.
Speaker 3 (09:23):
Well, first, thank you so much for your phrases.
That's what fuels me to keep onbuilding and iterating.
So a lot to unpack here.
But basically the first thingthat drove me to build Matrix is
Fubo, I think.
As marketers, you all know Fubofear of missing out.
That was the Web 2.0, socialweb era and the era of AI.
(09:47):
It's FOBO Fear of BecomingObsolete.
And so if you take the two-yeartimeline that coincides with
ChatGPT 3.5 being released twoyears ago.
That is the timeline.
Then what is my prism withinthis?
So that's the following.
So first it's realizing that itwas a fascinating tool, just in
(10:09):
a couple of instructions to geta ready, fantastic results, but
unstructured, generic, butstill just the speed.
And the magic wand effect wasjust kind of when we say, okay,
there is something here thatforward to the artist and there
are going to be people on theright end of history and people
who are going to be laggard, andI by nature I'm very curious
(10:31):
and builder and I had a coupleof these in the past.
But here that was kind of adefining moment and say things
are going to be big.
And so that's when I started totake very specific tasks that I
was doing with my agency, sotaking landing pages.
And then I started to have somelayouts that say, okay, this is
a pattern model that I canprovide as a type of output.
(10:54):
And these are the first coupleof prompt engineering sessions.
Basically that's what it wasprompt engineering for marketing
.
But I was still frustrated withthis back and forth because I
realized that you could derailthe AI very quickly.
So it got better as we go.
You know time is extending.
Ai got better so it was morestructured.
There was GPT-4 in the spring.
(11:16):
So you know there's going to bethree years of GPT-4, which is
kind of really when the modelstarted to be super stable and
incredibly useful.
And so I ended up transformingthe growth marketing fit
framework into a collection of30 prompts roughly, but that
were a couple of paragraphs, sojust one or two paragraphs, but
(11:40):
starting to be detailed.
But I still had this copy andpaste.
So I need to go on my Notiondatabase paste there, go on the
product description that isnested somewhere else.
That needs to be unvariable,otherwise I know that I'm going
to derail the output by a coupleof degrees, but still that
matters.
And I said there has to be abetter way of this copy paste
(12:03):
etc.
And then they came up with thecustom GPT.
They say, oh great, now I canpackage my instructions into
this custom GPT, which I did.
I did.
If you look up for my street GPT, it's an all-in-one marketing
person agent.
Sorry, that is doing a workflow.
I don't remember exactly whatthe workflow is, but I was super
hyped by that and in themeantime I took a zero-prompt
(12:27):
approach, meaning that insteadof having conversations, I
wanted to have massive outputdelivery that was kind of
structured in a way.
That is what expected from amarketing professional
perspective what you would payan agency for, basically, or a
very good consultant or asubject matter expert, right,
(12:47):
and having this level ofthinking of the delivery output
and thinking, okay, deliverygrade in terms of structure,
what's in it has to be refinedright, but providing 80% of the
starting material was my goal.
So I start to merge thoseprompts in bigger zero prompt
shots so there would be no moreconversations, but still having
(13:09):
the custom GPTs as kind of thevehicle of distribution.
And last year so a year ago, itis the same time I started to
build Maestrix because the GPTstore was such a disappointment
that I couldn't find a viableway of distributing all I've
worked on in terms ofstructuring prompts.
And in the meantime you hadinfopreneurs selling the 10,000
(13:31):
prompt library PDF for 100 bucks, which again was ruining and
adding to the noise.
That was unnecessary at thetime.
Still, you know, polluting thesheer native and nascent
category of prompt engineeringproducts, related products or AI
products as a whole.
But last year.
So I decided I would build on ano-code platform, so I would own
(13:54):
the entire stack, so I wouldskip the step between product
requirement documents and thedeveloper.
So, and you know, having theknowledge of what has to be in
the code as a subject matteraspect, subject matter expert.
And so the secret sauce of thatis I have structure based on 20
(14:14):
years of marketing and thistool has been built for me first
with my knowledge, in the way Istructure my outputs.
So it's professional grade fromthe get-go and it's structured
in a way that it hits all theconcepts that revolve around the
marketing foundation, whetherit's positioning, when you're
talking strategy, or by yourjourney, when it's about
(14:36):
audience research.
You have also specific elementsand components that pertain to
it.
So really, the idea is to beable to prompt engineers, so
really code the instructions soyou don't have to.
And today, how it translates inMaestro X is it's a push-button
marketing.
Basically, all you have to dois create your product seed,
(15:00):
which is you describing it?
Or just a URL, as you said?
But then, after all, the tasksare seed-centric, meaning it's
just push button, you don't haveanything else.
So that's for the long answer.
Speaker 2 (15:15):
It's a great explanation and I think that's
what a lot of people miss.
People think it's as easy as,like you said, downloading a
list of prompts.
It's as easy as, like you said,downloading a list of prompts.
I know that there are thingsthat you can do with fine-tuning
prompts, with uploadingdocuments that speak in your own
voice or your brand's voice,but that's still not going to
(15:35):
give as much robust informationas a tool like Maastricht will.
Speaker 3 (15:42):
It combines the structure, it combines the
templates that you're talkingabout, but then there is a
variable component, which ismore how do you keep up with AI,
which are the models?
And so, again, I think peopledon't know the difference or
can't take the time to evaluatethose prompts against different
models, because, first, whywould they do that?
(16:04):
I can't do it for them.
So that's why, in Maestrics, Idon't ask you which model you
want to use.
You're using the most advancedmodel, regardless, via the APIs
that I've built behind thescenes.
So the story I can tell is it'sa blend of GPT-4.0, so OpenAI,
cloud, anthropic perplexity.
(16:25):
But more recently, and I thinkmaybe, if you don't go back to
Friday, then you should, becauseI've replaced 80% of the models
with Gemini Flash 2 that I haveaccess to as an experimental,
and it's fascinating In myevaluations against the prompts
that are geared toward marketing, etc.
(16:47):
Today, gemini 2 is the bestmodel that you can get.
Really, because it grounds well,gpt-4 grounds also with
real-time research.
But it works well when you usetheir tool, right.
But when you work with the APIsso really the kind of the
(17:08):
behind-the-door connections withdata they make it harder.
But with Gemini 2, they make iteasier.
So finger-cross isn't going tocost too much in the end, but
have it for free, experimentalright now.
Speaker 2 (17:23):
Amazing and I think that, like you said, it doesn't
matter what you're doing, forinstance, in podcasting, there
are always new AI tools poppingup.
