July 2, 2025 • 48 mins
The technological landscape is shifting beneath our feet, and Dan Baird stands confidently at the intersection of entrepreneurship and artificial intelligence. In this riveting conversation, the co-founder and product lead at Wrench.ai delivers both a wake-up call and an invitation to the AI revolution that's already transforming business.
Baird's journey from selling beanbag chairs at Lovesac (which began as a joke and later went public) to developing patented AI-driven segmentation processes reveals the mind of a true builder and innovator. His candid insights cut through the hype surrounding artificial intelligence, distinguishing between different modalities like machine learning ("statistics on steroids") and generative AI ("autocomplete on steroids"), while emphasizing that understanding these distinctions is crucial for future success.
What sets this discussion apart is Baird's practical approach to AI implementation. He introduces the concept of building personal "agent armies" – specialized AI assistants for different tasks that form an organizational chart beneath you. This approach allows professionals to automate repetitive work while focusing on novel problem-solving. Perhaps most valuable is his exploration of "digital exhaust" – the data created during AI interactions that reveals preferences and decision-making patterns, which he argues may ultimately be more valuable than the tools themselves.
"You won't be replaced by AI, you'll be replaced by someone using AI," Baird warns, noting that professionals who don't embrace these technologies now risk becoming irrelevant within five years. Yet his message remains optimistic: "You are not late, you can absolutely catch up, and it actually really is fun and interesting." For students, professionals, and business leaders alike, this episode offers not just a glimpse into the future of work, but a practical roadmap for navigating it successfully.
Ready to explore how AI can transform your work and create new opportunities? Connect with Dan on social media and discover how these emerging technologies might help you build your own future.
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):
This is going to be a fun one.
I am so thrilled to have DanBaird on the show today.
Dan, you are co-founder andproduct lead at Wrenchai.
You also have an upcomingpodcast, burn the Map.
Your assistant found me onLinkedIn and introduced us to
start having conversationsLinkedIn and introduced us to
(00:46):
start having conversations, andit's been a lot of fun to learn
from you and to be able to, likeyou know, talk about what I'm
teaching and the AI involvementthat I have and where I should
go and where my students shouldalso go.
So thank you, first andforemost, for being here.
Speaker 3 (00:59):
Hey, my pleasure.
This is fun, interesting stuff.
I'm fascinated You're in areally, really cool space.
I'm excited to be here.
Speaker 2 (01:05):
Well, you're in a very cool space, so likewise, so
you have an MBA in globalbranding.
I've done a lot of differentwork in strategy design.
I want to talk about how yougot from there to now, figuring
out and designing tools andprocesses.
(01:25):
You have patented AI-drivensegmentation processes.
Speaker 3 (01:30):
I have always been a builder.
It is one of those things thatkind of runs in my blood.
Third generation entrepreneur.
My dad owns a foundry.
I grew up working for him.
He's a master craftsman.
He built planes in our garagewhen I was a little kid like
super cool stuff and not like wedo not come from money or
anything.
This is poor people building aplane in the garage.
(01:50):
So I was always surrounded bybuilding things and I've always
really had absolutely a passionand soft spot in my heart for
the people that go from zero toone.
I think it's one of the hardestthings you can do and growing
up I'm in Salt Lake City, utah,and happened to have one of my
friends start a little company abeanbag chair company, as a
joke called Lovesack.
Speaker 2 (02:10):
Oh, as a joke, yeah, as a joke.
Speaker 3 (02:13):
So we went to a drive-in movie theater, threw
that thing out.
Speaker 2 (02:16):
It was a joke.
Speaker 3 (02:16):
But literally everybody went I want one of
those things.
So we started a company.
Lo Sack went public in 2019.
So you can actually take thingsliterally as a joke, like we
would do that job because theeconomy was terrible and we
could give ourselves really cooljob titles that no one else
would give us.
But at 19, I ended up having 30employees and those beanbag
(02:38):
chairs if you look around thecountry you'll see them in malls
Travis, barker and other peopleare doing their commercials and
it's just a really fun and kindof cool concept.
But it was also born out of therags to riches kind of like.
We did guerrilla marketingbefore it was called guerrilla
marketing.
We bootstrap stuff before itwas called bootstrapping.
So that was kind of my firstforay.
I'd go to school until noon andthen I'd go work in the
(02:59):
lovesack factory until midnightjust to get stuff out.
We'd sell out multiple times aday.
It was just a really funconcept.
They tried to get me to quitschool and I was like I don't
know if I'm that guy.
They're like will you pleasemove to Tijuana and run our
production facility?
And I was going like I'm like ayear away.
Guys, I really wanted to.
So I ended up opening.
I said, instead of giving mefounder stock, I want to open a
(03:20):
bunch of franchises myself.
I did that in California,arizona.
I found Thunderbird.
Thunderbird is my MBA schooland a thing that I wanted to do
there was I was going.
What is the best way I wouldget you know, cause I learned
how to make product.
I learned how to make retailstores, but I was constantly
putting out fires and I wantedto go okay, how do people that
really do this globally do this?
(03:41):
So I wanted to go to get an MBAso I could learn how the world
was run.
And that I went and did.
Yeah, as you mentioned,branding and a dual capstone and
product management was theclosest thing I could find for
entrepreneurship and productmanagement.
Those were the two things,because if you're doing product
management, you're building aproduct inside of a corporation
from zero to one.
So I loved building that stuffand you know I had one of my
(04:02):
professors invented otter popsand I thought that was the
coolest thing ever and reallyalways liked the people that
were building things.
So I went and did innovationstrategy for ConAgra.
I worked on Healthy Choice andOrville Redenbacher and stuff
like that and our job a smallteam of 12 of us was go build a
platform.
It wasn't new flavors, they'relike you need to launch an
entire platform.
You'll get three-ish a year andif they don't do 20 million in
(04:24):
their first year, you're afailure.
So it was a really fun, cooljob.
You'd fly around the country,do focus groups, do prototypes.
You had really they reallyfunded that group.
So we got to have a lot of funbuilding with some really cool
companies.
And in food, which is superregulated and you have to be
really, really careful.
But if you go down to yourfrozen aisle today and you go
look at the frozen lasagna, youcan read my prose on top of your
(04:48):
Marie calendars.
You know it's that.
So fun stuff, cool stuff, hardstuff.
And after that I took a littlebit of a trip around the world
and came back was seeing a fewthings that I thought were
really cool.
Crowdfunding at the time was abig deal and I was seeing an
explosion of entrepreneurshipand it doesn't sound like a big
deal, but crowdfunding and justcrowd dynamic based businesses
(05:10):
are really really cool Causethey're kind of like open source
.
They flip the script and theyuse like the population and,
like it's a, when we were doingLove Sack and selling out
literally like multiple times aday, you would have to go buy
your product from China and youwould have to wait at least 90
days for it to show up, so youwould run out and then have
three months of nothing on theshelves until it got here and
the logistics actually killed us.
(05:31):
We had a really profitablebusiness in terms of margin, but
our cashflow got murdered justbecause it would take 90 days to
do so.
So crowdfunding actuallyallowed for the first time a
really big deal allowedeverybody to predict and measure
demand before they actuallypaid for inventory.
That doesn't sound like a bigdeal.
That is a really, really,really big deal.
That's like breaking down thefourth wall of a lot of
(05:52):
strategies, because you couldactually understand exactly how
many units do I need to buy.
I thought that was really cool,started a company called Crack
the Crowd and we opened thatthing up and we ended up being
kind of one of the premieragencies and very specifically
for equity crowdfunding in theUnited States.
We did some of the first multihundred million dollar raises
for real estate, stadiums andstuff like that Found out.
A good like this was a lot ofthe precursors to tokenomics,
(06:16):
right to a lot of the cryptotechnologies and things like
that.
So we were on the front end ofthat and it was really Wild West
.
There's a lot of stuff hey,we're raising money for a
submarine.
Hey, we're raising money for asubmarine.
Hey, we're raising money for alot of things we're like is this
legit?
A lot of them were, a lot ofthem weren't.
A lot of them were asking us todo things that were, ultimately
, that people didn't know this,but illegal, just like hey, you
can't.
(06:36):
Little things like when you'reraising money very specifically
for stocks and equities, youcan't say we will make money.
You can't even say we will,right, you have to be really,
really, really careful.
So it was a little too WildWest for us, because people are
going well, that guy's makingmoney, I'm going, he's making
money, he's also going to go tojail in five years.
And we pivoted into wrenchwrench, which is now what we do,
(06:57):
where we took a ton of thesegmentation and the work that
we did there and understandingwho is your audience, how do you
find that audience and who,what, where, who do you talk to,
what do you say, and how do yousay it?
We use a lot of AI, multimodalAI, to do natural language
processing, we do, yes,generative, we do machine
learning and we actually mix abunch of different models to
understand, again, who is myaudience, who am I talking to?
(07:19):
What can I do to enhance ourconversation and their
understanding of it?
How can I streamline theconversation and how can you do
it at scale?
So we focus on a lot ofpersonalization technology and
we focus on a lot of measuredcontent-driven strategy.
What do we say?
And again, why Not just hey?
The LLM told me to, anybody cando that.
We actually went hey.
(07:39):
When dollars are on the line,the differentiator between those
two is going to be who actuallyhas the data to justify their
actions.
And when you're at really bigcompanies, they demand that
level of understanding.
That's what we do at Wrench.
So that's how I got there.
I got here A lot of that typeof work.
Speaker 2 (07:55):
Yeah Well, and I appreciate that you mentioned
you mix a lot of different typesof what falls under the
umbrella of artificialintelligence.
Because that was one thing Italked to my students about and
they're like, oh, I didn't evenknow there was a difference
between AI and machine learningand deep learning and then
generative and the use cases foreach.
(08:16):
I'm like, yeah, there is adifference and we need to
understand that.
The nuances.
Speaker 3 (08:21):
It's a big deal.
I mean like it's all ultimatelymath, but you're talking about
arithmetic versus algebra,versus calculus, versus string
theory, versus.
They're different types, yes,but they accomplish different
things and a lot of them arereally good at one thing and bad
at another and for what it'sworth, we're going through.
You know, people talk aboutlike the hype cycle and things
of AI.
Like we've been doing wrenchfor seven, eight years now,
(08:43):
right, so like that's the joke.
We were doing it before.
It was cool is kind of a joke.
I'm a pioneer dinosaur and it'seight years, like not even old,
but it's a really big deal.
A lot of people are making a lotof mistakes and that's a really
really probably the biggestsingle mistake that they make is
not understanding thedifferences between the two.
Right, machine learning ispretty close to like statistics
(09:04):
on steroids, right, it'smathematical based, large data
sets, tabular data sets, meaningtable-based data sets, and then
generative.
It's like well, they took thosetabulized data sets and they're
just predicting the next word,right, I mean, they really are
fundamentally autocomplete onsteroids.
Most of the decisioning isbrand new.
They't been doing that and theamount of mistakes that they
(09:26):
make is like way more likeyou're like it looks really good
.
I'm like that's only becauseyou didn't check the answer.
You have to check the answeryeah and most of the time when
people don't, they're like, hey,it looks good and I go.
Well, here's the other thing,the uh, by definition like uh.
If you go through and you lookmachine learning, you move into
generative ai, llm, right, gbt'sbeing generative pre-trained
(09:48):
models.
They read the entire internet.
They put it all into a tablewhere they said, all right, what
word do I have?
Now it's like the word from,and they go okay, after reading
the entire internet, what's thenext most common word?
That gets placed right afterfrom.
And that's what they're doing,one word at a time all the way
down, which means most of thetime, after reading it, they
they spit out average answers,right, they summarize really
(10:10):
really well.
They're good at that.
But that also means again thatthey give out average answers.
That means most of the time.
If you're impressed by what theLLM did, you're probably
actually like a little bit belowaverage.
Or it's like, hey, look, it gotme an average answer really
quickly.
But if you're really good atsomething and you look at what
it gives you, you're like that'snot right.
Right, it's, it's.
You can almost tell that it'sfantastic, because that rising
(10:32):
tide raises all boats.
It does mean that we shouldactually see at least just
average content and output frommost of the product that comes
out of it, which means belowaverage is going to get higher
and higher and higher.
