World's First Chief AI Officer Shares Insights

Thank you for having me today.

You guys actually have some really amazing speakers.

So for me to be able to kick it off, I hope I get

to set the momentum and the energy.

And yes, the theme of today is turning data into action.

But what does that mean?

We hear a lot of keywords, we hear a lot of jargons.

What does it actually mean?

So what I wanted to do is because I took a chance

and took a look at your profiles, your backgrounds,

what you've done for a living.

And what I want you to do, if you wouldn't mind, is look

to the person on your right and say Congratulations.

Just do it. I know it's kind of funky.

There we go, pat in the back.

And then look at the person to your left

and say condolences.

Because I have a soft spot

for individuals in data governance

and security, cybersecurity

and anyone who's even coming close

to touching the space of fintech.

And it doesn't matter if you're in product creation,

customer experience, if you're an executive accountable

for making those trillions of transactions work.

The fact of the matter is, it's kind of like

being in a telco company.

We all get p****d off when the Fleming

and the electrical doesn't work,

but when it is working, we're just expecting it to work.

So no news is good news.

When we get news, it's usually not so good news.

And I think that there is not that people are taking it

for granted, but the fact

of the matter is whether you're in product,

consumer experience, triaging an executive,

it doesn't matter what your role or function is, at the end

of the day, you guys are enabling trillions

of transactions happening across the banks,

across the credit unions.

And us mere mortals are like, eh,

it's meant to work seamlessly.

So that's why congratulations and condolences.

Just a little bit of background just

to build some credibility only

'cause you guys have some amazing, amazing stories.

And quite frankly career director trajectories.

I was a rugby player. This was not my career path.

How many of you guys thought you'd ended

up in this industry?

Show of hands, like this is

what you wanted to do when you grew up.

Yeah, I don't think any of us did.

We just kind of were what I say, the wrong place, wrong time

and took the right job and then the career was chosen.

But I was actually a professional rugby player.

I played on the US Women's national team

and then it was kind of time to grow up

and so I got a real job and that was to be a data engineer.

And apparently I was the only super

extroverted data engineer and I like talking to people

and I love talking to the business.

And so I ended up taking on this awkward translator role

'cause neither were my tribe,

but I was the go in between

and things just kind of took a turn.

I went from a practitioner, an individual contributor,

to managing the entire P&L

for IBM's enterprise data management practice

to helping them launch Watson in 2011.

Lots of fun failures.

It was amazing training ground

'cause it was the first opportunity

to commercialize artificial intelligence for the workforce.

And then from there clients hired me

and I got to be the CDO, CAIO,

CAO, CDAO

for some amazing employers like Merck Pharmaceuticals,

Sony Music, Estee Lauder, Amazon.

I was the head of technology for North

America for their startups division.

So perplexity Anthropic,

who you guys know were my customers.

And then right now I'm the Chief Strategy Officer

for a cybersecurity company

because with the progressive growth of data,

there also needs to be an equal amount

of attention on protection.

So that is me Cliff note version.

Alright, just to kick us off, this is the PG version.

By the way. What do AI and teenage talk have in common?

For those of you guys who have teenagers,

the kids are talking about it.

Nobody really knows how to do it,

but everyone thinks the other person's doing it.

So then they claim that they're doing it.

You guys getting there? Or do we need more coffee?

Yeah, exactly. And that's kind of the space

of artificial intelligence.

And I've had over 200 deployments under my belt.

And so it's interesting

because we hear about all the amazing capabilities of

what data can do, what AI can do,

but no one is talking about these stats.

The fact of the matter is,

and if any of you guys saw the MIT report, only 5%

of gen AI projects are actually succeeding at scale.

And for applied AI,

only 12% are actually going into production.

The majority of them stop, pause, or cancel at POC

or MVP, I call it perpetual.

POC purgatory. Things just stop there.

And unfortunately the news isn't so good when it comes to

data transformations.

These massive, let's modernize our tech stack initiatives.

Data's a new oil, let's leverage it as an asset.

It's also kind of grim. So the good news is today's

conversation is gonna be the reasons why, how to avoid it so

that when you're sitting in meetings, a framework, an idea,

a thought will pop up

and you can have a dialogue and a discussion about it.

Because I'm gonna aggregate about 20 plus years

of data experience into a few slides

and the past 14 years

of my AI deployments into a few slides.

But to really understand the gravity

and why all this stuff happened.

'cause we understand the concept, yes, data's important.

I'm not sure most of us are aware of the timeline.

So I'm older, not as old as 1950s,

but I was around when

SQL databases were a thing and this was my job.

No joke. This is

how I started my career in data engineering.

And it was cobbling together pieces of information

to make sense so we can create pretty

reports for executives.

That was like the extent of it. And then things migrated.

The invention of the cloud, big data ecosystems.

You guys were around when Hadoop was the buzzword

and everyone wanted to leverage Hadoop and all of the above.

But that still was not the pervasive growth of information.

If we talk about the volume, the variety

and the velocity of how things boomed.

And by the way, our generation is very important

because we were there when the dial up

internet came about.

We were there when mobile phones were no longer

restricted to just drug lords.

We all got one like those Nokia flip phones.

We were there with mobile apps.

We were there when email came out.

We are here right now with AI

Metaverse 3.0, blockchain

and the data boom, I think the past 15, 20 years

of our careers has been full of transformations.

And I'm not sure we're fully

understanding the gravity at all.