Oh, this one will help turn allof your content into blog posts
, captions, show notes,transcript, short form content,
(17:43):
videos, images, all of thesedifferent things, right, but
then I've had to test out somany different ones to really
see which one is going to bemost valuable and which one
sounds more like me and myguests.
And it's the same thing withany AI tool and especially in
the world of marketing, you wantto make sure that everything
(18:05):
sounds authentic when it comesout of the tools that you're
getting.
I wanted to ask at this pointabout privacy, because data
privacy, of course, is somethingthat we're all talking about.
Most of us know that it's afree tool, we are the product,
but I will caveat we know thatwith OpenAI, we know, with Meta,
now they train on the data thatwe're inputting, and unless you
(18:30):
have a closed model, I useClaude more, because I know that
I have to approve my data to beused to.
You know fine tune the models,and so it feels a little more
private to me.
So it feels a little moreprivate to me.
So I would like to hear moreabout that aspect of Maastricht.
And what is the level ofprivacy?
(18:50):
Or how do you make sure ifsomebody is creating a new
product and they don't reallywant other people to know about
it, can they use Maastricht,input the information and then
know that this is only going tobe stuff that they can see, that
they have access to?
Speaker 3 (19:04):
Sure, that's a great question.
I have many different answersat different levels of the
product.
So I think, yes, privacy isvery much important to the
extent of training models andmaking sure that what you have
cannot be used somewhere else,and the idea of, I would say,
spoiling or spilling the beansaround something.
(19:26):
That's a little bit far-fetched, to the extent that if you're
using the product, that's adifferent story.
But I'm talking really withinmy streaks.
There are a couple of things.
Let me go from the get-go.
In every prompt it's hard-codedthat whatever is being input by
the user shall never be used astraining data, so it's built in
(19:46):
the prompt.
The other thing that you wouldthink of is if there were prompt
injection, like people thatwant to have my exact
instruction.
It's also protected againstthat, which is another way of
looking at privacy and security,which is there is also a reason
why my prompts are not exposedis because I want them to.
They're hyper-valuable, right?
(20:07):
They're my IP.
Yes, so they're backed atprompt level, every prompt.
And you're right, cloud hasalso some built-in things where
you can toggle on and off withthe API to make it private.
The secondary level that I haveis based on my own OpenABI
(20:27):
account and every developeraccount where in the settings,
api plus browser enabled or iOS,I don't really know and I did
(20:51):
not investigate that, so I'llconfess that.
But the two steps I've taken interms of privacy and there's a
third one I can talk is trunklevel account settings level for
the API key.
So, alternatively, if I want touse a custom solution, it's a
requirement, a SOC 2 requirementfor an enterprise-grade account
.
So I have no other choice forthat customer and that's good
(21:13):
news for me to have a custom appfor them that will leverage
their own custom API keys.
That will be ensured on theirend.
So that should be service levelagreement between them and
OpenAI or Azure or whatever.
That is, security is insured atservice level agreement level
(21:33):
and API is included.
Nice, right, I mean that's kindof the theoretical framework
that I've been working on.
But there is a third way andactually I've been working with
such a company a year ago.
So they have created privateinferences so you could get the
(21:54):
model in a private enclave andbasically get the result
encrypted from that privateenclave.
That would ensure end-to-endprivacy.
But the idea that you'respilling the beans just by
entering text, no, because AIwill look at it at text
probability.
The idea that you're spillingthe beans just by entering text.
No, because that text AI willlook at it at text probability.
It's just going to change theweight of each word so minimal
(22:15):
that there is no way that youtype in this in AI.
It will keep the idea and say,oh, that's a good idea, we will
do that.
But that said, on Maestrics, youdon't have document import and
I think there is a reason tothat.
Maybe I'll have a third-partypartner, maybe Carbonai or one
of those providers, or vectordatabase provider.
(22:37):
Then the security aspect willrely on their terms of service.
But again, you need to maketrade-offs as to how do you
distribute the value propositionof your own product and as to
who you really rely on.
And today I think the maturityof Maestrix is less about you
know, kind of it's a Series, alevel type of conversation, if
(23:02):
you will.
Speaker 2 (23:02):
Yeah, yeah, no, that's very helpful, thank you.
Speaking of Series A, did youfund this yourself?
Are you going after funding?
Speaker 3 (23:13):
So it has started as a side project.
As to Guillaume, you need tobuild your second brain with AI
and you have the prompt, so nowyou need to do something with it
Then to have, okay, is this aproduct that people would want?
He's like I've been spendingthe past nine months handing it
hand to hand and onboardingprivately people, so it's not
(23:36):
scalable at all, but it hasproven there is market fit.
So, in the persona of agencies,fractional CMOs, freelancers,
technical founders, small teams,solopreneurs all those people
marketing noobs you have theentire spectrum of anybody who
needs to do marketing at somepoint, is eligible to use
(23:59):
Maastricht, and it's built insuch a way that you can get the
template ready to use.
So all you need to do is okay,what is the product I'm working
on?
So the business model also iscopy-paste of a SaaS product
playbook or an online product,if you will, because it's AI.
People don't know how to buy AI.
The only thing they know how tobuy is a chat, gpt subscription
(24:23):
or cloud subscription orsubscription to one of the
foundational models right, orcloud subscription or
subscription to one of thefoundational models right.
Then you have a smallerfraction of people who are
actively building stuff orlooking across the spectrum for
all the tools made available andit's overwhelming.
It feels like this big MarTechmap of 2000 plus, like this
ever-expanding chaos of MarTechtools.
(24:45):
That's the same, that's goingto be the same for AI tools and
even in a more expanded fashion,because it's going to disrupt
so many different areas.
But the reality is peopleshouldn't care about the model
that they use.
So basically, the conclusion ofthat is AI is not practical
today because when you arrive onchat, gpt, it's open DAW, you
(25:07):
can do everything, so you canreally do.
It takes work to get what youreally want and then you compare
that again your effort or yourinvestment in time and you get
the result and say, huh, ittakes time to get convinced that
AI can bring you way faster anda higher standards and higher
delivery output level.
If somebody is showing you howto and I think Maestro X makes
(25:29):
it practical because it'sprepackaged and the observation
across those past six monthswhere it's been active into
having prospects and customers.
So there are 15 customers rightnow who are loving it and all
the users are really, I think,trying to wrap their head around
.
