Welcome back to Data Driven, one of the top 100 AI

podcasts where we navigate the ever evolving world of data

science, AI, and engineering. This week, Frank

and Andy are joined by a powerhouse in the AI and data

space, the amazing Lillian Pearson. As a globally

recognized AI growth strategist and author of the data and AI

imperative, Lillian shares her journey from professional engineering to

data science to fractional CMO and how she's leveraging AI

to revolutionize growth marketing. From breaking down the barriers

of early data science gatekeeping to the rise of agentic AI,

this conversation is packed with insights, wit, and a healthy dose

of industry reality checks. So buckle up for an

episode that proves why data driven is a must listen in AI.

Hello and welcome back to data driven the podcast where we explore the

emergent fields of data science, AI,

and data engineering. And with me this week is my most favoritest

data engineer in the world, Andy Leonard. How's it going, Andy?

Going well, Frank. A little cold, but well. A little cold.

Well, it is, if it's cold by you, it's absolutely freezing

by me. I think we're down about two or three degrees colder than you. Plus,

we're on top of the mountain. Right. Mountain and just a

very generous term. Yep. Hill, I suppose, the West Coasters

would call it. But today, I'm super excited. And do you know why?

I know why. But tell our audience why you're

super excited, Frank. Our guest today is someone I wanted on the show for a

little while now, but we couldn't make it work. She lives on the other side

of the planet, but she's kind enough to have it here. Our guest today

is Lillian Pearson, a global authority on AI

driven, growth. She's the author of this book, The Data and

AI Imperative, and she's actually written a bunch of other books

and LinkedIn materials. In fact, her LinkedIn learning

course on, foundations of data science or something like that was one

of the first courses I watched way back in the day. I

was in the Microsoft office in K Street

watching, watching these courses, because it was

pretty clear that the, front end client development was

ending. The world was changing there. I didn't wanna be part of it anymore. I

wanted to switch into data science following your decade long,

sales pitch to me to get into the field. And the thing that made

Lillian's course awesome Great. Was yeah. She's got a

blush now. So if you're watching well, the thing that the thing that made her

stuff awesome was, like, she was the first not,

like mathematics or like MIT PhD person. Like she was the

first real and approachable person to do this. Although I

didn't know what PE stood for, I thought it stood for Princeton educated because

at this time, right, like everybody who was doing data science content was

a PhD. They said, you know, you gotta get a PhD. You gotta get degree

in this. And this is, like, 2014, '20 '13. Right? So, like, that They

were so gatekeeping. I mean, they were, like, Absolutely.

They were, like, you cannot come in. This is our gold

mine. Yes. Thank you. So welcome to the show, Lillian. No. You're absolutely

right. Like and and, like, they don't and even even, like, the the well meaning

people were gatekeeping. Right? Like, so, like, when I went to, one of the

advantages of working at Microsoft is you you are kind of behind the firewall. I

don't know what it's like now, but back then it was like that. Right? So

I was able to, like, talk to Microsoft researchers working on stuff. And I

would go to them and say, hey. You know, what do you what's your advice?

Like, this is my career dilemma. And they'd be like, well, this one guy, smart

guy, he's like, just go back to school and get a PhD. Right? Like, you

know, in this. And, like Just just go get a PhD. Like, you and me

would go to, like, 07:11 and pick up, like, a coffee or, like, a Slurpee.

Right? Like, just go pick one up. And, like, I heard that. To be

fair, Frank, to be fair, you and I both know a lot of

very smart people. And I know what a PE is, and

I know fewer professional engineers than I know

PhDs. Yes. That is true. So when, like while I

first thought it stood for Princeton educated. Right? Because at the time, this is a

very gatekeep field, like you said, Lillian. And what worried me is I went to

Fordham University. So I can only imagine the two letters behind my

name, and that was a joke. Well,

the PE actually it does it means

something. It means something and it wasn't easy to get. And I

gotta tell you, I think it just I went to

college. I took, you know, like I was saying, like, I took thermodynamics.

I took linear algebra. I took differential equations. I, like,

got an engineering license, but I mean, degree. But then you have to work for

four years under a PE and build

systems in order to get someone to sign off that you

have done this work so that you can then sit for another exam four years

later and have like, it's like taking the board exam to get this. So I

did all of this and it was like something I, I was more

like, okay. So I, I completed the journey. It took eight years

to do this. So that's probably why you see less one

of the reasons why you see less PhDs than PEs. But,

then I got this this license, which I love, and it

gives credibility. And I think that's important. But,

my husband, who is

actually a software engineer, so a software developer.

He's just like, why are you even maintaining this thing? Because

it's, for environmental engineering. So am I building

environmental systems or doing anything related to that anymore

and not at all. But I still earned this license. And to

me, it means something. So thank you for saying all that

because I, I like the validation.

To me, I think this means something. Come on. It counts. That's a lot of,

like, it's years of your life. Like, that's not trivial. I mean, that's like I

mean, that's like being a, like, a cardiologist. Right?

Like, you know? It was a lot a lot of work, and I have to

maintain it. I have to do every continuing education every two

years and all this stuff. I'm keeping I'm gonna do that even though I'm not

building sewer systems or Right. Air pollution stacks or

whatever. You know? That's fine. Whatever. No. No. I

wouldn't knowing what that is, it's I you know, I I

have more respect. I don't have a little amount of respect for

PhDs. I have a lot of respect for people who go through that education and

that process. It's not trivial at all, but I have more respect for

professional engineers. Yeah. So you're the first,

PE to be on the show. So that's something And they should count it.

Like, people like, oh, you have a master's. Can you you have a master's? Like

and I'm like, actually, no. I don't have a master's. I have a PE, but

people don't know what that is. I'm like, well, that's okay.