We're going to raise the stakesfor what average is acceptable.
But simultaneously, a ton ofpeople are making a mistake for
like hey, I use generative.
It's like cool, you have nodefensive moat, you are totally
(10:55):
replaceable.
Anybody could have done thiswith that amount of effort.
And again, it is one of thosetools where it's like you need
to be using it, butsimultaneously, like again
they're right where they say hey, you won't be replaced by AI,
you'll be replaced by someoneusing AI.
And that's largely one of thereasons is they're going to get
to the average answer, but thenthey're going to be doing
(11:16):
additional research to get anabove average answer.
Yeah, yeah.
Yeah, and that's largely what ifthey're just using AI.
They're replaceable, like that.
If you're not, if you're notusing AI.
Sadly, right now, like I mean,we just reached one of the first
times in my company's historywhere we actually had people
that we were hiring and we hadsomeone that came up, good,
worked in the space, highlyrecommended, and uh, we
(11:39):
ultimately didn't hire thembecause they weren't using AI.
It was one of the first timeswhere we just literally went and
looked.
You probably have one, two yearsleft before you are wholly
irrelevant Unless you start now.
You need to go look so youmight be good.
You're not going to be good forlong.
It's too easy to catch up now.
So for any of the peoplewatching this, it is one of
those, like I just we hired somenew graduates as well who are
(12:01):
jumping in and they're neck deep, but I can't stress enough.
They told me that at schoolthey were discouraged, even like
it was threatened as as thoughit was plagiarism to use LLMs.
Screw that, that is stupid.
I don't care if USC tells youguys to do that, use it anyways.
I'm not telling you to have itwrite your papers.
You need to know how it works.
You need to know when.
(12:29):
You need to know quickly.
If you don't, in five years youwill not be employable.
So go play with it.
There are different tools thatreally accomplish and excel at
different things.
You can go and use, like youknow, google deep research to go
write nice empirically basedresearch papers.
Still, you need to check all ofthe data sources, gamma to do
presentations.
There's any number of differenttools for different kind of
accomplished goals, images andstuff.
Chatgpt is really good at now.
(12:49):
They're really good at raisingkids.
My kids play with them at homeand talk to it because it's
fantastic at that type of work.
But for what it's worth, dig in, jump in neck deep.
I can't tell you it really doesmake me nervous.
My job is wholly replaceable asa CEO.
I can, and that is actuallykind of the work I do.
I can actually even show you Ihave AI agents.
I have a Dan bot.
(13:09):
They will, and they have beenprogrammed with about three plus
years.
I'm on calls like this about 30hours a week.
So I have transcripts, alltypes of video.
I plug them in that like thereisn't a huge need to hire
additional customer supportanymore because we can totally
replicate everything that I telland do them.
They won't want to just ask danlike we'll ask my bot and then
(13:31):
you don't have to, but I can bein five places at once now yeah,
have you used your bot toreplace yourself in meetings yet
?
Speaker 2 (13:38):
because we've talked about this, I have delphi.
I have a delphi profile formyself.
I haven't turned on the videocomponent, but I can have her go
in to meetings.
I can connect her to Zoom or tomy meeting schedule and she can
go in and act as me.
Speaker 3 (13:55):
So I have not tried that one.
That's an interesting one.
What I will do is not attendthe meeting but send my Read AI
bot to it and read and summarizethe transcripts and the action
items after the fact so that Ican go hey, I missed it, but
I'll follow up with you like 20minutes after the meeting.
I got all the high points, andthat way, oftentimes too, I can
actually pay more attention thanI can, and you know, sometimes
(14:16):
it's like an hour, it's like Ineed to be there for five
minutes.
Speaker 2 (14:19):
You know Right, right yeah.
Speaker 3 (14:20):
So I have sent it.
In that sense I haven't sent itin my place yet, but again, I
have messed with.
Hey, jen is another one toowhere we have we've taken and
programmed like scripts wherethey recorded me takes about two
minutes and then now they havebasically a live deep fake that
they can feed a script to and Ican sit and like read off a
here's what this new featuredoes, type of stuff, without
(14:41):
actually recording it.
So we've used that one as well.
Speaker 2 (14:45):
So it is absolutely kind of that's the next thing
I'm going to be adding is AI formy videos, so that, yeah, I'll
be showing it for class, I'll beshowing for podcasts, but there
are other instances where Imight not need to.
If I want to create, like yousaid, promotional videos or
project related videos for work,that I do.
That's going to save so muchtime.
Speaker 3 (15:04):
Oh, yeah, and it's actually better for the student
experience as well.
Right, I don't know if you'veseen it, but you should go check
it out.
There was a study.
There's been some people thathave been following a school in
Africa where they'vesupplemented the learning
experience with artificialintelligence and personalized
tutors and they are doing likeorders of magnitude, more like
kind of adopted learning andinternalized learning, like they
(15:25):
basically said that theadvancement from kind of that
personalized experience wasorders of magnitude of just the
classroom experience alone.
So at the very least, it isn'tsaying the classroom experience
goes away, but it absolutelydoes say that there are
personalized styles of learning,totally reflected, like we do a
lot of this personalizationwhere we'll like I can look at
your LinkedIn profile and see ifyou're a data-driven person or
(15:48):
if you're more socially drivenand you'll take and learn and
make decisions based off socialcues.
And it doesn't mean that one isbetter than the other, but it
does usually mean that youprefer information that way,
which means same thing forstudents.
It's like if I give the one onthe left data and I give the one
on the right social proof,they'll actually believe me both
ways and it's like, hey, thisis just how they perceive to
(16:08):
ingest that information andyou're going wow, so we can
potentially get twice as muchdone just by removing that
psychological friction which Ididn't have it turned on, I let
people do like 10 anonymousquestions.
Speaker 2 (16:29):
I was like, no, no, no, no, I need it because I can
go in and see all the questionspeople asked.
And I realized, oh, some peopleare just using it as the way
that they should use chat orcloud or you know one of these
other tools, or say, oh, can youjust wordsmith this for me?
I'm like that's not the bestuse case.
The best use case is ask meabout branding, public relations
, podcasting, things that are inmy knowledge base, right, and
(16:53):
ask me those questions.
And it saved so much time fromhaving to figure out when can I
schedule another meeting inbetween all my other meetings.
And intro, you know like I'mlike just go talk to her because
she is me.
And the great thing is thatthen even people can.
Somebody said, oh, I didn'tscreenshot or save the
information.
I said well, it knows your IPaddress.
Go back in, it's going to pickup the conversation with you,
(17:16):
even if you call her instead oftext or voice text her.
When you call that number,she's going to say, oh, have you
thought more about blah, blah,blah, whatever was the last
thing that you texted?
And so it's so amazing and thisis something I'm really excited
about.
You mentioned, you know I'mbuilding having your own agents,
and that's something that I'mlearning, starting tomorrow, is
(17:37):
how to build my own agents and Ito your point, like things are
moving so fast.
Some people might say, why areyou getting your MBA with
specialty in AI?
But it's giving me thatfundamental knowledge base.
I'm getting to play with toolsfor business use cases in class,
so it actually is expanding theknowledge way beyond what I
would have done on my own,because it's really easy to
(17:59):
delay learning when you havemeetings, when you have a life,
when you have all these otherthings you're trying to fit into
your day.
Speaker 3 (18:05):
Well, if you think about it, so like we had to
hypothesize what the world wouldlook like, we literally sat
down, so like we were in a spotwhere we were going hey, we're
going to leave the tokenomicsand the Coinbase kind of space
right now, we have enough moneyto actually do this.
We can make a decision.
We're at an inflection pointwhere we go we're not going to
be poor, so let's have some fun.
Our goal is to die with smilewrinkles on our face.
(18:26):
We're going cool, let's put adent in the universe and let's
die with smile wrinkles.
And a big part of that wasgoing okay, well, what can we go
into?
What do we want to go into?
And we talked about AI, wetalked about drones, we talked
about 3D printing, with allexponential technologies, with
an exponential growth curve,when half of success is just
showing up in the right place.
Right, we ended up choosing AIbecause we had some fundamental
(18:46):
skills in it.
We knew some people in it andthings like that.
Again, it was early enough thatpeople were like hey, granted,
it's been around for 20, 30years, but it was still early
enough that it was like hey,which one of these do we really
want to do and you think aboutit.
The organization that youtypically deal with, these large
Fortune 100 companies and stufflike that.
You have the bottom and the topof the organizational chart.
(19:07):
The bottom is basicallycollecting data almost directly.
They hand that to their boss,who takes a lot of predictable
types of data and information.
They look at that data, theymake a decision and then they
actually do the same thing.
They hand it to their boss, whogoes all right, how's this
division doing?
Okay, cool, how's this regiondoing?
And they all just basicallyit's turtles all the way down.
(19:29):
We're going in the future,you're going to largely be
sitting around with novel dataproblems.
Your job will be to generatenew digital exhaust, for lack of
a better term.
The repetitive jobs are theones that are going away.
So when you do the AI stuff,it's like going back and I'm
going like hey, we used to use arotary dialer where it would
take forever just to even call.
(19:50):
You hated people that had likethe.
The zero was the joke.
If you had a zero in your phonenumber, it took so long to even
just spin it.
That friction of how long ittook just to prepare that data
to make that novel decision waslike 70 and 80% of your day Data
silos, even four or five yearsago, were substantial Marketing
couldn't talk to sales, couldn'ttalk to product, couldn't talk
(20:10):
to support.
It still is kind of thatproblem.
You're seeing the barriersalmost break down now where
largely like the trick is, withall of these AI tools and stuff
like that, it's like well, wait,my job is to be the novel
problem solver.
So if they can aggregate thatdata, no-transcript, well, I got
(20:54):
into this business so I couldbake pies, but I'm telling
everyone else how to bake pies.
I haven't baked a pie.
Through that job, you automateeverything that is replicable,
everything that is predictable,and then you move up one level
in the org chart.
You're doing that same thingwith agents, where you go, okay,
(21:15):
move up one level in the orgchart, all right, now I've got
to write manuals for the peoplebelow me.
I'm going to write a bunch ofmanuals.
Now I've packaged up this joband I'm going to move one step
up the org chart.
Where you're basically going,I'm going to build not even just
one agent.
You're going to build anorganizational chart.
Beneath you.
You will have one person forcommunication, one sort of org
chart I've got you know I callhim Prof D but Professor G,
(21:37):
right, scott Galloway.
So I've got Prof D who does alot of my copywriting and
research and stuff, because Ilike the early smarmy swear a
little bit type of copywriting.
And I've got other ones thatare doing finance and proposals
and things like that.
But like you're going to buildan org chart, that just, and I
actually think it'll follow youregardless of where you work,
(21:57):
and your goal will be how muchnovel new decision-making can I
kind of internalize?
How much can I build underneathme so that when people send me
that work I can hopefully sendit through a completely
automated workflow and otherthings?
I've already answered thatquestion.
So your goal will be the mostproductive kind of org chart of
you're going to have that swarmof agents that serve you.
(22:18):
So it'll be nifty, it's funstuff, yeah.
Speaker 2 (22:21):
Some people might hear this and go.
I'm going to have a swarm ofagents that are working for me,
like my own agent army.
But you know what if one goesrogue?
You know there are people whoare going to be really scared of
that.
But then there are those likeyou and me who I'm, like, so
excited about that phase and thehumanity it's bringing back
into, like me actually beingable to do my best work and live
(22:43):
in that space where, you know,I'm still the decision maker,
I'm still the one who'sprogramming.
Maybe, maybe I'll let somedecision making, you know, come
down with agents, but but whereit's, it still has to be my
unique knowledge base.
Speaker 3 (22:59):
Yes, it will, and like I mean, if you notice you
follow these same patterns innormal society anyways, right,
there are, like there's alwaystech and AI as a tool, just like
anything else in the kitchen.
You can make dinner with it,you can cut yourself with it,
right, and just like that.
So, and what we found whenwe're building agents, we do
enterprise agents, so we haveone and like they're fun and
(23:19):
like our goal was we're notgoing to necessarily just build
the fastest or the mostconvenient ones, we want to
build the smartest ones.