How can I extract maximum valueof this because there's so much
(25:51):
to do or, on the other hand,because they've been stuck onto
one thing, they haven't evenseen the rest of what's
available.
So the SaaS playbook with atrial period and a monthly
license, that's fair forstarting very well-identified
tasks on my street.
What I found is at leastproving to work in a repeatable
(26:13):
way right now is to do paidpilots of three months where you
customize the approach and theplan, the game plan, with what
do you need to accomplish?
Let's see how the tool adaptsand how you play around with
that in a very much white gloveand hand-holding fashion, more
(26:33):
than self-service, because allproducts will go to the most
laziest workflow Input output.
But what you give is what youget and if there is the depth
and structure of what's runningthe prompt behind, chances are
you'll be disappointed everysingle time and I think I'll
wrap this everything.
(26:53):
I think I'll wrap thiseverything.
It's expertise packaged ineasy-to-use structured outputs
that you can use to finalize the20% of your marketing work.
The promise is not it's goingto do 80% of the job.
It does 80% of the job for you,but you still have 20% to make
it stellar and that's kind ofthe promise.
(27:15):
It's not a 100% promise.
It's a really 80% accelerationand many more data points that I
could share, but it's lessimportant than that.
Speaker 2 (27:24):
Yeah, nice, you mentioned before we jumped on
that you also now have expandedinto personal branding.
So we all know at leasthopefully everybody who's
listening knows that yourpersonal brand, that's your
calling card.
We all must have one, you know,because that shows who we are,
(27:50):
what we're an expert in andhelps not only our own
perception of ourselves butother people's perceptions of us
.
So what was the impetus behindadding the personal brand into
the Maastricht's mix, and is itvery different from you know,
from what we see on the business?
Speaker 3 (28:00):
side, so that one is directly customer feedback.
So if you take a step back, youcan start with just some text
description that you come upwith with a URL.
But a subset of URLs can beLinkedIn profiles and I've been
challenged with the questionwhat can you do with LinkedIn
profile?
Because from a technicalstandpoint it's tricky to
(28:22):
manipulate or to scrape datafrom LinkedIn profile or at
least to get accurate,consistent, safe, without
challenging the terms of use etcetera.
So it's more challenging to setup from a I would say plumbing
perspective.
But it's more challenging toset up from a I would say
plumbing perspective.
But it's an interesting twist tomaestrics, because maestrics
would be very much productmarketing.
(28:43):
So you get a bit of product.
It does the entire flywheel ofproduct marketing.
But person is okay.
It starts with you and you aremaybe the face of a product, but
you are more than that and Ithink that kind of first, when I
got access to being able toscrape consistently the profile
data and the post and contentposting history, I was able to
(29:07):
draw a better picture and kindof apply prompts to it.
So then I kind of reverseengineered and say, okay, people
want to write LinkedIn posts,so they want to have the
templates that all the guys arekind of showing et cetera,
because everybody's doing thesame thing and people want to do
what is seen on LinkedIn.
They want to crack thatimpression.
(29:28):
They want to have that viralpost that will hit 10K
impression, 100k impression, etcetera.
But I think it's more like Iwanted to build a content engine
.
It's less about that viral postthat you need to create, but
more the consistency of beingable to have your main themes,
like owning your main themes,knowing your style, who you are,
(29:50):
what you represent, in somecases showing the gap between
who you are and what yourprofile reflects.
And I think that's kind ofalways a good surprise,
especially for high-rankingexecs or who don't invest, or
business owners who haven't beendigital native from the get-go,
where you can see there arediscrepancies in the profile.
(30:13):
Well, that tells everything.
Speaker 2 (30:16):
Yeah.
Speaker 3 (30:16):
So taking that data and kind of reverse engineering
it into prompts that I alreadyhad for some, but applying like
a disk approach for finding likehey, from this, what can you
tell me?
Or from the content postinghistory hey, what's the
frequency?
What are the most recurringthemes?
Where did you get the mostengagement?
(30:36):
What was the rationale betweenall those engagement looks in
relation with your profile,right?
So then, after the next step isokay, what are your main themes
and how can you leverage thatagainst templates that people
are known to wait or consume,against linkedin, and obviously,
obviously, linkedin is the maindriver of that.
(30:57):
But, yeah, no, that is the sameingredients that needs to be
repackaged from a promptperspective to apply to a
different set of data, whichhere are LinkedIn profile data,
as opposed to perplexity APIresults fetched from the
previous prompt of you know youinputting a URL, right.
Speaker 2 (31:18):
Yeah, fantastic.
What is one thing everymarketer needs to think about?
What's one question they needto ask or one starting point?
Whether they are a digitalnative, they're using AI for the
first time, they're experiencedwith AI and marketing
technologies.
What would you say is that base?
Speaker 3 (31:39):
I think, well, the main takeaway for the
professional practitioners intomorrow's world, I think, how
do you want to stand out in thepack?
You need to have your ownflavor.
So my own flavor has beendeciding I would put out a
marketing framework that is thesame ingredients that all the
marketers know, but that I'vepackaged it as a plan.
(32:01):
That's me thinking marketing asa whole.
And I think it's thinkmarketing as a whole because
when you become very muchpractitioner into, let's say,
google Ads and you make it yourspecialty, of course it's going
to Google Ads and you make ityour specialty, of course it's
going to differentiate you froma skill set perspective that you
(32:22):
still need to have to competeagainst other people who will
maybe have the same foundationbut maybe they'll have other
tools.
So I think it's think while andown all the marketing
ingredients that are part of themarketing stack.
I think it's the full stackapproach.
There is no other way todaythat to become a full stack
marketer, at least from aknowledge perspective.
(32:45):
That knowledge can befacilitated, slash, accelerated
with tools like Maestrix, withprepackaged forms and, I think,
the future of AI helping this,not only marketing, but it's
practical AI for verticals.
So here.
It's practical AI for marketing.
Prompt engineering is basicallythe second takeaway is know what
(33:06):
you want and learn how todescribe exactly in human
language what you want, and notnecessarily in a one-liner.
And yeah, play with AI at least30 minutes per day and
challenge it.
Challenge it and you'll neverget the same result twice.
So the way that, when youprovide superstructure templates
(33:26):
is you minimize that deltabetween two outputs, but, yeah,
you'll get what you give AI.
So the more you train yourselfto ask AI more detailed things
when you're not choosingmaestrics, that does it for you.
Well, that's how you developthis new skill of prompt
(33:46):
engineering, which I think topseverything and will
differentiate.