Anyway. Sure. Well, it became famous in The States. I don't

know what it's like, or or when exactly you left The US, but but

there was a court case, I think, in Oregon. There was an argument

over, something to do with the traffic light. There's something to do with

the traffic light. Yeah. So I remember this. So there was something to

do with the traffic light, and I guess, I I don't know the details of

the case. I'm sure Google I do. Oh, you do? The

the PE's wife, was charged with

running a red light. And he argued that

the yellow light didn't stay yellow long enough. Based on the

speed limit, and he did the math. He went to court and the

court. Hey. He he won his argument,

but the court didn't accept it. And they ended up appealing, and I'm

not sure exactly what happened on appeal, but I believe he did win on appeal

because the judge wasn't aware of what a p e was.

And they're like, No way. You know, the state certified that guy,

you know, as good as math. Okay? And other

things. So when he showed the math that there was no way she could have

stopped, Maybe. But the the, you know,

the fact that he he did the math and that wasn't accepted by the

court cause that's what caused the story. Yeah. That was a thing. Like, wait a

minute. Like, there he was a PE, and then everybody's like, what's that? Like,

so, so, like, not not I didn't know it took eight years,

but, like so it it definitely deserves more respect, in

the world than than I think it gets. That's okay. I don't even need it.

I'm not even, like, doing a technical role anymore,

really. Although it does help to have that background. Well, I mean, what are you

up to? Yeah. What are you up to? I couldn't even be to cut you

off. No. No. No. We had met on, like, a coaching call or something like

that. And because I think I reached out to you for career advice many, many

moons ago, like, and, you were like,

you know, I was like, but for me, the blocker was the math, like, getting

my head around the math. And this is, like, going on ten years ago. Yeah.

We got over all of that. Yeah. Oh, yeah. Yeah. Yeah. I mean, we're on

if I'm on the other side of the mountain now, you know, like, so, like,

at the time, you know, because you were like, oh, the math isn't wasn't really

a problem for me. And that's when I found out you were a PE. And,

and then I was like, oh, okay. Because but you were like, the coding was

the blocker. And I'm like, well, that's funny for me. The coding is not an

issue. The math was. Right? So it was interesting

because I think to your point and I'm sorry, Andy. We'll we'll get to your

question. Oh, it's okay. I'm just fanboying out. Right? So, like,

the, it's interesting how as a disciplined

data science right now, now I think the market's a little different because there's a

lot of experts out there. But, and

for those listeners, they didn't really see the the the wink at when I said

expert or the air quotes. But there were a lot

of disciplines kind of coming together that really formed data science. Right? You had kind

of the math the mathematicians, you had the coders, and then you had the subject

matter experts. Is that what you saw? Because you were in the game

at least three, four years before I was. Is that how it

started? Yeah. I mean, there were

statisticians who didn't that were, like,

essentially, filling the requirements of a data scientist, but then they would

call in the subject matter experts, that they

needed. And then there were yeah. I

mean,

I I had to hire. You know? I had to, like I was growing my

business, and I started in 2012. And I needed to hire people to

help me with requirements, and they needed they needed to basically be

data scientists. And there were no there were no

data scientists. So what I would have to do is I would have to take,

like, what one type of expert did, what another type of

expert did, and assimilate it into this thing that kind of

like a little bit of a Frankenstein in order to make it work.

Because there weren't and now it's so different. Now it's like, the market is

actually flooded. I mean, you can find people and it's, like, super easy, and it's,

like, all over the place. Like, if you go to Upwork, like, every job is

AI job. I'm like, this is not what it was. Let me tell

you a point. No. It's true. Like, people

forget. Like, when I made a decision to abandon kind of, you know, the the

front end development, GUI type stuff I was doing

and go into this direction. Even my wife who is a technologist,

right, but we're also a two engineer family, right,

was like, so you wanna study you wanna be an

actuary? Like, what what are you gonna do with this? Like, and and

in her defense, like, you know, ten, eleven years ago, this was

a risk. Now, fortunately, I backed the right horse after

after backing wrong horses a number of times, Silverlight,

Windows Phone, Windows eight. Right? So, you don't

have to get it right all the time, but you do have to hit it

once. Right? So now I think that's a good segue into what are you up

to now? Because I think what you're up to now, obviously, I have the book,

which is a really good book. I I haven't finished it yet, but,

I think you for getting it. I wish I had a good time your

review copy. Yeah. Well, that's your score. No problem. I

think I saw a post from you. Like, you said preorder it now, and I

was like, oh, I'll just preorder it now. And then it came,

like, right around New Year's. So,

very good book. I like the approach. But, so Andy

can ask his question or I can repeat it, but what are you up to

these days? Well,

I am acting I work as a fractional

CMO or I work as a growth adviser

and, strategist for

technology companies. So, actually, I'm not. I have done a

lot of work with b to b companies as you as you know, but I

have also the b to c, experience as

well as ecommerce d to c, marketing

experience. So I have just gone full throttle,

because I I had a role as a CMO

in 2022 for a data

SaaS company, a spreadsheet company. And as you know, I've

been advise advising founders and doing marketing, like, since

the beginning. So, like, that my first role in the data space was

even marketing, actually. So, and I grew

from there until, like, I got this job as a CMO, which I

thought was a bad word. I couldn't believe you wanted to call me a mark

marketing person. I was like, I like, put call me, like, chief product officer.

He's like, yeah. But my my investors are gonna like, they needed you

to be named for the function that you're doing, and you're doing a chief

marketing officer. And I would I didn't even know I was doing that. So then

I got that job, and I was so I gotta say I'm really good at

it. I've trained, like, ten years and spent over a hundred thousand dollars. Like, I

really this is, like and I didn't even know that's what it was called. And

once I did that, and once I saw, like, it was, like, then I

knew. So I so I've been doing ever since. And I just,

the data consulting, that was one of the reasons with the data and AI imperative.