And we've had clients that havelike built theirs and said I'm
not going to show my clientsthis, because if they knew this
was how smart they could be,they wouldn't hire us again.
We pulled them back in so theirclients wouldn't see them.
But what you find out is justlike humans, you get kind of a
(23:40):
saturation of memory overcapacity.
You can only pack so muchknowledge into one.
It slows it down.
And so what you do is youreally you have, okay, now I've
got one agent who's anorchestrator and they basically
their job is to take a task andthen to divvy it up into one of
the sub agents beneath it.
You specialize in finance.
I'm going to send it to youbecause this agent over here has
memory just dedicated tofinance and they're a specialist
(24:02):
in it.
So you actually start to findout that even the agents that
you build have an org chartthat's relatively similar to
your organizations and, justlike your organization, there's
governance.
So you'll actually have asecurity agent.
That security agent looks forstuff like wait, is this one
someone trying to jailbreak ouragent?
Are they saying that they'relooking for access to databases?
So like in terms of going rogue, just like in real life, you
(24:25):
actually have one person who'sgoing hey, wait, before we
publish anything, does this looklike this has any nefarious
steps in it, or anything likethat.
So you actually see it kind offollows some of the same
dynamics we have in society.
And again you go yep, I've gota security agent, I've got a
marketing agent, I've got asales agent, I've got a finance
agent and they're all workingcollaboratively together.
(24:45):
Final checks go through mysecurity agent before they
actually get shown to the publicor anything like that.
So, for what it's worth, it'slike yeah, will they, could they
?
Yes, is there going to be badactors?
For sure there's people thatuse knives and every other tool
for bad purposes, but it's not.
You know it's.
It's manageable and it's.
One of those things was likethere's not a monetary incentive
(25:05):
on doing it nefariously.
So 90% of them like weabsolutely have to be careful.
It still scares me how goodthey are and you absolutely have
to watch that ethics andeverything else in terms of data
, data integrity.
But overall, I think it's, it's, this is the new normal.
This is not like pie in the skystuff.
This is now, this is next year.
That's fast and it really.
(25:29):
The thing that really reallyscares me more is what will
happen to the people that don'tadopt it soon enough, because
it's going to be harder andharder and harder to catch up in
many ways.
So it's worth jumping in andjumping in neck deep right now.
Speaker 2 (25:38):
Yeah, well, and to that point, I have a lot of
students who do work at largeenterprise organizations.
I have many who work at verysmall organizations that are
starting their own businesses.
You work with enterprisesolutions.
There are companies out therethat will create agents for you.
Speaker 3 (25:55):
I'm trying to pull up my email and see which how many
this week.
Speaker 2 (25:57):
Well, just yeah, there's a couple that I get
emails from, but I've also beentold oh, that one, you know,
it's the one with all the cutesycartoon like, well, it'll be
all your agents for differentparts of your workflow.
But I've also been told they'rejust working off of chat GPT.
They're not really probablygoing to do the whole thing A
little more.
Which could be good, right, butthey might not be as good as
(26:18):
they say they are, but that's anoption for people who are
smaller businesses.
So I'd ask you, what are thebest ways for people with
smaller businesses to adapttechnology, still be considered
like maybe not the earliestadapters, but still on the
forefront of the small businessAI integration?
Should they learn how to buildtheir own agents?
(26:39):
Should they look at some ofthese tools and how do they
evaluate them appropriately toknow that these are the best
tools for them?
Because there's also, as yousaid.
I mean, even just looking atthe MarTech landscape a couple
of weeks ago for a class, wewere like, yeah, this year there
are 15,000 MarTech solutionsand there are a lot more AI ones
, and that's the smallpercentage increase from the
(27:00):
year before, but it's still.
There are a lot of smallcompanies that say that they're
the best and they simply aren't.
Speaker 3 (27:12):
And they're probably going to go away, right?
So, oh yeah, and the the inmany.
Much of the good news for yourstudents and everything else too
is, quite frankly, that goesfor actually a lot of the large
ones too.
We're still figuring it out.
The joke is, is that like aiagents and agentic ai and
everything else is, and peoplewill say ai employees now,
because they're going.
It's one level above agentic,it's not just workflows.
The joke is that it's liketeenage sex Everyone says
they're having it, everyonethinks everyone else is doing it
(27:34):
and most people haven't figuredit out at all and they're
totally lying, pretending theyknow what they're doing.
That is really the truth.
So in 2024, I was talking to asenior solutions architect with
AWS and I was making theobservation like he said
something to the tune of likehey, 40,000 new startups have
showed up in this space.
And I went, yeah, and as a joke, just because you know that,
(27:59):
maybe a month prior, a studycame out of Europe that said
that something like only 25% ofthe AI startups out of Europe
actually had AI at all.
And so, as a joke, I'm likeyeah, maybe 40% of them actually
have AI.
So 40,000 startups are likemaybe 40% of them actually have
AI.
As a joke, he stops, he getsreally serious.
I'd be surprised if 40 of themhad AI.
(28:19):
So this is someone who seesbehind the scenes and has.
So there is a ton of them thatdo that, and it includes really
large companies as well.
So, for what it's worth, llmops have existed, llm operations
, and what you're talking aboutis like where we actually ingest
, we do this, but the only ahandful of companies really do
right now.
Where you're going.
This is a big deal, because alot of the interface that you're
(28:42):
getting to in the web is goingto be very personalized.
It's going to be be like hey, Ireckon to your point earlier.
I recognize your IP address, soI know I'm going to show you
these products and I'm going tosay it this way and everything
else.
You're going to be likeliterally, the, the web of the
future, is personalized to yourpreferences, which is really
really cool.
But the way that and what thatmeans is, is that that interface
(29:02):
, even the chat interface thatyou see, is actually probably
now where some of the mostimportant data is going to be
exchanged.
You know they say data is thenew oil.
They don't mean it's valuable.
Data is the new oil.
Oil is valuable as a commodity,yes, but it's valuable as a
commodity because it's a fuel,because you can turn it into
plastic and rubber and all typesof different things.
(29:24):
When you start to use those LLMinterfaces with those agents,
when you're typing to them,you're actually revealing a lot
about yourself what you'relooking for, what you're
thinking of, what worries you,what way in which you present
the same type of thing.
I can use the word choice andfrequency of the words that you
choose to put into them tounderstand how you would prefer
to receive information.
And so a lot of those agentictools don't understand that.
(29:46):
Actually, one of the biggesttools that comes with, like,
when you use, go build an agent,go build it with chat, gpt,
it's 20 bucks, right, really,really cheap.
But no, like that.
Hey, you're not paying for itjust because it's 20 bucks.
Most of those accounts and thecost that it costs them, even to
run the GPUs that answer yourquestions, is well over $20 per
month.
They're doing it becausethey're trying to buy your
(30:07):
business in perpetuity, right,they're trying to build market
share.
They're trying to do it because, even though you're paying for
that much the words that youtype in are allowing them to
build a corpus, a body of textin Latin, surrounding you.
They're basically learning whatexactly should I be saying to
actually win the hearts andminds of these people?
And they're building a customerfor life.
Anybody who is building an AIagent right now if you're really
(30:28):
smart, you should actually belooking into the LLM operations
behind the scenes.
So I'd go okay, you built anagent.
Can you actually store and lookat the inputs and outputs from
that agent?
Can you say how many times ithad issues where it froze?
Can you tell me how many timesyou asked a question versus made
a statement, versus provide anysemantic feedback to it?
That digital exhaust that justcomes from interacting with the
(30:49):
agent is worth its weight ingold.
When you start to look at thosecompanies, or if you're running
one yourself, I would say, hey,start paying attention to where
you can get those.
The technology has existed fora long time.
But when you start to think, ifI build an LLM enterprise agent
and I go and I give it to afortune 100 company, the yes,
there is the tool and the factthat I can get those workflows
(31:09):
back, but I said I can also comeback to the employer and go hey
, as a heads up, I know a lotabout what your workforce is
thinking, what they're curiousabout, how much time they're
spending on spending on specifictasks, and so when you go look
at them, are there a lot thatare going to be a flash in the
pan?
Yes, I would say it's thosethat don't realize how valuable
that digital exhaust actually is, and the ones that do and again
(31:31):
, the same thing when you'retalking about students and how
valuable this information is theones that start to realize that
, like, whether you're doinggraphic design right, graphic
design with the number of imageand generative AI tools is low.
If you understand that, I'mgoing to actually take and
understand how people interactedwith that image to turn that
into an even better image.
(31:51):
All right, now you've got adefensible moat going forward.
Same with, like business and MBAstudents as well, I would be
looking into the differentmultimodal type of AI that I
have available to me and I'd belooking at all right, well,
where do I get the data sourcesthat actually helped me make
those decisions and how can Ibasically build a model where
I'm going okay, we're going tomake and do this workflow with
ML.
(32:11):
We're going to do this workflowwith generative.
I'm going to become kind of alittle bit of a master of what
widgets do we have or what toolsare in the toolbox and how do
we solve these problems so thatwe can say, hey, we actually do
know what we're doing there.
We do know how to get better.
We can actually measure ourprogress.
Whether it's the small startupsor, again, the really big
companies, a ton of them stilldon't realize this.
(32:32):
It's like hey, if you can'tmeasure it, you don't know how
good you are at it.
You don't know how much betteryou could be at it.
The people that are going to bereally surviving and thriving
in the near future are going tobe tracking their progress.
Again, llm operations is thename for it.
I would look into those and thecompanies that are doing that.
They're going to be the onesthat stick around, I think,
because they'll have a muchbetter idea.
Like it's too easy to switchfrom.
(32:54):
Like, hey, I want to use deepseek today and I want to use
anthropic tomorrow and I want touse switching cost is zero.
They can go burn all the cashin the universe from their
investors to try to build marketshare, but I don't know that it
gained them any sort of like.
I don't feel any.
I don't feel that I own anyresponsibility to one or the
other.
If I switch, I can switch andthere's zero concern for me to
(33:16):
do so.
What's really all said and doneat the end is who do you own
and who do you haverelationships with?
Right, not who do you own, butwhat relationships do you own
and whom do you have thoserelationships with?
That stuff is a really big dealin the future and that's what I
think those agents willmonetize.
Speaker 2 (33:34):
So Amazing, no, but really interesting and engaging
information.
I want to go back a little bitto talking about your brands.
Right, so you had Crack theCode, crowd, crack the Crowd,
and then you moved to Wrenchai.
What was that transition like?
Were you able to bring some ofyour previous clients along?
Did you have to start fresh?
How hard was it to rebrand thework that you're doing, along
(34:00):
with having this organization?
Speaker 3 (34:01):
It was hard.
I mean, one of the things thatwe found at Crack the Crowd that
was interesting for us is thatwe did a lot of analysis, always
been a really data-drivenindividual, like I always liked
working in highly regulatedindustries for that reason, and
one of the things that we foundthat the mom-and-pop shops like
even the little Kickstarterthings when we'd go and just
observe them, but when we'dwatch like just little
Kickstarter campaigns do well orpoorly, and then we'd go watch
(34:24):
really big fortune 100 companybacked companies and we'd watch
they had the same issues and itwas largely in their launch
process and again how they kindof they took their crowd and how
they organized the launchprocess.
So there's actually likethere's like there's like called
crossing the chasm.
But it's this phenomenon whichreally fun for MBAs.
If you've never read crossingthe chasm everybody watching you
(34:46):
should go read it.
Malcolm Gladwell said it's oneof the most important concepts
in business.
You should go check it out.
But the premise is there's likeinnovators, early adopters, late
adopters, but there's actuallykind of a timeline of adoption
of a new product or technologyand if you follow that kind of
chronological rollout processyou can actually help your
products go big.
Google, use this ring, use this, like some of the biggest
(35:06):
companies you've ever heard ofin the history of the world use
this premise.
And we would use that premise.
So we would go and we found outthe biggest thing that everyone
made a mistake of as theylaunched and they treated
everybody and they said the samething to everybody all at once.
And, uh, they just went hey, welaunched and I went, you
shouldn't do that.
You should get kind of an innercircle and you should go tell
them hey, we're going to launchand I need you guys very
(35:28):
specifically to help me.