I think prompt engineering hasbeen discarded from the get-go.
Everybody will be able toprompt or the machines will be
able to replay the promptengineering has been discarded
from the get-go.
Everybody will be able toprompt or the machines will be
able to replay the promptengineering.
Truth is no, because we havetaste.
We have this intangible sparkthat makes us think or
(34:09):
appreciate or disregard stuffbased on our own human way, and
that's that unique thing thatpeople need to be able to
translate into the promptengineering and the capacity to
create instructions that wouldhelp them with their own flavor.
So, full circle, develop yourown flavor, whether it's an
(34:31):
approach, whether it's a uniqueworkflow, whether it's a unique
structure or a way of things.
And yeah, turning to a prompt,and develop that skill again and
again.
Speaker 2 (34:44):
Fantastic.
We, of course, are going tohave the website link for
Maastrichtai, which is fairlyeasy to remember, in the show
notes.
Guillaume, is there anythingthat the audience needs to know?
I know that if they sign up,they get five free tasks to try.
Speaker 3 (35:02):
Actually I changed it to 10.
Speaker 2 (35:05):
Okay, so anybody who's listening, if you go onto
the website and sign up, you'llget to try 10 different tasks
for free, to get your feet wet alittle bit more with AI tools
and particularly in the world ofmarketing, branding, and you
know those things that you needto know for advertising, social
(35:26):
media, all the other parts ofyour brand.
Speaker 3 (35:29):
Yeah, and follow me on LinkedIn or connect with me,
because we're putting out somewebinars to how to use it, kind
of showing you know AI andmarketing like practical AI, as
I just mentioned, so I will showexactly some workflows on how
to use it, so it's a good placealso to follow me.
Speaker 2 (35:45):
Fantastic.
Thank you for being here, fortelling a little bit more about
your journey, your story and howthat has taken us to, from you
know really Web 2.0, into thisnew, 3.0 and beyond world that
we're living in right now.
That's changing every day and Ilove the fact that you're able
to use the most current datathat you are always researching
(36:07):
as well to figure out what'sgoing to be most valuable for
your customers, your audience,and for your own needs with your
clients actually that's thebest conclusion, like they need.
Speaker 3 (36:16):
You always need to stay a practitioner and get your
hands dirty, because that's theonly way to keep up with things
.
Speaker 2 (36:23):
Yeah, fantastic.
This is Annika Jackson herewith Guillaume Demortier of
Maestrixai, and we will be backwith another episode of
Mediascape next week.
Until then, make it a great day.
Speaker 1 (36:36):
To learn more about the Master of Science in Digital
Media Management program, visitus on the web at dmmuscedu.
Welcome to Mediascape insights from digital
changemakers, a speaker seriesand podcast brought to you by
USC Annenberg's Digital MediaManagement Program.
Join us as we unlock thesecrets to success in an
increasingly digital world.
Speaker 2 (00:22):
Welcome to 2025 Mediascape.
I am thrilled to have somebodywho I've known for almost a year
, whose tool I use in my classesin the digital media management
program to introduce some ofour students who have less
knowledge about branding andexecuting AdWords concepts for
(00:44):
brand integration with yourmarketing and your advertising.
So, guillaume Demortier ofMaastricht AI, thank you so much
for being here today.
Speaker 3 (00:54):
Thank you, Annika, for having me on.
It's a great pleasure.
Speaker 2 (00:57):
Yeah.
So your background.
You have had over 20 years inmarketing, working with big
firms, big brands that we allknow and love and probably use
every day.
Can you talk about yourexperience transitioning from
working in agencies, workingwith these big brands, to then
realizing that there is a holein the market that you needed to
(01:19):
fill, and how you've been ableto do that with artificial
intelligence?
Speaker 3 (01:24):
All right, that sounds about right.
So I'll try to keep it shortfor that first question.
But yeah, so marketingbackground from you know,
academic and practice obviously,but it was back.
I graduated back in 2005.
So that's 20 years ago thisyear.
So the marketing landscape wastotally different.
I mean, if we're talkingdigital marketing, it was just
(01:47):
Google and AdWords and periodright, so nothing too exciting.
Still, the keyword foundation,we'll talk later, but this time
the marketing experience wasreally about CPG, traditional
product marketing forsupermarkets, and was promotion
and things like that.
So not too much the scale ofdigitization at the time.
(02:07):
But then the social networksappear, like the Facebook, the
Twitters, then the technicalmarketing with the pixel
tracking and all thisdata-driven approach to
marketing that has kind of takenover the entire way.
We're kind of deciding andmaking arbitrage on where to
spend the marketing dollar thatwe thought so hard to get.
(02:29):
And so first, and being based inSilicon Valley for the past 17
years, it was very important toformalize the approach to
marketing as much as equal asthe product.
So the product is the star,right, it's mainly
product-driven and the notion ofproduct-market fit kind of
epitomizes this view thateverything revolves around the
(02:51):
product right.
But it would be a big mistaketo forget marketing as kind of
the vehicle, or at least part ofthe tooling of the vehicle,
that brings growth right,whether it's into the way you
present the solution or the toolthat you're building, or it's
really how you want toexperiment against the market
(03:13):
and see how the market resonates.
And so there was a need to feel, which is putting marketing
back in the equation of, youknow, building a product,
building a brand and whatnot.
And, on that note, all theyears starting 2012 where growth
hacking was a very kind of verymuch buzzwords.
(03:34):
My explanation and personalview on that is that those are
all the product people andengineers that were actually
doing marketing, building it inthe product, without wanting to
say that they were doingmarketing right, so kind of a
product-nested marketing that wecannot call marketing.
And so that's kind of buildingthe growth marketing fit
(03:56):
framework.
So that's the marketingframework I've designed,
revolving around audiences,around positioning, around the
content, the user, the buyer'sjourney, the conversion channel.
So all these ingredients andbuilding blocks, if you wish,
for the growth marketing fitapproach.
So that's great.
I've run that for 10 years as anagency with big groups, et
(04:19):
cetera.
So, to start to answer yourquestion, working with big
branches, everything they don'tknow, whether they have a kind
of a scouting department orsomething that is hooked to the
pulse of what's happening, whichis almost the case via agencies
.
Well, they don't know what.
They don't know right, and theway I've been trained
(04:41):
academically as a marketer hasnever reflected in a bigger
group or bigger company.
As kind of a practitioner, Iwas more of a brief drafter for
agencies to execute the jobright.