It was important to me to, one, up level help, like, up

level, like, the execution people, the implementation data

people that kinda wanna move into leadership to help them, like, to share that

strategic thinking. And the other part of it was, like

because the strategy advising work I did as a, day

data strategist, like, I charge like, I was able to make a thousand

dollars an hour for that work, and I don't offer it anymore.

And what I basically wanted to do was just give away

the keys to the kingdom in terms of how the the process I use

to actually build these technical strategies. So I've been building

technical strategies for twenty years since I graduated

college as like my first job. Yeah. So anyway.

Interesting. So that's what I did with the book and it's a segue. It's basically

my coming out party is like, as a growth leader.

Which so as you as you'll see, like, the first half of it is very

much into product led growth, growth marketing, and how AI

is is, is is,

driving these types of growth in a powerful way. And then the second half of

the book is technical strategy. So it was kind of my way of, like, publicly

coming out as a, you know, as a growth and marketing

person rather than a technical person, which I had been pigeonholed into,

a decade prior. Sorry for the long answer. No. It's a

it's a good background. I think it also speaks to the

nature of marketing is changing too. Right? It used to be you know, you think

Mad Men. Right? Like, you know, idea people in Madison Avenue

come up with crazy ideas. But I think increasingly because of

technology, because of data, it's increasingly a data heavy or data

driven role. Is that what you've seen too? I mean,

that's your background is is kind of the data side. I mean, everything is

is data, and

my marketing approach is very much, like, evidence based. Of

course, evidence based marketing. Like, everything needs to be strategic.

Everything needs to be backed by data. It needs to be based on the

market data and evidence. But,

you mentioned something. I'm sorry.

Yeah. I lost my train of thought there. Happens to the best of us.

That sounds very interesting to combine those two, and I can see

how you get, I don't wanna use the word synergy, but

that seems like the best word. It's the the VINs overlap quite a

bit or the Yuleers depending on, you know, what what exactly

you're drawing there for the diagram. But I was gonna go with I was gonna

go with peanut butter and chocolate, kinda like that. Yeah. The growth

marketing growth marketing is all basically just analytics

and data data informed everything with your

marketing. So Yeah. Actually, today, I just came out, and we're

trying to get my YouTube channel going again. And as you know, it's a lot

of work to have all the processes in place. But we did a really

cool interview with the CMO of

single store, Madhukar Kumar, and he covered multi a

multi agent AI and marketing. And,

it's such an interesting conversation and, like, it's

basically, I'm talking to him what is AI marketing strategy. And

to him, it's like basically taking the principles of

data science and machine learning and infusing

that into the marketing approach for the

company. And that yeah. I mean, that makes a lot of sense. And even,

like, a lot of the companies I support have, like, AI products and features. And

so, like, I can get in you know what I'm saying? It's like, you kind

of really need to understand. So this summer

how the product work. This summer, I co wrote a book called

Sentient Marketing and it's definitely not exactly the same what you're talking about,

but it's definitely the idea that the the the main takeaway of the book

is that marketing and I data people and IT people need to learn to work

together because that's where the field is going.

It's gonna be increasingly data driven and led by data as opposed

to intuition, right? Or however whatever

traditional marketing methods were. And, those are

not

historically, those are not really great. They don't get along

marketing and data and IT. Is that That's crazy. That's

crazy talk, Frank. But I mean, how do you see those worlds kind

of working together? Like, what have you seen? Right? Obviously, I think the

numbers tell the story, but, like, what's been your experience? Right? Because you're

kind of you're on the leading edge of of this transformation.

Thank you. And, yeah, I can tell you just, like, as a person who

came from the technology, engineering technology domain and

into, marketing. Yeah. That was a

hard adjustment because engineers and technical people really

looked down upon marketing people. I'm like, really

do. And I was like, don't call me a

marketer. I didn't want that. Like, I thought it was a stigma.

But, like, now working as a CMO and I work with

technical founders, that's my my, you know, tech tech startups is

my market. So,

no. I don't see I mean, they might still, like,

look down upon marketing people, but I don't see because you

what what needs to happen, especially with product led growth, like, there's a

lot of marketing and psychology that goes into all of,

like, the levers in a product, like, to to build

referrals and to get retention and to, like, optimize the

interactions of users with products in order to increase

select and value, retain customers, get, you know,

re referral referrals from

existing customers. Like, all of that stuff is evidence based

data. You get the data from the platform. You optimize,

and you have to understand psychology. You have to understand. So

it's very much marketing,

but but it's executed through automation

that's built by technologists. So

whether one side doesn't like the other or not, it's a moot

point because we have to work together to to make

this happen. And so there's not gonna be the retention rates we need for the

company to succeed. And and the same goes for sales. Like, a lot of

times, like, the sales team doesn't want to, like, listen to the marketing team, and,

like, the marketing wants to, like, do their own thing. But, no, they have to

be married. They have to be, like, really, deeply

integrated. And I think it it it

I don't see a separation. But I also work with smaller,

more early stage customers. So, like, when you're working with corporations,

I think that they get a lot more siloed and it's trickier.

Yeah. I know that answer. Go ahead, Andy. Sorry. I I love that

answer because I think you're you you hit on

probably the thing that's, that's different about especially

engineers and and marketing people. Engineers aren't

typically known for being into psychology,

and marketing relies on psychology an awful lot. It's

not I'm not saying one's better than the other, but,

you know, navigating the strengths. And

and I love your analogy of calling it a marriage because if you're,

you know, if you have two people in a relationship that are

identical, that doesn't work well. To what you need

is someone with opposing strengths to to yours. They

they'll they'll compensate for your weaknesses, and that needs to go both

ways. Like Yeah. That's one thing I love about my job

is, like, basically, I'm, a a

consumer or customer advocate. So because it's very when

you're building the product, it's very easy to be

very interested in the product and how the product works and all the things about

the product. And, like, so I'm always thinking about the customer. Does that

Mhmm. Like, what's in it for them? Like, why should they care?