And then you take yourinfluencers and you go hey, I'm
about to launch, I'm going to docan you do a product review and
are exclusive or some sort ofthing to help me promote it.
And then, after you basicallygot those two people or those
two segments on board, you go tothe early adopters.
They're the people that lean in.
They stand in line for thesetypes of products.
They're less price sensitive.
(35:52):
What we noticed is that, like,even though everybody learned
this in business school at thetime and everyone knows what an
early adopter is, as soon as youget out into the field, it goes
, disappears.
Glitter, no one goes, wait,where's my early adopters?
And you go, well, they're there.
And we realized, like when wewere at Con conagra, this is
what they would pay us to do isto figure out where those little
tiny focus groups of earlyadopters was at wrench.
(36:14):
One of the first things wefigured out how to do was
actually digitize that processand recognize that if I analyze
parts of speech or other thingsfrom authored content, I could
recognize someone based on thelanguage and how they people
that know a category well aremore likely to use its slang and
its jargon.
No rocket science there.
What they don't and didn'tthink about is that if you use
(36:36):
that to your advantage, youcould go do things like go in
and create pay-per-clickcampaigns and search where
they'd go hey, if you don't have12,000 searches nationally,
your campaign doesn't haveenough volume.
So they would tell everybodylike hey, if you're going to
build PPC campaigns, make sureyou have that search volume.
We're like well, those likeinnovator populations and those
really early adopters are verysmall, very tight but very
(36:58):
fanatical populations and theyuse their own slang and jargon.
They have their own language.
So if we go bid on thosekeywords that are really low
volume that people aren't evenlooking for.
We had PPC campaigns that wouldconvert at 25%.
You're going, hey, like justnot even supposed to be possible
.
So the first thing we did wasbasically go, hey, let's
(37:20):
actually try to build that.
That's what one of the patentsis for, is basically hey, you
can do that for any category,and if you can do that, you can
actually take, instead of doingpersonas, where most people go,
hey, we generated our personasand then we threw them in the
filing cabinet and we neverlooked at them again.
We don't know, you use personasbecause I can actually tell you
who someone is as they walkthrough the door.
The first time we can actuallylabel them and go hey, this one
(37:42):
is the early adopter, that one'sa late adopter.
One is not better than theother, but they take different
styles and if you adapt to those, you can literally quintuple
your conversions.
And so the transition wasactually relatively simple.
But based on that just becauseI mean it seems like a joke
We've been doing segmentationsince the 50s, of 1950, right,
(38:04):
rapid prototyping and stuff.
People thought that, like, thiswas all new and digital.
No, no, no, no, no, no, no, no,no, no.
But until very recently, peoplewould say, like my segment is
males, 18 to 24.
When you're on the internet,your gender and age matters less
than it almost ever did, right,like.
And so these are things wherepeople are on the web who they
(38:24):
want to be, they search for andthey switch between products
with zero cost of switching.
And so, for that matter, I wasgoing hey look, the future is
going to be personalized content.
Personalized content, and if wecan jump on that bandwagon and
understand early who should wedo?
We can build go-to-marketstrategies with this.
We can double conversion ratesof existing agencies, blah, blah
(38:44):
, blah, blah, blah.
So the transition for us fromthat perspective was actually a
lot easier than we kind ofthought it would be, just
because that need for anyonethat was going.
It still happens.
The marketing teams you go allright, we're going to.
What's our new strategy?
They look around who has theblackest turtleneck in the room?
(39:06):
That person gets to make allthe cool decisions because they
look the part.
Right, they have those MartinScorsese glasses, those huge
producer glasses, right, itwasn't and still isn't largely a
space where a ton of data isactually really looked at.
And in the future and I'mtalking next 12 months, it's
going to start being the placewhere it's like did you bring
data In?
God, we trust all others bringdata, yeah.
So I think that transition iswhat actually made our lives
(39:28):
kind of easier.
It allowed us to self-fund.
We're profitable.
We're one of the only AIcompanies that was able to do
that.
Thank you.
That's huge.
That was not easy.
(39:56):
I've got the blood, sweat,tears and scars to prove know
professors and PhDs in datascience and stuff like that.
You don't need to know it anddo it.
People confuse me with a datascientist all the time I have to
crack them like no, no, no, no,no no, no, I'm not.
But it absolutely is somethingyou can learn.
And a big part again.
I wouldn't tell someone to goactually learn to code.
I would tell them to go learnobject-oriented programming,
(40:19):
meaning you need to understandhow the building blocks work at
an abstract level.
That's very doable, it's notvery hard, that's a very good
use of time.
I would maybe learn some Pythonand things like that, but I
would argue like, look just,your fingers aren't going to be
on a keyboard in the near future.
So do you want to understandhow they work?
Very much so.
Do you need to be the personthat's putting ones and zeros
(40:40):
together?
No, I don't think you need tobe that person.
But if you can and you're kindof continuing your study, I
would absolutely focus onknowing those bits of
information.
They're around, they're free,right, they're oftentimes well,
not always free, but like youhave MOOCs and you have Coursera
and things like that and thingslike that, you're in a world,
too, where it's like, whetheryou have a degree or not, the
(41:01):
person who is the most up todate on the newest information
and is the most into it is goingto be probably the most
pragmatic hire.
Speaker 2 (41:06):
So yeah, again, so much great information here.
This is such a fully packedepisode.
I do want to turn to yourupcoming podcast.
Yeah, talk about Burn the Map.
What's behind the name?
Speaker 3 (41:21):
So you actually just heard some of it.
So this is good.
So crossing the chasm in thatpremise is the most interesting
thing for me, because I used tolove that space.
I still love that space whenyou get into talk to like those
innovators.
So imagine a bell curve.
That's what those, thosesegment groups, look like.
People imagine their targetmarket and they imagine a
bullseye.
It's like if you took yourbullseye and you rotated it
(41:43):
sideways so you could see thatthe innovators were the bullseye
.
Sure, they're the people thatactually work on it.
They have almost no pricesensitivity.
They like this problem so muchthat they work on it in their
free time.
And then you get to influencersand you get to early adopters
people.
You're a couple rings out nowlate adopters.
These are the people that wantdiscounts.
(42:04):
They're not into the space.
You really have to adapt yourcommunication to get them
interested because they're justtotally not into it.
Those innovators arefascinating to me.
They're the people that doinsane things.
There's an old quote from neverunderestimate the ability of a
small group of people to changethe world.
Indeed, it's the only thingthat really has right.
(42:25):
It's so true when you go findout there are very, very small
groups of people that areactually the ones that put a
dent in the universe, and mostthe rest of us are just kind of
along for the ride.
Earn the map is the name of thepodcast, and it's because I
wanted to actually go talk tothose people not a not even
necessarily just the people thatwere the visionaries and the
pioneers that had accomplishedit, but I wanted to talk to the
(42:47):
people that were in the middleof the thick of it, because
they're actually in a spot, notnecessarily where a map is
irrelevant, but where they'vestudied it so hard they realize
they're no longer on thechartered map, right.
So so it wasn't anything to saywe disrespect the map, we love
the map.
It's just we've learnedeverything the map can teach us
at this point, and we'restarting to move into a new
space, and so the team is usingeven some of our technology to
(43:11):
start to highlight some people.
You basically go into a skillset and we look for and we
analyze their online behaviorsand we look for people that have
a specialty, almost to likekind of the fanatical level.
So we can actually we did thisyesterday with my assistant,
where I pulled up her map likeimagine a little scatter plot
bubble chart and it has likepeople from zero to 100 that
(43:31):
have an affinity for you,whatever else you want.
But I was explaining to her Iwas like, hey, look, so I put in
your bio, I measured everyoneelse's bio and how many times
they use words, phrases andabstract concept language the
same way you did.
And this is a ranking ofeverybody that's in your social
chart of how much they relateand have an affinity for you and
(43:53):
like, yeah, whoever this is isprobably a good match.
It ended up being her boyfriend, like out of literally hundreds
of dots on the screen.
She's like, is that Colby?
That's my boyfriend.
I'm like, yeah, there you go.
So we're doing that to analyzeand basically have conversations
with those people because theyare delightful Like.
One of our first episodes iswith Bridger Jensen, who has
(44:14):
been on the forefront ofstudying.
He helped develop one of ourpersonality models with us.
This guy knows the differencebetween like Myers-Briggs and
disc profiles and all the prosand cons of those.
He ended up being on theforefront of therapy.
He wanted to actually build anon-psychopharmacological I'm
not even sure if I'm saying thatword correctly.
He wanted to see if he couldactually build and measure the
(44:37):
utility of like, say, going toCrossFit instead of like taking
a pill.
Here's one.
There has to be solutions tosome of these problems that
don't require medicine.
He went into the medicinemedicinal space and he ended up
being like one of the peoplethat's leading.
He's the first federallylicensed psilocybin therapist
and he's in Utah.
They've rewritten and they'rerewriting the definition of the
(44:59):
terms, like religion, because ofhim and this guy had, I mean,
phenomenal fun, interestingstory, really genuine Like, hey,
this this is a reallyresponsible use of medicines.
And the police raided hisfacility.
They then found out that theyraided it and that was actually
harassment.
They had to give him backpsilocybin.
(45:19):
It is a fascinating story.
He's like, why did you do thisman?
And he was just absolutely,just absolutely positively sure,
even though he had no datapoints surrounding him.
He's like, no, this is theright thing to do and the right
way to go about it.
So it's one of the mostfascinating just kind of you get
to talk to these people andthey're just absolutely
enthralling.
(45:39):
They're super, super fun totalk to.
So we're making a space wherewe can actually talk to them and
learn how they made thosedecisions, how they navigated
them and how they live, to tellthe tale.
So that is, bring the man.
Speaker 2 (45:50):
Amazing.
And when is that launchingofficially?
Speaker 3 (45:52):
That is amidst launch now, so it should be releasing
in the next like week or so, orthe first episode.
So yeah, you're on one.
So yes, I am, thank you, I amexcited.
You're absolutely fascinating,fun conversation, so I'm looking
forward to it yeah, but muchlike this one.
Speaker 2 (46:08):
I'm really enjoying this and, uh, I am really
wanting to know because, ofcourse, we'll put the socials
for the podcast for ranchai butwhat is one key takeaway that
you want anybody listening,whether they are a grad student,
an undergrad, an alum or justsomebody out in the world of
business who has stumbled uponmediascape insights from digital
(46:31):
change makers one of thebiggest insights I would say is
just, that's never.
Speaker 3 (46:36):
You're still very much on the front end of this
stuff.
So, in terms of like theadoption of AI, the coolest
thing about it is that a lot ofpeople think they're going to be
leaving behind their creativeside, and I would.
I would go.
I actually think you'reprobably closer to being able to
unlock it.
Any generation has been subjectto the tools at its disposal
and you're looking at adifferent set of tools and I
(46:58):
would just go look.
The sooner and the more that youget involved, I think, the more
fun you're going to have doingit, cause there is a ton of this
stuff that is absolutelyfascinating, super fun stuff.
It does absolutely like the testof a vocation is the love, the
love of its drudgery, and I gocool.
If you jump into the space, youstart learning kind of different
modalities, different AImethodologies, and you
(47:19):
understand what they can andcannot do, you'll be able to
insulate yourself, to makeyourself as as a non replicable,
as humanly possible, and, uh, Ithink you'll be able to get
more done in less time, and Iactually do really think it's.
Actually I'm very optimisticabout the future because of it,
but it is one of those where Igo.
I think the people that don'tthey look at it as a chore, that
are going to feel it as an everand ever chore just because
(47:41):
there'll be further and furtherbehind, feeling like it's harder
and harder to catch up, and Ithink there's the others that
jump in that deep and I thinkthey'll be going.
This is actually a really funride and ride the wave.
So get started.
You are not late, you canabsolutely catch up and it
actually really is fun andinteresting and I think you'll
actually be pretty amazed withwhat you find.
Speaker 2 (48:00):
Fantastic Thank you.
This was such a funconversation, as always.
I love learning things.
I've written down someresources and, while we've been
on this interview, this callthis Zoom.
So danbairdwrenchai Awesome.
Thank you so much.
Thank you, Annika.