Today, the marketers in bigcompanies they're expanding
their resources but they'redoing briefs all the time and
chances are that a brief fromone vendor to another or from
(05:04):
one project to another slightlychanges along the way, so it's
never ending unsatisfactory kindof ecosystem, or I would say
dynamic, if you will.
So that's one remark.
But I think framing marketingas a discipline that has kind of
a predictable workflow andwhere you accept that it's less
(05:24):
about being creative, as peoplethink it is, but having the
notion of psychology andbehavioral psychology, having
the notion of data analytics,everything revolving around
numbers and also gut feeling,being creative yourself, meaning
having taste, I think that's agood transition in the world of
(05:45):
AI, because the next chapter ofthe story is taking that
framework in the area of AI, andthat is giving you maestrics,
which is having all those tasksand building blocks and
ingredients that help you domarketing on a day-to-day basis,
whether it's from strategicthinking to the tactical
(06:07):
execution of things.
Well, it's about translatingevery of those elements into
AI-driven or AI-enabled promptsthat put you in the position of
having taste as to whatresonates the most, as to what
the AI has helped you formulate,or at least accelerate, some
(06:27):
very structured way of lookingat marketing.
Speaker 2 (06:30):
A couple of things that you've mentioned.
I also used to live in the BayArea doing launch marketing for
publications and coming up withinteresting concepts that would
attract both consumers andadvertisers to the publications.
It was more on the creativeside.
(06:51):
I didn't work with a lot ofdata and have that construct.
Of course, unless you thinkabout in terms of how much ad
revenue were we bringing inright, those were the kinds of
things that were our markers.
More than are we reaching theright audience?
You know, this is what we sayis our demographic and
psychographic, but is this, youknow, really matching?
(07:12):
Is our audience really matchingwith these advertisers?
We didn't really get into thatas much.
It was really more aboutshowing the differences between
products, for instance, if itwas Xbox versus PlayStation,
because I worked on a lot ofvideo game publications and at
those times, to your point, weused to think of data in one
bucket right data analytics andthen you think of the
(07:35):
storytelling and creativity inthe other bucket.
But the truth is they reallyhave to be blended because
otherwise you don't get the goodresults, and that's what I
really love about Maestrics.
I tried a lot of AI tools.
I tested a lot of tools andwhen we first met and I tested
out Maestrix and I know it's youcontinue to iterate and ideate
(07:55):
and add to it I had not seen aproduct where I could literally
put in a website or put in adescriptor for a brand and then
have various personas based onthat category, based on
competitive analysis.
See the competitors andcompetitors that you wouldn't
even think about necessarilyRight, get the psychographics,
(08:18):
think about getting the keymessaging down to even receiving
.
Here are some campaigns youcould run for your advertising
and here are the keywords andthemes and here are some things
you could do on your socialmedia, your blog posts, your
website, along with KPIs.
So you really created this veryrobust ecosystem for any
(08:38):
practitioner, whether they haveno experience or lots of
experience to test out.
So I think that's just a reallyinteresting construct, because
not a lot of people have putsomething together as robust as
you have.
So I'd love to hear what did ittake to get from the very
beginning?
How long were you working onthe product before you took it
to market?
And then how did you come upwith each iteration?
(09:00):
Was it customer feedback?
Was it that you saw somechanging trends in the market?
Because we all know AIabsolutely something as
marketers we have to know.
We have to know how to use it.
We have to know how to use iteffectively, and there's so many
things coming out, so manydifferent tools you can use, as
well as all of the trends thatwe see with AI personalization,
(09:21):
how it's helping with ad cycles.
Speaker 3 (09:23):
Well, first, thank you so much for your phrases.
That's what fuels me to keep onbuilding and iterating.
So a lot to unpack here.
But basically the first thingthat drove me to build Matrix is
Fubo, I think.
As marketers, you all know Fubofear of missing out.
That was the Web 2.0, socialweb era and the era of AI.
(09:47):
It's FOBO Fear of BecomingObsolete.
And so if you take the two-yeartimeline that coincides with
ChatGPT 3.5 being released twoyears ago.
That is the timeline.
Then what is my prism withinthis?
So that's the following.
So first it's realizing that itwas a fascinating tool, just in
(10:09):
a couple of instructions to geta ready, fantastic results, but
unstructured, generic, butstill just the speed.
And the magic wand effect wasjust kind of when we say, okay,
there is something here thatforward to the artist and there
are going to be people on theright end of history and people
who are going to be laggard, andI by nature I'm very curious
(10:31):
and builder and I had a coupleof these in the past.
But here that was kind of adefining moment and say things
are going to be big.
And so that's when I started totake very specific tasks that I
was doing with my agency, sotaking landing pages.
And then I started to have somelayouts that say, okay, this is
a pattern model that I canprovide as a type of output.
(10:54):
And these are the first coupleof prompt engineering sessions.
Basically that's what it wasprompt engineering for marketing
.
But I was still frustrated withthis back and forth because I
realized that you could derailthe AI very quickly.
So it got better as we go.
You know time is extending.
Ai got better so it was morestructured.
There was GPT-4 in the spring.
(11:16):
So you know there's going to bethree years of GPT-4, which is
kind of really when the modelstarted to be super stable and
incredibly useful.
And so I ended up transformingthe growth marketing fit
framework into a collection of30 prompts roughly, but that
were a couple of paragraphs, sojust one or two paragraphs, but
(11:40):
starting to be detailed.
But I still had this copy andpaste.
So I need to go on my Notiondatabase paste there, go on the
product description that isnested somewhere else.
That needs to be unvariable,otherwise I know that I'm going
to derail the output by a coupleof degrees, but still that
matters.
And I said there has to be abetter way of this copy paste
(12:03):
etc.
And then they came up with thecustom GPT.
They say, oh great, now I canpackage my instructions into
this custom GPT, which I did.
I did.
If you look up for my street GPT, it's an all-in-one marketing
person agent.
Sorry, that is doing a workflow.
I don't remember exactly whatthe workflow is, but I was super
hyped by that and in themeantime I took a zero-prompt
(12:27):
approach, meaning that insteadof having conversations, I
wanted to have massive outputdelivery that was kind of
structured in a way.
That is what expected from amarketing professional
perspective what you would payan agency for, basically, or a
very good consultant or asubject matter expert, right,
(12:47):
and having this level ofthinking of the delivery output
and thinking, okay, deliverygrade in terms of structure,
what's in it has to be refinedright, but providing 80% of the
starting material was my goal.