And, like, how do we get them to time to value down to, like, they

wanna give, like, two like, they care two craps. They do not

care about, you know, generally, like, people do not care about the solution. They just

want the out. They want the result, and they want it as easily as possible

with doing as little brain work or

investment of energy and time as possible. So I'm always, like,

advocating for that. Whereas when you're building the solution, myself

included, when I'm building the solution, it's so easy to get into the

details that you've, like, it's all about the solution, but it's, like,

you know, from my world, it's all about the customers and, like, the results.

Sure. Yeah. There's all these trees, and it

turns out there's a forest. Exactly.

No. And it's particularly, if you come from the technology angle, it's very easy to

get distracted by the shiny objects,

especially the new stuff. How do you see?

You mentioned agentic. Right? And that is, you know, we're recording

this on, 01/21/2025. Agentic seems like

it's gonna be the buzzword of the year. What's your take

on this, and how do you see it changing?

You think so? Yeah? Agentic? I just seems like

just reading the the tea leaves and kind of, like, you know, a

lot of the research papers, a lot of the buzz is all around agentic.

I'm personally not convinced just yet, but it

seems like a lot of people I think part of it is

founders that went down the rabbit hole and they invested, like, their whole top level

of marketing messaging around agentic. And then,

like, then they came to me, like, at the end of last

year, and we're like, no one knows what agentic is. No one knows

what agentic marketing is. It's like I think, like, in,

like, like, Silicon Valley, they know there there's,

like, a lot of hype around this. And I think that, yeah, there's a lot

of possibilities. But I also think, like, in the real world,

people don't know what that means, and it's probably pretty hard to sell.

Because you gotta look at the market size and the problem

you're solving and the urgency. You know? So if it's nice to have

and, like, how how easy is it to reach these people. And I think,

like, there's so few people that even know what's what that

means. There's also no yeah. Absolutely.

And there's no there's no consensus

definition of what makes something agentic. Yeah. Right?

So I I just I'm not sure if

it's a, you know, hey, look, we're, you know, the generative AI hype wave now

is two, two and a half years old. You know, now we need a new

thing called agentic generative AI. Right? We need a new adjective to make

it kinda continue. That's kinda my you know what I mean? But I

also think that there might be some legs to this. Right? Because I think that

the there there is the notion of like, and again,

it all depends on how you define agentic. Right? So for my purposes of

thinking about this, agentic is an AI that can

do something. Right? Like, you know, so bit like the Nest

thermostat. Right? Oh, it's gotten cold. You know, it's this time of

year. Raise the temperature. In a sense, in

a very kind of way, it has some kind of agency. Right? So in my

mind, that that that's agentic. Now I've seen

people take robotic process automation and

slap a new coat of paint on it and call it agentic, which

from one definite one look of it, like, I could see where you could justify

that. I don't think it's true agentic AI. Like, what's your take on this? Because

you you mentioned you work with startups. They they they went hard on this,

agentic message. Do you think maybe they did it too soon

or, like, it's just it's a evolving market? In this

case, I think that they were in Silicon Valley, and they

so it was, like, it it was being pushed really, like, a lot of hype

around this in the end of last year. And, I know I

think there's a ton of possibility. And I've been interested in

in, like, AI agents. I see some Facebook ads

like AI agents. And, honestly, today,

after, like it's interesting because I I'm, like,

publishing this video on multi agent marketing,

multi agent marketing, and then I'm, like, trying to build this process

for my team member so she can, like, SEO optimize everything and take it over

the blog and SEO optimize everything because I need to delegate this whole thing over,

and she's never done it. And I'm just, like, looking at all this, and I'm

like, why am I doing all this? There's gotta be an agent

that can integrate between WordPress and YouTube to

to do, like, an integration with some

sort of agent generative AI agent to, like,

populate, like you know what I'm saying? I'm just like, that's gotta already

exist. You you you're speaking my language

because I have a system I wrote called Dingo,

that does this. Okay. Does something very similar. It.

It basically, if you go to franksworld.com, this isn't

an ad for Dingo because I'm I'm I'm I'm I'm actually on the fence about,

like, should I open source this? Should I make it a SaaS? And this

is something that Andy and I can be going back and forth with for a

while. But, if you go to franksworld.com, I basically have,

it's called Dingo. Originally, it was named for one of the dogs you saw because

he looks like a Dingo. Yeah. And, I I I back

acronymed it to data in data goes in, data goes out. Right? That's the idea.

Right? So, like, you could I could give it a right now, it exists as

a command line program where I basically can

take a YouTube URL, and it goes out. It

pulls the transcript for the YouTube URL, generates a blog

post based on my writing style,

and generates the blog post, pulls the YouTube

metadata. So I have the tags. I have everything kinda, and it's all pre placed

into my WordPress blog. This is how Frank's world, you know, can

get hundreds of blog posts per month because I can automate the process to such

a degree that I do that. Now does it do the SEO? Doing?

That's what I'm doing. Yeah. So, as luck

would have it or misfortune would have it, whatever you

wanna call it, when I noticed that, OpenAI

knows my writing style because it was trained on a lot of my articles for

MSDN. Now I was really mad for about thirty

six hours, because I spent a lot of time with lawyers

over the last couple of years, one of which was the custody case. I

kinda asked my lawyer, like, hey. Like, they totally took my writing

style. Right? They totally trained it on my algo because I asked it a

bunch of questions. I could tell they pulled it from my MSDN articles, and there

was a lot of them, like, maybe fifty, sixty of them. And I asked my

lawyer, like, what can I do? And when she was done laughing, which is never

a good sign when your lawyer starts laughing,

she's like, there's not really much you can do. And part of it was that

when I signed the contract to write for MSDN, it was kinda like they

own the content. I was like, oh, well. But then after

about twelve hours of calming down from that, I realized that, no. Wait a minute.