Speaker 3 (48:15):
Appreciate it Likewise.
Thanks everybody.
Awesome, thank you so much.
Thank you, annika.
Appreciate it Likewise.
Thanks everybody.
Speaker 2 (48:19):
Yeah, thank you.
Everybody who's watching thisepisode, listening to it,
connect with Dan on socials.
He's really open to talking topeople about all of this
wonderful journey that we're allon and, you know, use him as a
resource.
Speaker 3 (48:32):
Yes, happy to chat.
Thank you, appreciate it.
Speaker 1 (48:35):
Thank you, appreciate it.
To learn more about the Masterof Science in Digital Media
Management program, visit us onthe 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):
This is going to be a fun one.
I am so thrilled to have DanBaird on the show today.
Dan, you are co-founder andproduct lead at Wrenchai.
You also have an upcomingpodcast, burn the Map.
Your assistant found me onLinkedIn and introduced us to
start having conversationsLinkedIn and introduced us to
(00:46):
start having conversations, andit's been a lot of fun to learn
from you and to be able to, likeyou know, talk about what I'm
teaching and the AI involvementthat I have and where I should
go and where my students shouldalso go.
So thank you, first andforemost, for being here.
Speaker 3 (00:59):
Hey, my pleasure.
This is fun, interesting stuff.
I'm fascinated You're in areally, really cool space.
I'm excited to be here.
Speaker 2 (01:05):
Well, you're in a very cool space, so likewise, so
you have an MBA in globalbranding.
I've done a lot of differentwork in strategy design.
I want to talk about how yougot from there to now, figuring
out and designing tools andprocesses.
(01:25):
You have patented AI-drivensegmentation processes.
Speaker 3 (01:30):
I have always been a builder.
It is one of those things thatkind of runs in my blood.
Third generation entrepreneur.
My dad owns a foundry.
I grew up working for him.
He's a master craftsman.
He built planes in our garagewhen I was a little kid like
super cool stuff and not like wedo not come from money or
anything.
This is poor people building aplane in the garage.
(01:50):
So I was always surrounded bybuilding things and I've always
really had absolutely a passionand soft spot in my heart for
the people that go from zero toone.
I think it's one of the hardestthings you can do and growing
up I'm in Salt Lake City, utah,and happened to have one of my
friends start a little company abeanbag chair company, as a
joke called Lovesack.
Speaker 2 (02:10):
Oh, as a joke, yeah, as a joke.
Speaker 3 (02:13):
So we went to a drive-in movie theater, threw
that thing out.
Speaker 2 (02:16):
It was a joke.
Speaker 3 (02:16):
But literally everybody went I want one of
those things.
So we started a company.
Lo Sack went public in 2019.
So you can actually take thingsliterally as a joke, like we
would do that job because theeconomy was terrible and we
could give ourselves really cooljob titles that no one else
would give us.
But at 19, I ended up having 30employees and those beanbag
(02:38):
chairs if you look around thecountry you'll see them in malls
Travis, barker and other peopleare doing their commercials and
it's just a really fun and kindof cool concept.
But it was also born out of therags to riches kind of like.
We did guerrilla marketingbefore it was called guerrilla
marketing.
We bootstrap stuff before itwas called bootstrapping.
So that was kind of my firstforay.
I'd go to school until noon andthen I'd go work in the
(02:59):
lovesack factory until midnightjust to get stuff out.
We'd sell out multiple times aday.
It was just a really funconcept.
They tried to get me to quitschool and I was like I don't
know if I'm that guy.
They're like will you pleasemove to Tijuana and run our
production facility?
And I was going like I'm like ayear away.
Guys, I really wanted to.
So I ended up opening.
I said, instead of giving mefounder stock, I want to open a
(03:20):
bunch of franchises myself.
I did that in California,arizona.
I found Thunderbird.
Thunderbird is my MBA schooland a thing that I wanted to do
there was I was going.
What is the best way I wouldget you know, cause I learned
how to make product.
I learned how to make retailstores, but I was constantly
putting out fires and I wantedto go okay, how do people that
really do this globally do this?
(03:41):
So I wanted to go to get an MBAso I could learn how the world
was run.
And that I went and did.
Yeah, as you mentioned,branding and a dual capstone and
product management was theclosest thing I could find for
entrepreneurship and productmanagement.
Those were the two things,because if you're doing product
management, you're building aproduct inside of a corporation
from zero to one.
So I loved building that stuffand you know I had one of my
(04:02):
professors invented otter popsand I thought that was the
coolest thing ever and reallyalways liked the people that
were building things.
So I went and did innovationstrategy for ConAgra.
I worked on Healthy Choice andOrville Redenbacher and stuff
like that and our job a smallteam of 12 of us was go build a
platform.
It wasn't new flavors, they'relike you need to launch an
entire platform.
You'll get three-ish a year andif they don't do 20 million in
(04:24):
their first year, you're afailure.
So it was a really fun, cooljob.
You'd fly around the country,do focus groups, do prototypes.
You had really they reallyfunded that group.
So we got to have a lot of funbuilding with some really cool
companies.
And in food, which is superregulated and you have to be
really, really careful.
But if you go down to yourfrozen aisle today and you go
look at the frozen lasagna, youcan read my prose on top of your
(04:48):
Marie calendars.
You know it's that.
So fun stuff, cool stuff, hardstuff.
And after that I took a littlebit of a trip around the world
and came back was seeing a fewthings that I thought were
really cool.
Crowdfunding at the time was abig deal and I was seeing an
explosion of entrepreneurshipand it doesn't sound like a big
deal, but crowdfunding and justcrowd dynamic based businesses
(05:10):
are really really cool Causethey're kind of like open source
.
They flip the script and theyuse like the population and,
like it's a, when we were doingLove Sack and selling out
literally like multiple times aday, you would have to go buy
your product from China and youwould have to wait at least 90
days for it to show up, so youwould run out and then have
three months of nothing on theshelves until it got here and
the logistics actually killed us.
(05:31):
We had a really profitablebusiness in terms of margin, but
our cashflow got murdered justbecause it would take 90 days to
do so.
So crowdfunding actuallyallowed for the first time a
really big deal allowedeverybody to predict and measure
demand before they actuallypaid for inventory.
That doesn't sound like a bigdeal.
That is a really, really,really big deal.
That's like breaking down thefourth wall of a lot of
(05:52):
strategies, because you couldactually understand exactly how
many units do I need to buy.
I thought that was really cool,started a company called Crack
the Crowd and we opened thatthing up and we ended up being
kind of one of the premieragencies and very specifically
for equity crowdfunding in theUnited States.
We did some of the first multihundred million dollar raises
for real estate, stadiums andstuff like that Found out.
A good like this was a lot ofthe precursors to tokenomics,
(06:16):
right to a lot of the cryptotechnologies and things like
that.
So we were on the front end ofthat and it was really Wild West
.
There's a lot of stuff hey,we're raising money for a
submarine.
Hey, we're raising money for asubmarine.
Hey, we're raising money for alot of things we're like is this
legit?
A lot of them were, a lot ofthem weren't.
A lot of them were asking us todo things that were, ultimately
, that people didn't know this,but illegal, just like hey, you
can't.
(06:36):
Little things like when you'reraising money very specifically
for stocks and equities, youcan't say we will make money.
You can't even say we will,right, you have to be really,
really, really careful.
So it was a little too WildWest for us, because people are
going well, that guy's makingmoney, I'm going, he's making
money, he's also going to go tojail in five years.
And we pivoted into wrenchwrench, which is now what we do,
(06:57):
where we took a ton of thesegmentation and the work that
we did there and understandingwho is your audience, how do you
find that audience and who,what, where, who do you talk to,
what do you say, and how do yousay it?
We use a lot of AI, multimodalAI, to do natural language
processing, we do, yes,generative, we do machine
learning and we actually mix abunch of different models to
understand, again, who is myaudience, who am I talking to?
(07:19):
What can I do to enhance ourconversation and their
understanding of it?
How can I streamline theconversation and how can you do
it at scale?
So we focus on a lot ofpersonalization technology and
we focus on a lot of measuredcontent-driven strategy.
What do we say?
And again, why Not just hey?
The LLM told me to, anybody cando that.
We actually went hey.
(07:39):
When dollars are on the line,the differentiator between those
two is going to be who actuallyhas the data to justify their
actions.
And when you're at really bigcompanies, they demand that
level of understanding.
That's what we do at Wrench.
So that's how I got there.
I got here A lot of that typeof work.
Speaker 2 (07:55):
Yeah Well, and I appreciate that you mentioned
you mix a lot of different typesof what falls under the
umbrella of artificialintelligence.
Because that was one thing Italked to my students about and
they're like, oh, I didn't evenknow there was a difference
between AI and machine learningand deep learning and then
generative and the use cases foreach.
(08:16):
I'm like, yeah, there is adifference and we need to
understand that.
The nuances.
Speaker 3 (08:21):
It's a big deal.
I mean like it's all ultimatelymath, but you're talking about
arithmetic versus algebra,versus calculus, versus string
theory, versus.
They're different types, yes,but they accomplish different
things and a lot of them arereally good at one thing and bad
at another and for what it'sworth, we're going through.
You know, people talk aboutlike the hype cycle and things
of AI.
Like we've been doing wrenchfor seven, eight years now,
(08:43):
right, so like that's the joke.
We were doing it before.
It was cool is kind of a joke.
I'm a pioneer dinosaur and it'seight years, like not even old,
but it's a really big deal.
A lot of people are making a lotof mistakes and that's a really
really probably the biggestsingle mistake that they make is
not understanding thedifferences between the two.
Right, machine learning ispretty close to like statistics
(09:04):
on steroids, right, it'smathematical based, large data
sets, tabular data sets, meaningtable-based data sets, and then
generative.
It's like well, they took thosetabulized data sets and they're
just predicting the next word,right, I mean, they really are
fundamentally autocomplete onsteroids.
Most of the decisioning isbrand new.
They't been doing that and theamount of mistakes that they
(09:26):
make is like way more likeyou're like it looks really good
.
I'm like that's only becauseyou didn't check the answer.
You have to check the answeryeah and most of the time when
people don't, they're like, hey,it looks good and I go.
Well, here's the other thing,the uh, by definition like uh.
If you go through and you lookmachine learning, you move into
generative ai, llm, right, gbt'sbeing generative pre-trained
(09:48):
models.
They read the entire internet.
They put it all into a tablewhere they said, all right, what
word do I have?
Now it's like the word from,and they go okay, after reading
the entire internet, what's thenext most common word?
That gets placed right afterfrom.
And that's what they're doing,one word at a time all the way
down, which means most of thetime, after reading it, they
they spit out average answers,right, they summarize really
(10:10):
really well.
They're good at that.
But that also means again thatthey give out average answers.
That means most of the time.
If you're impressed by what theLLM did, you're probably
actually like a little bit belowaverage.
Or it's like, hey, look, it gotme an average answer really
quickly.
But if you're really good atsomething and you look at what
it gives you, you're like that'snot right.
Right, it's, it's.
You can almost tell that it'sfantastic, because that rising
(10:32):
tide raises all boats.
It does mean that we shouldactually see at least just
average content and output frommost of the product that comes
out of it, which means belowaverage is going to get higher
and higher and higher.
We're going to raise the stakesfor what average is acceptable.
But simultaneously, a ton ofpeople are making a mistake for
like hey, I use generative.
It's like cool, you have nodefensive moat, you are totally
(10:55):
replaceable.
Anybody could have done thiswith that amount of effort.
And again, it is one of thosetools where it's like you need
to be using it, butsimultaneously, like again
they're right where they say hey, you won't be replaced by AI,
you'll be replaced by someoneusing AI.
And that's largely one of thereasons is they're going to get
to the average answer, but thenthey're going to be doing
(11:16):
additional research to get anabove average answer.
Yeah, yeah.
Yeah, and that's largely what ifthey're just using AI.
They're replaceable, like that.
If you're not, if you're notusing AI.
Sadly, right now, like I mean,we just reached one of the first
times in my company's historywhere we actually had people
that we were hiring and we hadsomeone that came up, good,
worked in the space, highlyrecommended, and uh, we
(11:39):
ultimately didn't hire thembecause they weren't using AI.