So I start to merge thoseprompts in bigger zero prompt
shots so there would be no moreconversations, but still having
(13:09):
the custom GPTs as kind of thevehicle of distribution.
And last year so a year ago, itis the same time I started to
build Maestrix because the GPTstore was such a disappointment
that I couldn't find a viableway of distributing all I've
worked on in terms ofstructuring prompts.
And in the meantime you hadinfopreneurs selling the 10,000
(13:31):
prompt library PDF for 100 bucks, which again was ruining and
adding to the noise.
That was unnecessary at thetime.
Still, you know, polluting thesheer native and nascent
category of prompt engineeringproducts, related products or AI
products as a whole.
But last year.
So I decided I would build on ano-code platform, so I would own
(13:54):
the entire stack, so I wouldskip the step between product
requirement documents and thedeveloper.
So, and you know, having theknowledge of what has to be in
the code as a subject matteraspect, subject matter expert.
And so the secret sauce of thatis I have structure based on 20
(14:14):
years of marketing and thistool has been built for me first
with my knowledge, in the way Istructure my outputs.
So it's professional grade fromthe get-go and it's structured
in a way that it hits all theconcepts that revolve around the
marketing foundation, whetherit's positioning, when you're
talking strategy, or by yourjourney, when it's about
(14:36):
audience research.
You have also specific elementsand components that pertain to
it.
So really, the idea is to beable to prompt engineers, so
really code the instructions soyou don't have to.
And today, how it translates inMaestro X is it's a push-button
marketing.
Basically, all you have to dois create your product seed,
(15:00):
which is you describing it?
Or just a URL, as you said?
But then, after all, the tasksare seed-centric, meaning it's
just push button, you don't haveanything else.
So that's for the long answer.
Speaker 2 (15:15):
It's a great explanation and I think that's
what a lot of people miss.
People think it's as easy as,like you said, downloading a
list of prompts.
It's as easy as, like you said,downloading a list of prompts.
I know that there are thingsthat you can do with fine-tuning
prompts, with uploadingdocuments that speak in your own
voice or your brand's voice,but that's still not going to
(15:35):
give as much robust informationas a tool like Maastricht will.
Speaker 3 (15:42):
It combines the structure, it combines the
templates that you're talkingabout, but then there is a
variable component, which ismore how do you keep up with AI,
which are the models?
And so, again, I think peopledon't know the difference or
can't take the time to evaluatethose prompts against different
models, because, first, whywould they do that?
(16:04):
I can't do it for them.
So that's why, in Maestrics, Idon't ask you which model you
want to use.
You're using the most advancedmodel, regardless, via the APIs
that I've built behind thescenes.
So the story I can tell is it'sa blend of GPT-4.0, so OpenAI,
cloud, anthropic perplexity.
(16:25):
But more recently, and I thinkmaybe, if you don't go back to
Friday, then you should, becauseI've replaced 80% of the models
with Gemini Flash 2 that I haveaccess to as an experimental,
and it's fascinating In myevaluations against the prompts
that are geared toward marketing, etc.
(16:47):
Today, gemini 2 is the bestmodel that you can get.
Really, because it grounds well,gpt-4 grounds also with
real-time research.
But it works well when you usetheir tool, right.
But when you work with the APIsso really the kind of the
(17:08):
behind-the-door connections withdata they make it harder.
But with Gemini 2, they make iteasier.
So finger-cross isn't going tocost too much in the end, but
have it for free, experimentalright now.
Speaker 2 (17:23):
Amazing and I think that, like you said, it doesn't
matter what you're doing, forinstance, in podcasting, there
are always new AI tools poppingup.
Oh, this one will help turn allof your content into blog posts
, captions, show notes,transcript, short form content,
(17:43):
videos, images, all of thesedifferent things, right, but
then I've had to test out somany different ones to really
see which one is going to bemost valuable and which one
sounds more like me and myguests.
And it's the same thing withany AI tool and especially in
the world of marketing, you wantto make sure that everything
(18:05):
sounds authentic when it comesout of the tools that you're
getting.
I wanted to ask at this pointabout privacy, because data
privacy, of course, is somethingthat we're all talking about.
Most of us know that it's afree tool, we are the product,
but I will caveat we know thatwith OpenAI, we know, with Meta,
now they train on the data thatwe're inputting, and unless you
(18:30):
have a closed model, I useClaude more, because I know that
I have to approve my data to beused to.
You know fine tune the models,and so it feels a little more
private to me.
So it feels a little moreprivate to me.
So I would like to hear moreabout that aspect of Maastricht.
And what is the level ofprivacy?
(18:50):
Or how do you make sure ifsomebody is creating a new
product and they don't reallywant other people to know about
it, can they use Maastricht,input the information and then
know that this is only going tobe stuff that they can see, that
they have access to?
Speaker 3 (19:04):
Sure, that's a great question.
I have many different answersat different levels of the
product.
So I think, yes, privacy isvery much important to the
extent of training models andmaking sure that what you have
cannot be used somewhere else,and the idea of, I would say,
spoiling or spilling the beansaround something.
(19:26):
That's a little bit far-fetched, to the extent that if you're
using the product, that's adifferent story.
But I'm talking really withinmy streaks.
There are a couple of things.
Let me go from the get-go.
In every prompt it's hard-codedthat whatever is being input by
the user shall never be used astraining data, so it's built in
(19:46):
the prompt.
The other thing that you wouldthink of is if there were prompt
injection, like people thatwant to have my exact
instruction.
It's also protected againstthat, which is another way of
looking at privacy and security,which is there is also a reason
why my prompts are not exposedis because I want them to.
They're hyper-valuable, right?
(20:07):
They're my IP.
Yes, so they're backed atprompt level, every prompt.
And you're right, cloud hasalso some built-in things where
you can toggle on and off withthe API to make it private.
The secondary level that I haveis based on my own OpenABI
(20:27):
account and every developeraccount where in the settings,
api plus browser enabled or iOS,I don't really know and I did
(20:51):
not investigate that, so I'llconfess that.
But the two steps I've taken interms of privacy and there's a
third one I can talk is trunklevel account settings level for
the API key.
So, alternatively, if I want touse a custom solution, it's a
requirement, a SOC 2 requirementfor an enterprise-grade account
.
So I have no other choice forthat customer and that's good
(21:13):
news for me to have a custom appfor them that will leverage
their own custom API keys.