Sam Altman did me a giant favor because now with the right

tuning and the right prompt, I can get it to produce articles

that it looks like I wrote because it was trained on

all my material. Mhmm. Or not all my material, but a

large corpus of documents that were that write like me.

And I've tested it, and that's basically what informs

Dingo. Right? So, like and it's also modular too. Like, you

can change out kind of the things. Right? I didn't you can also point it

to different blogs. Right? So, like, I have a a quantum computing blog that, you

know, if I change the parameter, it'll go that. But don't wanna go down this

rabbit hole, but stuff like that does exist. And, you know, as you are a

SaaS expert, we should probably talk offline after this and get your

thoughts on this. But, but no. I mean I mean, you're right. I

mean, like, I guess that's kinda agentic. I didn't wouldn't think of that as agentic

because I still kinda have to kick it off. But, I guess

what I'm thinking of is and this is probably the buzzword for 2026,

is autonomous agentic AI because we all you know, you know how it is. Like,

you know, the hype cycle you need to add in that adjective every every so

often to keep people interested. Yeah. Yes. Sorry. I I I

totally, like, went on a tangent. You could tell I had good coffee. No. I

see this on your website, and it's really interesting.

I like the idea of Dingo. Yeah.

Yeah. And Frank, Frank's being as as you know, you if

you're listening to this for the very first time and you're hearing Frank talk about

it, you're like, well, Frank talked an awful lot about that. He didn't cover

half of the functionality. So Dingo grew out of

trying to automate things to get the podcast,

out, generate transcripts. And That's right. He's taken

it even he's taking it even farther. I'm not gonna let the cat out of

the bag, but he's been experimenting for probably six

months now with yet another feature that that I won't

mention. But I get to listen to the results, and it's

awesome. I have this tool called cast

magic, and it basically I love cast magic. Yeah. And it

sounds like that's where I get I'm getting my content girl, like, just don't

even bother. Just, like, just use cast because you can put in custom

prompts, and you can train it on your writing style. So we use

GPTs, and I have purchased GPTs, and then I build them based on my

own writing style. But, like, Cast Magic

pretty much So Cast Magic everything. CastMagic is really good for pulling

the transcript, and it's really good at pulling transcripts. We should

we should totally have a Did you see the content, though? You can do custom

prompts in there now. You can do custom prompts. And what's interesting is is

that well, I love CastMagic. First off, like, I could totally fanboy out

on this. I bought it off AppSumo, which if the folks don't know what AppSumo

is, awesome. So you know what AppSumo is. AppSumo is freaking awesome. And

his book is even more awesome. Noah something. I forget his last name.

He has a hundred Noah Kagan. He, hundred million dollar weekend or something

like that. Something like million dollar weekend. Excellent book.

But there's a lot of good tools there. And you

basically you buy, like, you know, the deal that if you bought it off,

off AppSumo Cast Magic off AppSumo, we have sweetheart deals that

no one else could get anymore. Right? Like, so, like, I put a lot of

stuff on Cast Magic. Tell you. And,

I put a lot of stuff in cast magic. Magic is good. I got my

like, I'm just like yeah. Because I used to have, teams,

like, a content repurposing team and, like Right. Honestly,

is they didn't charge too much and and I kind of, like, everything was

ironed out and would just get done for me. So now, like, building a new

bringing in a new person starting from scratch is kind of a lot of work.

But, like Right. Also, I'm, like, I don't need I would rather have

someone that uses AI and just, like, I don't need to

overpay for this. You know what I'm saying? Right. Right. It should just be a

a a function of compute, particularly, like, once you train it on

I think it was you and I on that initial call many, many years

ago, I was talking about, like, you know, well, I do this and I do

that. And you're like, you basically said something to the point of why the

hell are you doing all this yourself, Frank? You should

hire you said this, something like that.

If you said hell, I don't remember, but it was kinda like with that tone

of, like, that's how I remember it. Right? So, like, you were like, why are

you doing this yourself? You should get a virtual assistant. And then when you start

coming to pay for things like that, my wife and I are not on the

same page speaking of marriage and and complimentary skill

sets. Right? Points of view. Right? We're not always in the same page. So, like,

that has led me to be very

creative in terms of, like, how can I automate those? Right? So I have a

lot of things that I built. Like, for instance, if you drop if

somebody drops something on my Outlook calendar, like my personal,

office tenant calendar, I actually have a script that

will copy that to all my other calendars, work related

and per like, a bunch of other calendars. In fact, they even modified

it that, if it has the word podcast in

it, Andy gets a copy. Yeah. And so far that

mostly works. How does it know that something's

spilled, like, some physical liquids spilled on

the computer. How would it detect that? It just No. It doesn't it

doesn't do that. No. No. I'm saying they they put something on my calendar. Not

a Oh. So, like, if you schedule Sorry. I missed the No. That's okay.

It's conversation. When I said dropped it on my calendar, that's

probably the verb. I Okay. It was a poor verb choice on my part.

But, no. So, like, originally, I built it because I was

I left Microsoft to join a startup, and this startup was

one of the worst work experiences I ever had. And I realized very quickly that

I needed to get out before the SEC got involved

or before they ran out of money. Like, it was one of those situations. It

was, like, all hands on deck. And I realized I had a I had a

lot of meetings. I had I had to coordinate with, you know, the day job.