It was one of the first timeswhere we just literally went and
looked.
You probably have one, two yearsleft before you are wholly
irrelevant Unless you start now.
You need to go look so youmight be good.
You're not going to be good forlong.
It's too easy to catch up now.
So for any of the peoplewatching this, it is one of
those, like I just we hired somenew graduates as well who are
(12:01):
jumping in and they're neck deep, but I can't stress enough.
They told me that at schoolthey were discouraged, even like
it was threatened as as thoughit was plagiarism to use LLMs.
Screw that, that is stupid.
I don't care if USC tells youguys to do that, use it anyways.
I'm not telling you to have itwrite your papers.
You need to know how it works.
You need to know when.
(12:29):
You need to know quickly.
If you don't, in five years youwill not be employable.
So go play with it.
There are different tools thatreally accomplish and excel at
different things.
You can go and use, like youknow, google deep research to go
write nice empirically basedresearch papers.
Still, you need to check all ofthe data sources, gamma to do
presentations.
There's any number of differenttools for different kind of
accomplished goals, images andstuff.
Chatgpt is really good at now.
(12:49):
They're really good at raisingkids.
My kids play with them at homeand talk to it because it's
fantastic at that type of work.
But for what it's worth, dig in, jump in neck deep.
I can't tell you it really doesmake me nervous.
My job is wholly replaceable asa CEO.
I can, and that is actuallykind of the work I do.
I can actually even show you Ihave AI agents.
I have a Dan bot.
(13:09):
They will, and they have beenprogrammed with about three plus
years.
I'm on calls like this about 30hours a week.
So I have transcripts, alltypes of video.
I plug them in that like thereisn't a huge need to hire
additional customer supportanymore because we can totally
replicate everything that I telland do them.
They won't want to just ask danlike we'll ask my bot and then
(13:31):
you don't have to, but I can bein five places at once now yeah,
have you used your bot toreplace yourself in meetings yet
?
Speaker 2 (13:38):
because we've talked about this, I have delphi.
I have a delphi profile formyself.
I haven't turned on the videocomponent, but I can have her go
in to meetings.
I can connect her to Zoom or tomy meeting schedule and she can
go in and act as me.
Speaker 3 (13:55):
So I have not tried that one.
That's an interesting one.
What I will do is not attendthe meeting but send my Read AI
bot to it and read and summarizethe transcripts and the action
items after the fact so that Ican go hey, I missed it, but
I'll follow up with you like 20minutes after the meeting.
I got all the high points, andthat way, oftentimes too, I can
actually pay more attention thanI can, and you know, sometimes
(14:16):
it's like an hour, it's like Ineed to be there for five
minutes.
Speaker 2 (14:19):
You know Right, right yeah.
Speaker 3 (14:20):
So I have sent it.
In that sense I haven't sent itin my place yet, but again, I
have messed with.
Hey, jen is another one toowhere we have we've taken and
programmed like scripts wherethey recorded me takes about two
minutes and then now they havebasically a live deep fake that
they can feed a script to and Ican sit and like read off a
here's what this new featuredoes, type of stuff, without
(14:41):
actually recording it.
So we've used that one as well.
Speaker 2 (14:45):
So it is absolutely kind of that's the next thing
I'm going to be adding is AI formy videos, so that, yeah, I'll
be showing it for class, I'll beshowing for podcasts, but there
are other instances where Imight not need to.
If I want to create, like yousaid, promotional videos or
project related videos for work,that I do.
That's going to save so muchtime.
Speaker 3 (15:04):
Oh, yeah, and it's actually better for the student
experience as well.
Right, I don't know if you'veseen it, but you should go check
it out.
There was a study.
There's been some people thathave been following a school in
Africa where they'vesupplemented the learning
experience with artificialintelligence and personalized
tutors and they are doing likeorders of magnitude, more like
kind of adopted learning andinternalized learning, like they
(15:25):
basically said that theadvancement from kind of that
personalized experience wasorders of magnitude of just the
classroom experience alone.
So at the very least, it isn'tsaying the classroom experience
goes away, but it absolutelydoes say that there are
personalized styles of learning,totally reflected, like we do a
lot of this personalizationwhere we'll like I can look at
your LinkedIn profile and see ifyou're a data-driven person or
(15:48):
if you're more socially drivenand you'll take and learn and
make decisions based off socialcues.
And it doesn't mean that one isbetter than the other, but it
does usually mean that youprefer information that way,
which means same thing forstudents.
It's like if I give the one onthe left data and I give the one
on the right social proof,they'll actually believe me both
ways and it's like, hey, thisis just how they perceive to
(16:08):
ingest that information andyou're going wow, so we can
potentially get twice as muchdone just by removing that
psychological friction which Ididn't have it turned on, I let
people do like 10 anonymousquestions.
Speaker 2 (16:29):
I was like, no, no, no, no, I need it because I can
go in and see all the questionspeople asked.
And I realized, oh, some peopleare just using it as the way
that they should use chat orcloud or you know one of these
other tools, or say, oh, can youjust wordsmith this for me?
I'm like that's not the bestuse case.
The best use case is ask meabout branding, public relations
, podcasting, things that are inmy knowledge base, right, and
(16:53):
ask me those questions.
And it saved so much time fromhaving to figure out when can I
schedule another meeting inbetween all my other meetings.
And intro, you know like I'mlike just go talk to her because
she is me.
And the great thing is thatthen even people can.
Somebody said, oh, I didn'tscreenshot or save the
information.
I said well, it knows your IPaddress.
Go back in, it's going to pickup the conversation with you,
(17:16):
even if you call her instead oftext or voice text her.
When you call that number,she's going to say, oh, have you
thought more about blah, blah,blah, whatever was the last
thing that you texted?
And so it's so amazing and thisis something I'm really excited
about.
You mentioned, you know I'mbuilding having your own agents,
and that's something that I'mlearning, starting tomorrow, is
(17:37):
how to build my own agents and Ito your point, like things are
moving so fast.
Some people might say, why areyou getting your MBA with
specialty in AI?
But it's giving me thatfundamental knowledge base.
I'm getting to play with toolsfor business use cases in class,
so it actually is expanding theknowledge way beyond what I
would have done on my own,because it's really easy to
(17:59):
delay learning when you havemeetings, when you have a life,
when you have all these otherthings you're trying to fit into
your day.
Speaker 3 (18:05):
Well, if you think about it, so like we had to
hypothesize what the world wouldlook like, we literally sat
down, so like we were in a spotwhere we were going hey, we're
going to leave the tokenomicsand the Coinbase kind of space
right now, we have enough moneyto actually do this.
We can make a decision.
We're at an inflection pointwhere we go we're not going to
be poor, so let's have some fun.
Our goal is to die with smilewrinkles on our face.
(18:26):
We're going cool, let's put adent in the universe and let's
die with smile wrinkles.
And a big part of that wasgoing okay, well, what can we go
into?
What do we want to go into?
And we talked about AI, wetalked about drones, we talked
about 3D printing, with allexponential technologies, with
an exponential growth curve,when half of success is just
showing up in the right place.
Right, we ended up choosing AIbecause we had some fundamental
(18:46):
skills in it.
We knew some people in it andthings like that.
Again, it was early enough thatpeople were like hey, granted,
it's been around for 20, 30years, but it was still early
enough that it was like hey,which one of these do we really
want to do and you think aboutit.
The organization that youtypically deal with, these large
Fortune 100 companies and stufflike that.
You have the bottom and the topof the organizational chart.
(19:07):
The bottom is basicallycollecting data almost directly.
They hand that to their boss,who takes a lot of predictable
types of data and information.
They look at that data, theymake a decision and then they
actually do the same thing.
They hand it to their boss, whogoes all right, how's this
division doing?
Okay, cool, how's this regiondoing?
And they all just basicallyit's turtles all the way down.
(19:29):
We're going in the future,you're going to largely be
sitting around with novel dataproblems.
Your job will be to generatenew digital exhaust, for lack of
a better term.
The repetitive jobs are theones that are going away.
So when you do the AI stuff,it's like going back and I'm
going like hey, we used to use arotary dialer where it would
take forever just to even call.
(19:50):
You hated people that had likethe.
The zero was the joke.
If you had a zero in your phonenumber, it took so long to even
just spin it.
That friction of how long ittook just to prepare that data
to make that novel decision waslike 70 and 80% of your day Data
silos, even four or five yearsago, were substantial Marketing
couldn't talk to sales, couldn'ttalk to product, couldn't talk
(20:10):
to support.
It still is kind of thatproblem.
You're seeing the barriersalmost break down now where
largely like the trick is, withall of these AI tools and stuff
like that, it's like well, wait,my job is to be the novel
problem solver.
So if they can aggregate thatdata, no-transcript, well, I got
(20:54):
into this business so I couldbake pies, but I'm telling
everyone else how to bake pies.
I haven't baked a pie.
Through that job, you automateeverything that is replicable,
everything that is predictable,and then you move up one level
in the org chart.
You're doing that same thingwith agents, where you go, okay,
(21:15):
move up one level in the orgchart, all right, now I've got
to write manuals for the peoplebelow me.
I'm going to write a bunch ofmanuals.
Now I've packaged up this joband I'm going to move one step
up the org chart.
Where you're basically going,I'm going to build not even just
one agent.
You're going to build anorganizational chart.
Beneath you.
You will have one person forcommunication, one sort of org
chart I've got you know I callhim Prof D but Professor G,
(21:37):
right, scott Galloway.
So I've got Prof D who does alot of my copywriting and
research and stuff, because Ilike the early smarmy swear a
little bit type of copywriting.
And I've got other ones thatare doing finance and proposals
and things like that.
But like you're going to buildan org chart, that just, and I
actually think it'll follow youregardless of where you work,
(21:57):
and your goal will be how muchnovel new decision-making can I
kind of internalize?
How much can I build underneathme so that when people send me
that work I can hopefully sendit through a completely
automated workflow and otherthings?
I've already answered thatquestion.
So your goal will be the mostproductive kind of org chart of
you're going to have that swarmof agents that serve you.
(22:18):
So it'll be nifty, it's funstuff, yeah.
Speaker 2 (22:21):
Some people might hear this and go.
I'm going to have a swarm ofagents that are working for me,
like my own agent army.
But you know what if one goesrogue?
You know there are people whoare going to be really scared of
that.
But then there are those likeyou and me who I'm, like, so
excited about that phase and thehumanity it's bringing back
into, like me actually beingable to do my best work and live
(22:43):
in that space where, you know,I'm still the decision maker,
I'm still the one who'sprogramming.
Maybe, maybe I'll let somedecision making, you know, come
down with agents, but but whereit's, it still has to be my
unique knowledge base.
Speaker 3 (22:59):
Yes, it will, and like I mean, if you notice you
follow these same patterns innormal society anyways, right,
there are, like there's alwaystech and AI as a tool, just like
anything else in the kitchen.
You can make dinner with it,you can cut yourself with it,
right, and just like that.
So, and what we found whenwe're building agents, we do
enterprise agents, so we haveone and like they're fun and
(23:19):
like our goal was we're notgoing to necessarily just build
the fastest or the mostconvenient ones, we want to
build the smartest ones.
And we've had clients that havelike built theirs and said I'm
not going to show my clientsthis, because if they knew this
was how smart they could be,they wouldn't hire us again.
We pulled them back in so theirclients wouldn't see them.
But what you find out is justlike humans, you get kind of a
(23:40):
saturation of memory overcapacity.
You can only pack so muchknowledge into one.
It slows it down.
And so what you do is youreally you have, okay, now I've
got one agent who's anorchestrator and they basically
their job is to take a task andthen to divvy it up into one of
the sub agents beneath it.
You specialize in finance.
I'm going to send it to youbecause this agent over here has
memory just dedicated tofinance and they're a specialist
(24:02):
in it.
So you actually start to findout that even the agents that
you build have an org chartthat's relatively similar to
your organizations and, justlike your organization, there's
governance.
So you'll actually have asecurity agent.
That security agent looks forstuff like wait, is this one
someone trying to jailbreak ouragent?
Are they saying that they'relooking for access to databases?