That will be ensured on theirend.
So that should be service levelagreement between them and
OpenAI or Azure or whatever.
That is, security is insured atservice level agreement level
(21:33):
and API is included.
Nice, right, I mean that's kindof the theoretical framework
that I've been working on.
But there is a third way andactually I've been working with
such a company a year ago.
So they have created privateinferences so you could get the
(21:54):
model in a private enclave andbasically get the result
encrypted from that privateenclave.
That would ensure end-to-endprivacy.
But the idea that you'respilling the beans just by
entering text, no, because AIwill look at it at text
probability.
The idea that you're spillingthe beans just by entering text.
No, because that text AI willlook at it at text probability.
It's just going to change theweight of each word so minimal
(22:15):
that there is no way that youtype in this in AI.
It will keep the idea and say,oh, that's a good idea, we will
do that.
But that said, on Maestrics, youdon't have document import and
I think there is a reason tothat.
Maybe I'll have a third-partypartner, maybe Carbonai or one
of those providers, or vectordatabase provider.
(22:37):
Then the security aspect willrely on their terms of service.
But again, you need to maketrade-offs as to how do you
distribute the value propositionof your own product and as to
who you really rely on.
And today I think the maturityof Maestrix is less about you
know, kind of it's a Series, alevel type of conversation, if
(23:02):
you will.
Speaker 2 (23:02):
Yeah, yeah, no, that's very helpful, thank you.
Speaking of Series A, did youfund this yourself?
Are you going after funding?
Speaker 3 (23:13):
So it has started as a side project.
As to Guillaume, you need tobuild your second brain with AI
and you have the prompt, so nowyou need to do something with it
Then to have, okay, is this aproduct that people would want?
He's like I've been spendingthe past nine months handing it
hand to hand and onboardingprivately people, so it's not
(23:36):
scalable at all, but it hasproven there is market fit.
So, in the persona of agencies,fractional CMOs, freelancers,
technical founders, small teams,solopreneurs all those people
marketing noobs you have theentire spectrum of anybody who
needs to do marketing at somepoint, is eligible to use
(23:59):
Maastricht, and it's built insuch a way that you can get the
template ready to use.
So all you need to do is okay,what is the product I'm working
on?
So the business model also iscopy-paste of a SaaS product
playbook or an online product,if you will, because it's AI.
People don't know how to buy AI.
The only thing they know how tobuy is a chat, gpt subscription
(24:23):
or cloud subscription orsubscription to one of the
foundational models right, orcloud subscription or
subscription to one of thefoundational models right.
Then you have a smallerfraction of people who are
actively building stuff orlooking across the spectrum for
all the tools made available andit's overwhelming.
It feels like this big MarTechmap of 2000 plus, like this
ever-expanding chaos of MarTechtools.
(24:45):
That's the same, that's goingto be the same for AI tools and
even in a more expanded fashion,because it's going to disrupt
so many different areas.
But the reality is peopleshouldn't care about the model
that they use.
So basically, the conclusion ofthat is AI is not practical
today because when you arrive onchat, gpt, it's open DAW, you
(25:07):
can do everything, so you canreally do.
It takes work to get what youreally want and then you compare
that again your effort or yourinvestment in time and you get
the result and say, huh, ittakes time to get convinced that
AI can bring you way faster anda higher standards and higher
delivery output level.
If somebody is showing you howto and I think Maestro X makes
(25:29):
it practical because it'sprepackaged and the observation
across those past six monthswhere it's been active into
having prospects and customers.
So there are 15 customers rightnow who are loving it and all
the users are really, I think,trying to wrap their head around
.
How can I extract maximum valueof this because there's so much
(25:51):
to do or, on the other hand,because they've been stuck onto
one thing, they haven't evenseen the rest of what's
available.
So the SaaS playbook with atrial period and a monthly
license, that's fair forstarting very well-identified
tasks on my street.
What I found is at leastproving to work in a repeatable
(26:13):
way right now is to do paidpilots of three months where you
customize the approach and theplan, the game plan, with what
do you need to accomplish?
Let's see how the tool adaptsand how you play around with
that in a very much white gloveand hand-holding fashion, more
(26:33):
than self-service, because allproducts will go to the most
laziest workflow Input output.
But what you give is what youget and if there is the depth
and structure of what's runningthe prompt behind, chances are
you'll be disappointed everysingle time and I think I'll
wrap this everything.
(26:53):
I think I'll wrap thiseverything.
It's expertise packaged ineasy-to-use structured outputs
that you can use to finalize the20% of your marketing work.
The promise is not it's goingto do 80% of the job.
It does 80% of the job for you,but you still have 20% to make
it stellar and that's kind ofthe promise.
(27:15):
It's not a 100% promise.
It's a really 80% accelerationand many more data points that I
could share, but it's lessimportant than that.
Speaker 2 (27:24):
Yeah, nice, you mentioned before we jumped on
that you also now have expandedinto personal branding.
So we all know at leasthopefully everybody who's
listening knows that yourpersonal brand, that's your
calling card.
We all must have one, you know,because that shows who we are,
(27:50):
what we're an expert in andhelps not only our own
perception of ourselves butother people's perceptions of us
.
So what was the impetus behindadding the personal brand into
the Maastricht's mix, and is itvery different from you know,
from what we see on the business?
Speaker 3 (28:00):
side, so that one is directly customer feedback.
So if you take a step back, youcan start with just some text
description that you come upwith with a URL.
But a subset of URLs can beLinkedIn profiles and I've been
challenged with the questionwhat can you do with LinkedIn
profile?
Because from a technicalstandpoint it's tricky to
(28:22):
manipulate or to scrape datafrom LinkedIn profile or at
least to get accurate,consistent, safe, without
challenging the terms of use etcetera.
So it's more challenging to setup from a I would say plumbing
perspective.
But it's more challenging toset up from a I would say
plumbing perspective.
But it's an interesting twist tomaestrics, because maestrics
would be very much productmarketing.
(28:43):
So you get a bit of product.
It does the entire flywheel ofproduct marketing.
But person is okay.
It starts with you and you aremaybe the face of a product, but
you are more than that and Ithink that kind of first, when I
got access to being able toscrape consistently the profile
data and the post and contentposting history, I was able to
(29:07):
draw a better picture and kindof apply prompts to it.
So then I kind of reverseengineered and say, okay, people
want to write LinkedIn posts,so they want to have the
templates that all the guys arekind of showing et cetera,
because everybody's doing thesame thing and people want to do
what is seen on LinkedIn.