And I also had to coordinate a lot of these interview calls. So I was,

like, at the time, I just picked up Office three sixty five for

my family. And I was like, I'll write a

Power Automate agent. And that's basically what it is. It's a, it's a thing. So

when an event happens and the event is add a cap, add an object

to my outlook calendar, it then fires off a whole

series of tasks that then spread it off across different

calendars. So You're reminding me. I can't you know, like and that's

good. You need technical people to build these things. Because then it's like you

you built these automations. Like, I built this whole membership, and then, like, I decided

after I launched it that I just, like, didn't like the sales numbers, and it

was a whole learning experience. But then I built all these freaking automations and had

someone spend a lot of time and, actually, a lot of money, like, building. Like,

I didn't just buy the solution. I was like, no. We should build this on

WordPress, and we should, like, let's do it in let's

do it in Discord. So you need an automation for everything. And then it's like

something changes, and, like, the whole thing need breaks

and, like That's kind of the catch with No. I don't want anything. Like,

automation seems to be lean and nimble and

That's it. No. A %. Like, that's why, like, I haven't really modified

it because I want it to be single focused. So that

way, if one thing breaks, God forbid, it only affects

one thing. But that's where I see a lot of these RPA systems. In

fact, I did have an RPA system that predated Dingo that was

like a Rube Goldberg machine. Like like, you know, it did this. It downloaded this.

It did this. It did this. It did this. But one break in the chain

and the whole thing went kablooey. And then that became a

crisis. So you're right. Like, automation needs to be lean,

isolated, and, I don't know. I'll probably come with another word later,

but, the no. But I mean the whole thing with

agents as well. So it's better to keep them in just doing multitasking.

The agent the multi agents is basically just like anything else.

Like, you guys do you're a data engineer, Andy. So you know

MapReduce and you understand about others and master.

There's a master managing many different tasks at the same time. So I

think it's pretty much probably the same type of architecture with a multi

agent system. And that's where I see the

whole, you know, when people say agentic, what I hear

as a as a data engineer is, first

first, the engineering part of that is I want systems that are

decoupled and independently resilient. And

then I want when I start using them together, I don't want them to

be coupled, but I do want them to communicate. I want these

dotted lines between these disparate systems, and

I want that to also be somewhat resilient.

Not so coupled not so much that I would use the word couple to

describe it, but I want them to be able to communicate.

And I like the idea I learned this

from managing teams and working with people, is

that you get people who are experts in many different

fields that have many different strengths and accompanying weaknesses,

but who cares if they're AI agents, what their

weaknesses are, as long as they can

complement each other. And so that's where that dotted line

comes in between these systems with different focus.

And it's a little like, you know, wrangling cats at

times. And I I too have, some custom

G. P. T. S. That I I think we're around with every now and then

and even some stuff running locally. But it's interesting

to see how these systems when you get them just a little

right, they don't have to be perfect yet, but they'll start feeding

each other. And that's what I think of

when I hear the word agentic, and in my mind, my mind

goes to the word community. A community of

AI, bots, agents, GPTs, whatever you wanna call

them, that will work off each other.

And so far, I've managed to get them to go through maybe

two or three passes where one feeds the other and then

the other feeds back and that. But after that, it gets stupid.

They just start making crazy suggestions, but they're getting

better. That's so interesting. Yeah. That's what I do sometimes, like, when I'm, like,

needing to fill out forms and, like, develop,

like, yeah. I I use one one

chat channel one chat GPT channel to, like,

synthesize the information, the answer to the question, and

then fill that. Because usually, when I'm using

GPTs in order to execute something for

me, there's, like, a series of questions, and I have all this

source data. So I have to, like I use one chat

channel to synthesize the information from the source

data to fill in the answers of the other the GPT so

it can synthesize. So it's like I'm actually running it, but I'm

using almost most of the reasoning.

ChatGPT is doing it. I'm just, like, manually feeding one

channel to the other. I don't see that as an issue.

Myself. No. Right. I don't I don't see that issue with copying and pasting, you

know, pulling the response and pasting into another. That's how I started with

it. There's a guy on Twitter, Doug Doug Finke.

I think did we interview Doug? I know we talked about it,

but he's I don't think we've I don't think we have. But

Doug specializes in PowerShell, and he got into

AI. And when he started doing and he does these free webinars

all the time where he's literally hooking these together, so

he's doing the automation. And if that's the path

you're on right now, Lillian, you may wanna we'll we'll have to send you a

link to Doug's channel, and you can watch

every at least once a week. He's doing something for free and

just out there sharing. And I think it's amazing because it's

it's the I word, integration. And I

I love integrating these because that's a heart of

automation. Mhmm. That does not yeah. That

sounds interesting. I'd love to just check out what he's doing. And I and

I do have to say, like,

excuse me. Rhett Bless you. Bless you. I

did sort of thank you. I

I have a friend who brought me into

one of these companies. I I'm not

even gonna give a plug on it because I don't feel like they earned a

plug. But I will say that

this company used clay

to basically string together a bunch of

reason. It is obvious to me that they had string together a bunch of

reasoning nodes, using that were

all connected to OpenAI's API. So the

reasoning nodes were all driven by ChatGPT.

And, like because I I paid I needed to get I, like,

overbooked, and I needed to get a, market

analysis done, and I needed to travel to Bangkok. So I hired my

friend who's an MBA, and he's, like, a product leader from CNN,

and he really knows what he's doing. And he comes back the next

day with recommendations. And then I'm like,

I need some supporting data

reports. I need some you can't just give me recommendations. I need to see

where this is coming from. And he said, well, it came from, like, over a

hundred reports. And, so I'll pull the most

credible of them. So I go and he pulls the credible ones and

it supports the recommendation. I'm like, okay. That's great. Like, this is so

much more thorough than if I had done it and took, like, less time and

you got it done, like, amazing. But then I'm kinda like,

how did you get, like, a hundred reports? Because these were not

these were, like, Harvard business review. These were, like, serious,

seriously credible re reports. So he brings me into this

partner of his to see their technology. They built with Clay, and

I look at this thing. They press run, and there are some there are, like,

nodes. There's, like, 40 nodes. And the thing spits

out, like, 250 where where it had gone

to, like, 200 of all the partners of this

done all of the market research, all of the output data, and, like,

basically automated the entire assessment. And, like, that's how he

got over the report. He used CLiT. And it's like,

you know, if okay. So great. So, like, I'm like, okay. Great. So that

basically takes care of most of the target market,

like, market research stuff. Like, that's done and that's

but, like, as someone who understands, like, the full depth of what's

required for, like, go to market and to, like, product market

fit, like, you would be really dumb to just, like that's not like, I'm not

worried it's gonna take my job, but I'm also, like, you can spit out, like,

250. Like, this is like gold. Yeah.

So that's have you guys used clay? I've not used clay,

but I'm gonna put it on my list of things. Yeah. I just wrote it

down. And it's that's the thing, for so

many jobs that are are out there and if people

would would pivot into that mindset, it's like that's

probably a couple of weeks worth of work, and he came back with

it the next day. And so what that does is

it's a force a force multiplier. So instead of serving

20 clients a year, you can serve a 20.