So like in terms of going rogue, just like in real life, you
(24:25):
actually have one person who'sgoing hey, wait, before we
publish anything, does this looklike this has any nefarious
steps in it, or anything likethat.
So you actually see it kind offollows some of the same
dynamics we have in society.
And again you go yep, I've gota security agent, I've got a
marketing agent, I've got asales agent, I've got a finance
agent and they're all workingcollaboratively together.
(24:45):
Final checks go through mysecurity agent before they
actually get shown to the publicor anything like that.
So, for what it's worth, it'slike yeah, will they, could they
?
Yes, is there going to be badactors?
For sure there's people thatuse knives and every other tool
for bad purposes, but it's not.
You know it's.
It's manageable and it's.
One of those things was likethere's not a monetary incentive
(25:05):
on doing it nefariously.
So 90% of them like weabsolutely have to be careful.
It still scares me how goodthey are and you absolutely have
to watch that ethics andeverything else in terms of data
, data integrity.
But overall, I think it's, it's, this is the new normal.
This is not like pie in the skystuff.
This is now, this is next year.
That's fast and it really.
(25:29):
The thing that really reallyscares me more is what will
happen to the people that don'tadopt it soon enough, because
it's going to be harder andharder and harder to catch up in
many ways.
So it's worth jumping in andjumping in neck deep right now.
Speaker 2 (25:38):
Yeah, well, and to that point, I have a lot of
students who do work at largeenterprise organizations.
I have many who work at verysmall organizations that are
starting their own businesses.
You work with enterprisesolutions.
There are companies out therethat will create agents for you.
Speaker 3 (25:55):
I'm trying to pull up my email and see which how many
this week.
Speaker 2 (25:57):
Well, just yeah, there's a couple that I get
emails from, but I've also beentold oh, that one, you know,
it's the one with all the cutesycartoon like, well, it'll be
all your agents for differentparts of your workflow.
But I've also been told they'rejust working off of chat GPT.
They're not really probablygoing to do the whole thing A
little more.
Which could be good, right, butthey might not be as good as
(26:18):
they say they are, but that's anoption for people who are
smaller businesses.
So I'd ask you, what are thebest ways for people with
smaller businesses to adapttechnology, still be considered
like maybe not the earliestadapters, but still on the
forefront of the small businessAI integration?
Should they learn how to buildtheir own agents?
(26:39):
Should they look at some ofthese tools and how do they
evaluate them appropriately toknow that these are the best
tools for them?
Because there's also, as yousaid.
I mean, even just looking atthe MarTech landscape a couple
of weeks ago for a class, wewere like, yeah, this year there
are 15,000 MarTech solutionsand there are a lot more AI ones
, and that's the smallpercentage increase from the
(27:00):
year before, but it's still.
There are a lot of smallcompanies that say that they're
the best and they simply aren't.
Speaker 3 (27:12):
And they're probably going to go away, right?
So, oh yeah, and the the inmany.
Much of the good news for yourstudents and everything else too
is, quite frankly, that goesfor actually a lot of the large
ones too.
We're still figuring it out.
The joke is, is that like aiagents and agentic ai and
everything else is, and peoplewill say ai employees now,
because they're going.
It's one level above agentic,it's not just workflows.
The joke is that it's liketeenage sex Everyone says
they're having it, everyonethinks everyone else is doing it
(27:34):
and most people haven't figuredit out at all and they're
totally lying, pretending theyknow what they're doing.
That is really the truth.
So in 2024, I was talking to asenior solutions architect with
AWS and I was making theobservation like he said
something to the tune of likehey, 40,000 new startups have
showed up in this space.
And I went, yeah, and as a joke, just because you know that,
(27:59):
maybe a month prior, a studycame out of Europe that said
that something like only 25% ofthe AI startups out of Europe
actually had AI at all.
And so, as a joke, I'm likeyeah, maybe 40% of them actually
have AI.
So 40,000 startups are likemaybe 40% of them actually have
AI.
As a joke, he stops, he getsreally serious.
I'd be surprised if 40 of themhad AI.
(28:19):
So this is someone who seesbehind the scenes and has.
So there is a ton of them thatdo that, and it includes really
large companies as well.
So, for what it's worth, llmops have existed, llm operations
, and what you're talking aboutis like where we actually ingest
, we do this, but the only ahandful of companies really do
right now.
Where you're going.
This is a big deal, because alot of the interface that you're
(28:42):
getting to in the web is goingto be very personalized.
It's going to be be like hey, Ireckon to your point earlier.
I recognize your IP address, soI know I'm going to show you
these products and I'm going tosay it this way and everything
else.
You're going to be likeliterally, the, the web of the
future, is personalized to yourpreferences, which is really
really cool.
But the way that and what thatmeans is, is that that interface
(29:02):
, even the chat interface thatyou see, is actually probably
now where some of the mostimportant data is going to be
exchanged.
You know they say data is thenew oil.
They don't mean it's valuable.
Data is the new oil.
Oil is valuable as a commodity,yes, but it's valuable as a
commodity because it's a fuel,because you can turn it into
plastic and rubber and all typesof different things.
(29:24):
When you start to use those LLMinterfaces with those agents,
when you're typing to them,you're actually revealing a lot
about yourself what you'relooking for, what you're
thinking of, what worries you,what way in which you present
the same type of thing.
I can use the word choice andfrequency of the words that you
choose to put into them tounderstand how you would prefer
to receive information.
And so a lot of those agentictools don't understand that.
(29:46):
Actually, one of the biggesttools that comes with, like,
when you use, go build an agent,go build it with chat, gpt,
it's 20 bucks, right, really,really cheap.
But no, like that.
Hey, you're not paying for itjust because it's 20 bucks.
Most of those accounts and thecost that it costs them, even to
run the GPUs that answer yourquestions, is well over $20 per
month.
They're doing it becausethey're trying to buy your
(30:07):
business in perpetuity, right,they're trying to build market
share.
They're trying to do it because, even though you're paying for
that much the words that youtype in are allowing them to
build a corpus, a body of textin Latin, surrounding you.
They're basically learning whatexactly should I be saying to
actually win the hearts andminds of these people?
And they're building a customerfor life.
Anybody who is building an AIagent right now if you're really
(30:28):
smart, you should actually belooking into the LLM operations
behind the scenes.
So I'd go okay, you built anagent.
Can you actually store and lookat the inputs and outputs from
that agent?
Can you say how many times ithad issues where it froze?
Can you tell me how many timesyou asked a question versus made
a statement, versus provide anysemantic feedback to it?
That digital exhaust that justcomes from interacting with the
(30:49):
agent is worth its weight ingold.
When you start to look at thosecompanies, or if you're running
one yourself, I would say, hey,start paying attention to where
you can get those.
The technology has existed fora long time.
But when you start to think, ifI build an LLM enterprise agent
and I go and I give it to afortune 100 company, the yes,
there is the tool and the factthat I can get those workflows
(31:09):
back, but I said I can also comeback to the employer and go hey
, as a heads up, I know a lotabout what your workforce is
thinking, what they're curiousabout, how much time they're
spending on spending on specifictasks, and so when you go look
at them, are there a lot thatare going to be a flash in the
pan?
Yes, I would say it's thosethat don't realize how valuable
that digital exhaust actually is, and the ones that do and again
(31:31):
, the same thing when you'retalking about students and how
valuable this information is theones that start to realize that
, like, whether you're doinggraphic design right, graphic
design with the number of imageand generative AI tools is low.
If you understand that, I'mgoing to actually take and
understand how people interactedwith that image to turn that
into an even better image.
(31:51):
All right, now you've got adefensible moat going forward.
Same with, like business and MBAstudents as well, I would be
looking into the differentmultimodal type of AI that I
have available to me and I'd belooking at all right, well,
where do I get the data sourcesthat actually helped me make
those decisions and how can Ibasically build a model where
I'm going okay, we're going tomake and do this workflow with
ML.
(32:11):
We're going to do this workflowwith generative.
I'm going to become kind of alittle bit of a master of what
widgets do we have or what toolsare in the toolbox and how do
we solve these problems so thatwe can say, hey, we actually do
know what we're doing there.
We do know how to get better.
We can actually measure ourprogress.
Whether it's the small startupsor, again, the really big
companies, a ton of them stilldon't realize this.
(32:32):
It's like hey, if you can'tmeasure it, you don't know how
good you are at it.
You don't know how much betteryou could be at it.
The people that are going to bereally surviving and thriving
in the near future are going tobe tracking their progress.
Again, llm operations is thename for it.
I would look into those and thecompanies that are doing that.
They're going to be the onesthat stick around, I think,
because they'll have a muchbetter idea.
Like it's too easy to switchfrom.
(32:54):
Like, hey, I want to use deepseek today and I want to use
anthropic tomorrow and I want touse switching cost is zero.
They can go burn all the cashin the universe from their
investors to try to build marketshare, but I don't know that it
gained them any sort of like.
I don't feel any.
I don't feel that I own anyresponsibility to one or the
other.
If I switch, I can switch andthere's zero concern for me to
(33:16):
do so.
What's really all said and doneat the end is who do you own
and who do you haverelationships with?
Right, not who do you own, butwhat relationships do you own
and whom do you have thoserelationships with?
That stuff is a really big dealin the future and that's what I
think those agents willmonetize.
Speaker 2 (33:34):
So Amazing, no, but really interesting and engaging
information.
I want to go back a little bitto talking about your brands.
Right, so you had Crack theCode, crowd, crack the Crowd,
and then you moved to Wrenchai.
What was that transition like?
Were you able to bring some ofyour previous clients along?
Did you have to start fresh?
How hard was it to rebrand thework that you're doing, along
(34:00):
with having this organization?
Speaker 3 (34:01):
It was hard.
I mean, one of the things thatwe found at Crack the Crowd that
was interesting for us is thatwe did a lot of analysis, always
been a really data-drivenindividual, like I always liked
working in highly regulatedindustries for that reason, and
one of the things that we foundthat the mom-and-pop shops like
even the little Kickstarterthings when we'd go and just
observe them, but when we'dwatch like just little
Kickstarter campaigns do well orpoorly, and then we'd go watch
(34:24):
really big fortune 100 companybacked companies and we'd watch
they had the same issues and itwas largely in their launch
process and again how they kindof they took their crowd and how
they organized the launchprocess.
So there's actually likethere's like there's like called
crossing the chasm.
But it's this phenomenon whichreally fun for MBAs.
If you've never read crossingthe chasm everybody watching you
(34:46):
should go read it.
Malcolm Gladwell said it's oneof the most important concepts
in business.
You should go check it out.
But the premise is there's likeinnovators, early adopters, late
adopters, but there's actuallykind of a timeline of adoption
of a new product or technologyand if you follow that kind of
chronological rollout processyou can actually help your
products go big.
Google, use this ring, use this, like some of the biggest
(35:06):
companies you've ever heard ofin the history of the world use
this premise.
And we would use that premise.
So we would go and we found outthe biggest thing that everyone
made a mistake of as theylaunched and they treated
everybody and they said the samething to everybody all at once.
And, uh, they just went hey, welaunched and I went, you
shouldn't do that.
You should get kind of an innercircle and you should go tell
them hey, we're going to launchand I need you guys very
(35:28):
specifically to help me.
And then you take yourinfluencers and you go hey, I'm
about to launch, I'm going to docan you do a product review and
are exclusive or some sort ofthing to help me promote it.
And then, after you basicallygot those two people or those
two segments on board, you go tothe early adopters.
They're the people that lean in.
They stand in line for thesetypes of products.
They're less price sensitive.
(35:52):
What we noticed is that, like,even though everybody learned
this in business school at thetime and everyone knows what an
early adopter is, as soon as youget out into the field, it goes
, disappears.
Glitter, no one goes, wait,where's my early adopters?
And you go, well, they're there.
And we realized, like when wewere at Con conagra, this is
what they would pay us to do isto figure out where those little
tiny focus groups of earlyadopters was at wrench.
(36:14):
One of the first things wefigured out how to do was
actually digitize that processand recognize that if I analyze
parts of speech or other thingsfrom authored content, I could
recognize someone based on thelanguage and how they people
that know a category well aremore likely to use its slang and
its jargon.
No rocket science there.