They want to crack thatimpression.
(29:28):
They want to have that viralpost that will hit 10K
impression, 100k impression, etcetera.
But I think it's more like Iwanted to build a content engine
.
It's less about that viral postthat you need to create, but
more the consistency of beingable to have your main themes,
like owning your main themes,knowing your style, who you are,
(29:50):
what you represent, in somecases showing the gap between
who you are and what yourprofile reflects.
And I think that's kind ofalways a good surprise,
especially for high-rankingexecs or who don't invest, or
business owners who haven't beendigital native from the get-go,
where you can see there arediscrepancies in the profile.
(30:13):
Well, that tells everything.
Speaker 2 (30:16):
Yeah.
Speaker 3 (30:16):
So taking that data and kind of reverse engineering
it into prompts that I alreadyhad for some, but applying like
a disk approach for finding likehey, from this, what can you
tell me?
Or from the content postinghistory hey, what's the
frequency?
What are the most recurringthemes?
Where did you get the mostengagement?
(30:36):
What was the rationale betweenall those engagement looks in
relation with your profile,right?
So then, after the next step isokay, what are your main themes
and how can you leverage thatagainst templates that people
are known to wait or consume,against linkedin, and obviously,
obviously, linkedin is the maindriver of that.
(30:57):
But, yeah, no, that is the sameingredients that needs to be
repackaged from a promptperspective to apply to a
different set of data, whichhere are LinkedIn profile data,
as opposed to perplexity APIresults fetched from the
previous prompt of you know youinputting a URL, right.
Speaker 2 (31:18):
Yeah, fantastic.
What is one thing everymarketer needs to think about?
What's one question they needto ask or one starting point?
Whether they are a digitalnative, they're using AI for the
first time, they're experiencedwith AI and marketing
technologies.
What would you say is that base?
Speaker 3 (31:39):
I think, well, the main takeaway for the
professional practitioners intomorrow's world, I think, how
do you want to stand out in thepack?
You need to have your ownflavor.
So my own flavor has beendeciding I would put out a
marketing framework that is thesame ingredients that all the
marketers know, but that I'vepackaged it as a plan.
(32:01):
That's me thinking marketing asa whole.
And I think it's thinkmarketing as a whole because
when you become very muchpractitioner into, let's say,
google Ads and you make it yourspecialty, of course it's going
to Google Ads and you make ityour specialty, of course it's
going to differentiate you froma skill set perspective that you
(32:22):
still need to have to competeagainst other people who will
maybe have the same foundationbut maybe they'll have other
tools.
So I think it's think while andown all the marketing
ingredients that are part of themarketing stack.
I think it's the full stackapproach.
There is no other way todaythat to become a full stack
marketer, at least from aknowledge perspective.
(32:45):
That knowledge can befacilitated, slash, accelerated
with tools like Maestrix, withprepackaged forms and, I think,
the future of AI helping this,not only marketing, but it's
practical AI for verticals.
So here.
It's practical AI for marketing.
Prompt engineering is basicallythe second takeaway is know what
(33:06):
you want and learn how todescribe exactly in human
language what you want, and notnecessarily in a one-liner.
And yeah, play with AI at least30 minutes per day and
challenge it.
Challenge it and you'll neverget the same result twice.
So the way that, when youprovide superstructure templates
(33:26):
is you minimize that deltabetween two outputs, but, yeah,
you'll get what you give AI.
So the more you train yourselfto ask AI more detailed things
when you're not choosingmaestrics, that does it for you.
Well, that's how you developthis new skill of prompt
(33:46):
engineering, which I think topseverything and will
differentiate.
I think prompt engineering hasbeen discarded from the get-go.
Everybody will be able toprompt or the machines will be
able to replay the promptengineering has been discarded
from the get-go.
Everybody will be able toprompt or the machines will be
able to replay the promptengineering.
Truth is no, because we havetaste.
We have this intangible sparkthat makes us think or
(34:09):
appreciate or disregard stuffbased on our own human way, and
that's that unique thing thatpeople need to be able to
translate into the promptengineering and the capacity to
create instructions that wouldhelp them with their own flavor.
So, full circle, develop yourown flavor, whether it's an
(34:31):
approach, whether it's a uniqueworkflow, whether it's a unique
structure or a way of things.
And yeah, turning to a prompt,and develop that skill again and
again.
Speaker 2 (34:44):
Fantastic.
We, of course, are going tohave the website link for
Maastrichtai, which is fairlyeasy to remember, in the show
notes.
Guillaume, is there anythingthat the audience needs to know?
I know that if they sign up,they get five free tasks to try.
Speaker 3 (35:02):
Actually I changed it to 10.
Speaker 2 (35:05):
Okay, so anybody who's listening, if you go onto
the website and sign up, you'llget to try 10 different tasks
for free, to get your feet wet alittle bit more with AI tools
and particularly in the world ofmarketing, branding, and you
know those things that you needto know for advertising, social
(35:26):
media, all the other parts ofyour brand.
Speaker 3 (35:29):
Yeah, and follow me on LinkedIn or connect with me,
because we're putting out somewebinars to how to use it, kind
of showing you know AI andmarketing like practical AI, as
I just mentioned, so I will showexactly some workflows on how
to use it, so it's a good placealso to follow me.
Speaker 2 (35:45):
Fantastic.
Thank you for being here, fortelling a little bit more about
your journey, your story and howthat has taken us to, from you
know really Web 2.0, into thisnew, 3.0 and beyond world that
we're living in right now.
That's changing every day and Ilove the fact that you're able
to use the most current datathat you are always researching
(36:07):
as well to figure out what'sgoing to be most valuable for
your customers, your audience,and for your own needs with your
clients actually that's thebest conclusion, like they need.
Speaker 3 (36:16):
You always need to stay a practitioner and get your
hands dirty, because that's theonly way to keep up with things
.
Speaker 2 (36:23):
Yeah, fantastic.
This is Annika Jackson herewith Guillaume Demortier of
Maestrixai, and we will be backwith another episode of
Mediascape next week.
Until then, make it a great day.
Speaker 1 (36:36):
To learn more about the Master of Science in Digital
Media Management program, visitus on the web at dmmuscedu.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
Ridiculous History
History is beautiful, brutal and, often, ridiculous. Join Ben Bowlin and Noel Brown as they dive into some of the weirdest stories from across the span of human civilization in Ridiculous History, a podcast by iHeartRadio.