And so if you're able to, you know, people pay for the result of

your work, if that's your value proposition, and it should be,

then all of a sudden, you've just, you know, multiplied,

you know, multiplied your income. Yeah. There's that and there's

also just, like, the sheer volume of

what you can get done, like, in the unit

time. So That's right. What I sell now, like, I just

couldn't have I yeah. It's basically the same thing you're saying. Like, I

couldn't have, like, sold the results I sell now with the time I

had without AI. Like, I'm totally dependent on it because,

like, a brain can't even reason that much. You know? It can carry so much

of the heavy lifting. You know? You just, like, oversight. But you don't have to

know. But the thing is, if you don't know, if you do not know the

ins and outs of what you're doing, there are, like,

it's a minefield. So, like, I'm not at all worried, like, oh, someone's gonna

come and take my job. No. No. No. No. No. Because go ahead and try

and use that. Try to use Chachi Petit to build a strategy

and, like, just fall right in, you know, fall right in it because you have

to really know what you're doing. You know what I'm saying? Sure. And it's

cool, but it's not gonna replace I'm not worried. He's like Very

much stuff. Ready. Yeah. So they I've I've

a very good friend of mine who is a, data scientist and does a show

every Friday at 2PM eastern. Oh, was that Lev? Lev

selector. He he and I had a conversation months

ago about this idea, and I believe the term he used was

reflection. And the idea is in training,

any type of AI. It's more of a training philosophy.

But the idea is if you want, a

helper, an expert helper, you

you ask it questions that you know the answer to,

And then when it gives you the answer, you give it feedback on that. Now

granted, you have to have a GPT that's at that level. I'm doing it with

something local. And over time, I've built virtual

Andy, who is also a data engineer. And

I'll, you know, it'll I'll ask it a question about how to build a pipeline

and some technology, and it'll come back with an answer. And it sounds

so confident because that's what they're designed to do, but it'll be

incorrect. And I'll remind it. Well,

yes. You can it it's a good idea, but

it's not physically possible to build, say, an Azure

data factory pipeline in this way because you're talking

about nesting iteration, and the it physically

will not allow you to do that. And then I'll let it,

you know, noodle on that for a few sentences, and I'll finally explain

to it what I would do. You know, real landing, not

virtual. And the trick is that in

the first iteration, your outer iteration, if you will,

you call a child pipeline, and that child pipeline

performs the inner iteration. That's how I work around it. But

as soon as I told it that, from then on, whenever I ask

it how to solve the problem, it didn't just say, well, just put this iterator,

an until operator, for instance, inside of a for

each. It would say, build a for each, call a pipeline, and in the

pipeline have an until. Stuff like that. And

that's the system becoming an expert because you tell it

to. And I think that's what I I'm not sure if the term was reflection,

but it's using what I know to make the make

the automation better at helping me. And it's

a it is very much a, you know, a positive

spiral and an accelerator. And it becomes even more

of a force, multiplier in in helping me

to to with ideas how to solve this problem.

I really like, I really like, even just

arguing against the reasoning,

with ChatGPT. Like a lot of times, not a lot of times,

but sometimes, I'll get done in an

analysis and assessment of something, and it'll come up

with something that I don't agree with. So

then I'm like, well, yeah, did you consider x, y, and

z thing? And then it might return back

an answer that's, like, something I hadn't thought of. You know

what I'm saying? So it was, like, a value, a lot of value. And,

like, the the, like,

I mean, because I'm one person, and this is an agglomeration

of the perspectives of millions of

people. So, like, it's not always right, but there are a lot of

perspectives and approaches that are right. And I have one

I have a lifetime forty five years of experience, but

it's collected generated, like, forty five million years of experience.

So, like, you can harness that. That's

true. Have you played with notebook l m at all? I'm just curious.

I have not. No. So it's interesting. So it's a Google

product, Google AI product, and it's really good. What it'll generate

is it'll give it a PDF, you know, document,

even audio file, video file, YouTube video, and it will generate,

among other things, a two person kinda NPR

podcast style interview where they're talking about

what it was trained on. And I find it useful because I

like audio content. It's, you know, in the car quite a bit

and whatnot. But it also can help

me think about things I hadn't considered.

So when they have these two hosts kinda debate a topic, it but it also

kinda it shakes loose kinda like the the the biological neural network. You know what

I mean? Like, where it it it it kinda like I hadn't

considered that angle. Right? And it's just I don't know. I find it

enriches, like, my brainstorming. And to your point, right, it is a

sounding board that is awake twenty four seven, assuming, you know,

it the servers are up and running. But it's

I find it fascinating in that it could be used that way.