What they don't and didn'tthink about is that if you use
(36:36):
that to your advantage, youcould go do things like go in
and create pay-per-clickcampaigns and search where
they'd go hey, if you don't have12,000 searches nationally,
your campaign doesn't haveenough volume.
So they would tell everybodylike hey, if you're going to
build PPC campaigns, make sureyou have that search volume.
We're like well, those likeinnovator populations and those
really early adopters are verysmall, very tight but very
(36:58):
fanatical populations and theyuse their own slang and jargon.
They have their own language.
So if we go bid on thosekeywords that are really low
volume that people aren't evenlooking for.
We had PPC campaigns that wouldconvert at 25%.
You're going, hey, like justnot even supposed to be possible
.
So the first thing we did wasbasically go, hey, let's
(37:20):
actually try to build that.
That's what one of the patentsis for, is basically hey, you
can do that for any category,and if you can do that, you can
actually take, instead of doingpersonas, where most people go,
hey, we generated our personasand then we threw them in the
filing cabinet and we neverlooked at them again.
We don't know, you use personasbecause I can actually tell you
who someone is as they walkthrough the door.
The first time we can actuallylabel them and go hey, this one
(37:42):
is the early adopter, that one'sa late adopter.
One is not better than theother, but they take different
styles and if you adapt to those, you can literally quintuple
your conversions.
And so the transition wasactually relatively simple.
But based on that just becauseI mean it seems like a joke
We've been doing segmentationsince the 50s, of 1950, right,
(38:04):
rapid prototyping and stuff.
People thought that, like, thiswas all new and digital.
No, no, no, no, no, no, no, no,no, no.
But until very recently, peoplewould say, like my segment is
males, 18 to 24.
When you're on the internet,your gender and age matters less
than it almost ever did, right,like.
And so these are things wherepeople are on the web who they
(38:24):
want to be, they search for andthey switch between products
with zero cost of switching.
And so, for that matter, I wasgoing hey look, the future is
going to be personalized content.
Personalized content, and if wecan jump on that bandwagon and
understand early who should wedo?
We can build go-to-marketstrategies with this.
We can double conversion ratesof existing agencies, blah, blah
(38:44):
, blah, blah, blah.
So the transition for us fromthat perspective was actually a
lot easier than we kind ofthought it would be, just
because that need for anyonethat was going.
It still happens.
The marketing teams you go allright, we're going to.
What's our new strategy?
They look around who has theblackest turtleneck in the room?
(39:06):
That person gets to make allthe cool decisions because they
look the part.
Right, they have those MartinScorsese glasses, those huge
producer glasses, right, itwasn't and still isn't largely a
space where a ton of data isactually really looked at.
And in the future and I'mtalking next 12 months, it's
going to start being the placewhere it's like did you bring
data In?
God, we trust all others bringdata, yeah.
So I think that transition iswhat actually made our lives
(39:28):
kind of easier.
It allowed us to self-fund.
We're profitable.
We're one of the only AIcompanies that was able to do
that.
Thank you.
That's huge.
That was not easy.
(39:56):
I've got the blood, sweat,tears and scars to prove know
professors and PhDs in datascience and stuff like that.
You don't need to know it anddo it.
People confuse me with a datascientist all the time I have to
crack them like no, no, no, no,no no, no, I'm not.
But it absolutely is somethingyou can learn.
And a big part again.
I wouldn't tell someone to goactually learn to code.
I would tell them to go learnobject-oriented programming,
(40:19):
meaning you need to understandhow the building blocks work at
an abstract level.
That's very doable, it's notvery hard, that's a very good
use of time.
I would maybe learn some Pythonand things like that, but I
would argue like, look just,your fingers aren't going to be
on a keyboard in the near future.
So do you want to understandhow they work?
Very much so.
Do you need to be the personthat's putting ones and zeros
(40:40):
together?
No, I don't think you need tobe that person.
But if you can and you're kindof continuing your study, I
would absolutely focus onknowing those bits of
information.
They're around, they're free,right, they're oftentimes well,
not always free, but like youhave MOOCs and you have Coursera
and things like that and thingslike that, you're in a world,
too, where it's like, whetheryou have a degree or not, the
(41:01):
person who is the most up todate on the newest information
and is the most into it is goingto be probably the most
pragmatic hire.
Speaker 2 (41:06):
So yeah, again, so much great information here.
This is such a fully packedepisode.
I do want to turn to yourupcoming podcast.
Yeah, talk about Burn the Map.
What's behind the name?
Speaker 3 (41:21):
So you actually just heard some of it.
So this is good.
So crossing the chasm in thatpremise is the most interesting
thing for me, because I used tolove that space.
I still love that space whenyou get into talk to like those
innovators.
So imagine a bell curve.
That's what those, thosesegment groups, look like.
People imagine their targetmarket and they imagine a
bullseye.
It's like if you took yourbullseye and you rotated it
(41:43):
sideways so you could see thatthe innovators were the bullseye
.
Sure, they're the people thatactually work on it.
They have almost no pricesensitivity.
They like this problem so muchthat they work on it in their
free time.
And then you get to influencersand you get to early adopters
people.
You're a couple rings out nowlate adopters.
These are the people that wantdiscounts.
(42:04):
They're not into the space.
You really have to adapt yourcommunication to get them
interested because they're justtotally not into it.
Those innovators arefascinating to me.
They're the people that doinsane things.
There's an old quote from neverunderestimate the ability of a
small group of people to changethe world.
Indeed, it's the only thingthat really has right.
(42:25):
It's so true when you go findout there are very, very small
groups of people that areactually the ones that put a
dent in the universe, and mostthe rest of us are just kind of
along for the ride.
Earn the map is the name of thepodcast, and it's because I
wanted to actually go talk tothose people not a not even
necessarily just the people thatwere the visionaries and the
pioneers that had accomplishedit, but I wanted to talk to the
(42:47):
people that were in the middleof the thick of it, because
they're actually in a spot, notnecessarily where a map is
irrelevant, but where they'vestudied it so hard they realize
they're no longer on thechartered map, right.
So so it wasn't anything to saywe disrespect the map, we love
the map.
It's just we've learnedeverything the map can teach us
at this point, and we'restarting to move into a new
space, and so the team is usingeven some of our technology to
(43:11):
start to highlight some people.
You basically go into a skillset and we look for and we
analyze their online behaviorsand we look for people that have
a specialty, almost to likekind of the fanatical level.
So we can actually we did thisyesterday with my assistant,
where I pulled up her map likeimagine a little scatter plot
bubble chart and it has likepeople from zero to 100 that
(43:31):
have an affinity for you,whatever else you want.
But I was explaining to her Iwas like, hey, look, so I put in
your bio, I measured everyoneelse's bio and how many times
they use words, phrases andabstract concept language the
same way you did.
And this is a ranking ofeverybody that's in your social
chart of how much they relateand have an affinity for you and
(43:53):
like, yeah, whoever this is isprobably a good match.
It ended up being her boyfriend, like out of literally hundreds
of dots on the screen.
She's like, is that Colby?
That's my boyfriend.
I'm like, yeah, there you go.
So we're doing that to analyzeand basically have conversations
with those people because theyare delightful Like.
One of our first episodes iswith Bridger Jensen, who has
(44:14):
been on the forefront ofstudying.
He helped develop one of ourpersonality models with us.
This guy knows the differencebetween like Myers-Briggs and
disc profiles and all the prosand cons of those.
He ended up being on theforefront of therapy.
He wanted to actually build anon-psychopharmacological I'm
not even sure if I'm saying thatword correctly.
He wanted to see if he couldactually build and measure the
(44:37):
utility of like, say, going toCrossFit instead of like taking
a pill.
Here's one.
There has to be solutions tosome of these problems that
don't require medicine.
He went into the medicinemedicinal space and he ended up
being like one of the peoplethat's leading.
He's the first federallylicensed psilocybin therapist
and he's in Utah.
They've rewritten and they'rerewriting the definition of the
(44:59):
terms, like religion, because ofhim and this guy had, I mean,
phenomenal fun, interestingstory, really genuine Like, hey,
this this is a reallyresponsible use of medicines.
And the police raided hisfacility.
They then found out that theyraided it and that was actually
harassment.
They had to give him backpsilocybin.
(45:19):
It is a fascinating story.
He's like, why did you do thisman?
And he was just absolutely,just absolutely positively sure,
even though he had no datapoints surrounding him.
He's like, no, this is theright thing to do and the right
way to go about it.
So it's one of the mostfascinating just kind of you get
to talk to these people andthey're just absolutely
enthralling.
(45:39):
They're super, super fun totalk to.
So we're making a space wherewe can actually talk to them and
learn how they made thosedecisions, how they navigated
them and how they live, to tellthe tale.
So that is, bring the man.
Speaker 2 (45:50):
Amazing.
And when is that launchingofficially?
Speaker 3 (45:52):
That is amidst launch now, so it should be releasing
in the next like week or so, orthe first episode.
So yeah, you're on one.
So yes, I am, thank you, I amexcited.
You're absolutely fascinating,fun conversation, so I'm looking
forward to it yeah, but muchlike this one.
Speaker 2 (46:08):
I'm really enjoying this and, uh, I am really
wanting to know because, ofcourse, we'll put the socials
for the podcast for ranchai butwhat is one key takeaway that
you want anybody listening,whether they are a grad student,
an undergrad, an alum or justsomebody out in the world of
business who has stumbled uponmediascape insights from digital
(46:31):
change makers one of thebiggest insights I would say is
just, that's never.
Speaker 3 (46:36):
You're still very much on the front end of this
stuff.
So, in terms of like theadoption of AI, the coolest
thing about it is that a lot ofpeople think they're going to be
leaving behind their creativeside, and I would.
I would go.
I actually think you'reprobably closer to being able to
unlock it.
Any generation has been subjectto the tools at its disposal
and you're looking at adifferent set of tools and I
(46:58):
would just go look.
The sooner and the more that youget involved, I think, the more
fun you're going to have doingit, cause there is a ton of this
stuff that is absolutelyfascinating, super fun stuff.
It does absolutely like the testof a vocation is the love, the
love of its drudgery, and I gocool.
If you jump into the space, youstart learning kind of different
modalities, different AImethodologies, and you
(47:19):
understand what they can andcannot do, you'll be able to
insulate yourself, to makeyourself as as a non replicable,
as humanly possible, and, uh, Ithink you'll be able to get
more done in less time, and Iactually do really think it's.
Actually I'm very optimisticabout the future because of it,
but it is one of those where Igo.
I think the people that don'tthey look at it as a chore, that
are going to feel it as an everand ever chore just because
(47:41):
there'll be further and furtherbehind, feeling like it's harder
and harder to catch up, and Ithink there's the others that
jump in that deep and I thinkthey'll be going.
This is actually a really funride and ride the wave.
So get started.
You are not late, you canabsolutely catch up and it
actually really is fun andinteresting and I think you'll
actually be pretty amazed withwhat you find.
Speaker 2 (48:00):
Fantastic Thank you.
This was such a funconversation, as always.
I love learning things.
I've written down someresources and, while we've been
on this interview, this callthis Zoom.
So danbairdwrenchai Awesome.
Thank you so much.
Thank you, Annika.
Speaker 3 (48:15):
Appreciate it Likewise.
Thanks everybody.
Awesome, thank you so much.
Thank you, annika.
Appreciate it Likewise.
Thanks everybody.
Speaker 2 (48:19):
Yeah, thank you.
Everybody who's watching thisepisode, listening to it,
connect with Dan on socials.
He's really open to talking topeople about all of this
wonderful journey that we're allon and, you know, use him as a
resource.
Speaker 3 (48:32):
Yes, happy to chat.
Thank you, appreciate it.
Speaker 1 (48:35):
Thank you, appreciate it.
To learn more about the Masterof Science in Digital Media
Management program, visit us onthe web at dmmuscedu.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
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.
The Clay Travis and Buck Sexton Show
The Clay Travis and Buck Sexton Show. Clay Travis and Buck Sexton tackle the biggest stories in news, politics and current events with intelligence and humor. From the border crisis, to the madness of cancel culture and far-left missteps, Clay and Buck guide listeners through the latest headlines and hot topics with fun and entertaining conversations and opinions.