And to your point, right, it's And also we have our perspective. Like like,

especially if it's anything where you're needing to speak, like,

an an executive level. Like yeah, I know everyone and their

mother, like, uses ChatuchyPT to, like, refine if they

need to send an email, they refine their email or whatever. But, like, it

also introduces perspectives. Like, if you're ever like, if you're in an

emotionally charged situation where you need to, like, it will

introduce like, I really like to use just critique this. And, like

Yes. You know, and, like, okay. Great. Like, actually, yeah,

you're introducing the perspective of the other person because, yeah, I'm a

CMO, and I think as a customer, except for when I'm in I'm in

the middle of it, and then it's a little bit harder, you know, for me

to, like, see the perspective of the other person. So, like, I don't need

it to necessarily call my best friend now and share

those, like, details. We could just, like, get a critique and, like, kinda,

level up that way. Absolutely. It

it is interesting that, that I find I kinda put that in

the category of of the empathetic AIs, and that became a

buzzword a few months ago. And the use

cases some of the use cases I find astounding, especially

with, with people who have experienced

military related, post traumatic stress. They're dealing

with those sorts of things. There's just a lot of success stories

coming out of that. But the LLM PTSD?

Yeah. Yeah. There's a lot of Could you share a little bit more about that?

I'm interested how an AI could actually help with PTSD.

It it's it's talk therapy. And because

it's trained in, you know, in in a lot of talk

therapy, it, it it does a a fair job

of that. I haven't, I'm not recommending it. I don't

know enough about it. I haven't looked into it enough to see that. But,

that's a topic near and dear to my heart is helping people who are struggling,

you know, with PTSD. I'm a, I was in the National Guard for

six years. We you know, it's nothing like what people who have

seen action. You know, I never saw anything like that, but

it's it's just a soft spot in my heart for people who struggle with

that. Yeah. I I same. I mean, I I

was late for work on 09:11, so I have my own little dance with PTSD.

So Yeah. Right. Yeah. I,

had I not been late for work, see, there's a fifty fifty

chance I'd still be alive. Yeah. And it's just the

number of that it kind of weaving that into what you said

about, you know, what mindset is front of mind for for

you, getting ready to to speak to

some situation. And you may not be in the best mindset.

You may be because not any no malice. No, you know, there's

nothing wrong with what you're you're feeling or anything, but it's

just different. You're thinking customer and you need to talk

to a CEO. That's different. And and

necessarily so. It's not that the CEO is evil. It's not that the

customers are evil. It's that there's a difference there. So

just, you know, pull the emotion out of it and the judgment out of it

and just think about the communication style.

And it it kind of fold into this, a tweet I

saw. Gosh. It was 2023,

and data scientists working with LLMs. And he

said a lot of people make a big deal out of

LLMs hallucinations, and some of them are funny and some are

tragic. We we totally get that. But his point

was they always hallucinate. They don't know how to do anything

else. They're not that bright. It's only it's only called salus

hallucination when it's wrong. When it's wrong and and and or

ludicrous or, you know, infuriating. Glue on

pizza? It's fair. Yeah. It's

fair. But at the same time, when you think about what it's doing, you know,

especially down on, like, the vector database level, It is just, you

know, it's it's nearest neighbor or some other algorithm that

it's using to identify the next word based on to to

your point, Lillian, twenty five million

documents that it's looking at. And you've got you and, you

know, access to a handful of friends and experience and the

conversations you've had at one second per second,

you know, human speed. And it's put together this huge big

data, solution, which sounds

awesome. Attention is all you need. Yep. Yeah.

And so, you know, they the there are fallacies of

big data. Nassim Taleb has mentioned several in

his books, in his insert, where he talks

about the things that big data the past big data can lead you down

that are, you know, incorrect or wrong, and especially in the fields

of predictive analytics. The quality of your data can

be north of 99%, way north of it.

And, you know, but there'd be fallacies introduced into it,

and the analysis will produce the wrong result. Sometimes tragically,

horribly wrong. And, so I I'm I'm not

I'm gonna stop there because I'm starting to get a wandery. But

Yeah. No. No. This is great. This is what we do, Lily, and we kinda,

like, go on these different trains of thoughts. One of one person said we should,

we we get off track a bit. One person suggested that we sponsor an

off road racing team because we're always off the but I wanna

be respectful of your time. Plus in your time zone, it's probably pushing

11PM. May not may not. Yeah. I actually am getting a little

little tired. We wanna be respectful of time, but, as always, you're

welcome back in the show. Thank you so much for having me on.

And I enjoyed our conversation and, like, it's fun that we have.

Like, synergy about, like, AI, you know, like, we're

we're we're all we're, like, at least you guys I think we're all

kind of working in a different capacity, but we're,

having a lot of overlap in our experience, and I feel like that's

nice to kind of hear. Absolutely. It's always gonna be a different opinion

around the we're all dancing around the same problem. Right? And, like, some of us

are in a different orbit and things like that. Where could folks find

out more about you, what you're up to? LinkedIn.

LinkedIn is good. I'm starting a LinkedIn newsletter, by the way.

Cool. Awesome. I think so. Yeah. Well, I'll connect.

Yeah. Definitely. Thank you. Yeah. Let's connect, Eddie. I'll I'll find you now

so I can make sure that k.

I have there's a cartoon me as my avatar. So the

guy it's a little cartoon with a beard that goes off the frame.

Alright. Alright. Okay. So And,

Craig, thank you so much for having me on. That was great. Thanks for thanks

for joining us, and, I'm very glad we had this talk.

It's awesome. Oh, thank you. Okay. So I'm gonna connect. And

then, yeah, if you guys ever wanna talk shop about your startup

idea, I'm always Cool. Awesome. Please let me on speed dial.

Awesome. Alright. Well, with that, we'll let our British AI

who I suppose she's a bit agentic, Bailey finish the show.

That's a wrap for this episode of Data Driven. A huge thanks

to Lillian Pearson for sharing her insights on AI, growth

strategy, and the evolving landscape of data driven marketing.

If you enjoyed this episode, be sure to subscribe, leave a

review, and share it with your fellow data enthusiasts. And

don't forget you can find data driven among the top 100

AI podcasts. Number 38 to be precise,

so clearly you have excellent taste in podcasts. Until

next time, stay curious, stay data driven, and

maybe, just maybe, start training your own AI agentic

overlord. Cheers.