But it's also what's contributing to our inability

to keep pace with everything.

'cause the pace of change is just astronomical.

But the data boom actually happened between 2010 and 2015.

'cause everyone wanted commerce.

Everyone wanted IOT devices.

Everyone wanted to become digitally transformed.

And so we were gobbling, gobbling,

gobbling information which then allowed us

to create these amazing capabilities.

And voila, gen AI was born in 2023.

In my humble opinion, AI is nothing but a data product.

It cannot survive, it cannot live, it cannot breathe,

it cannot be sustained

unless the data is living even pre like breathing

and fundamentally up to date.

So now we take the history of data

and we outline the history of AI.

For some of you gearhead

and geeks like me, we know AI is not new.

It has been in R and D for about 80, 90 years.

Alright, fun fact.

Does anyone know how artificial

intelligence the term was coined?

70s Dartmouth, lots of substances at a conference.

It's literally how it came about.

It was a term that was coined at a conference in Dartmouth.

And it's a little bit

misleading and we'll get into that.

But the first commercial application was Watson

and then of course we now have it accessible to us.

But what are the biggest differences between Watson

and the AI that we know today?

Well, Watson was kind of like a winery.

You needed a lot of money to make a little bit of money.

It was only available to enterprises.

It was very restricted in inter industries

and it was not consumer available.

It was fundamentally built to be enterprise grade.

But then ChatGPT came out and voila, what happened?

We all got to play with it. We all got to tinker with it.

And that helped the democratization of it all.

And it learned off the worldwide web,

which is why it's also creating a lot of ruffles for us

ndividuals who are writers have a lot of IP.

I think you guys have recently heard of the lawsuit

that Anthropic has to pay a few.

They owe me quite a bit of money because I did a search.

And quite a few of the foundational models are leveraging

my patents, my book, my IP, without even asking

for permission or paying me.

So there's some, you know, goods

and bads that's come of it.

But you overlay the data boom with the AI boom

and voila, what's happening is our inability

to absorb information, to retain information.

We're constantly chasing our to-dos.

And for the obvious reasons, knowledge

used to double in the 1900s. Every 100 years.

It's why some of the old geniuses could memorize novels

and poems and you're like, oh, how could they memorize that?

I don't even know my sister's phone number by heart.

And then the mid 1900s information doubled

every 25 years.

Right now information is doubling every 12 hours

that LinkedIn post, that news article, that blog

the next day, it's completely refreshed.

That's the pace that we're living in.

So it's really, really difficult to keep up.

And for those of us that are in our field,

like I'm getting whiplash.

I have to dedicate two hours a day just to read

and understand what's going on.

I can't even imagine what it's like in other industries.

And the core fundamental challenges of a lot

of these things is we're not even grasping the fundamentals

because we hear a lot of the PR

and the tagline AI's amazing, data's the new oil, monetize.

But there's a lot of work

and a lot of effort that goes into it.

You don't have to raise your hands,

but let's do it a different way.

I'm not gonna ask you if you truly understand

AI 'cause I'm sure we all do.

How many, raise a show of hands.

If you don't think your boss knows

AI, we'll do it a different way.

Yeah, it's a two letter term. We'll just use AI.

Well yeah, AI can do it.

Well, how's it gonna fix customer data?

Yeah, AI's gonna take care of it.

Okay, but if on,

we're already done with summer,

but Thanksgiving, Christmas you guys go

or you're having conversations

or you're actually trying to explain it to the board,

the several executives back in 2018 back in a napkin

with MIT, this was created and I absolutely love it

and I adore it and it's mother proof,

which means you can understand it.

Well I'll give you an example.

I always use the test of the senses. Can it see?

Well if someone's trying to sell you a piece of software

about computer vision or oh, we can identify this

and I can identify that.

Can it see if the answer's no, it's not AI.

If it can see great,

can it actually identify the image that it sees?

If it cannot, it's a camera folks not AI.

If it can identify the image it sees

and with an element of accuracy,

now you're getting into

machine learning and computer vision.

I'll tell you where I can still trick

AI even though it's extremely advanced,

I will give it a picture of a tiger, no problem.

I will give it a picture of a dog taking a

nap behind a fence.

And there are shadowed lines

because of the way the sun is positioned.

I ask, can you recognize this image? It says a tiger.

It can't detect that it's shadows from the fence,

but it assumes lines on a body of a certain structure.

It must be a tiger. So it can't pick up the nuances.

And this is where we become really, really relevant.

So part of my coaching, part of my education, part of

what I do with boards

and companies, it's actually

to teach them critical thinking.

That is a one thing we cannot displace in the age of AI

because the goal is to outsource

tasks, not critical thinking.

So we continue to stay relevant

and pick up on the nuances that AI cannot yet.

Now if we talk about the macro view, there's a lot

of investments going on.

You can't turn the corner without this unicorn

company, this VC.

And yeah, the numbers are about $1.85 trillion

by the end of 2030.

So fun fact and kind of a sad fact.

Show of hands, how much of that do you think is needed

to solve world hunger?

30%? 50%?

Okay. 70%?

Okay. It's just above. It's just below 50%.

We are investing more in this capability than we are in the

largest humanitarian effort our world has ever seen.

That's how important this is to us.

So if you're avoiding it, if you're overwhelmed by it

or if you feel that the team is actively rushing into these

decisions, just understand AI is here.

It's not gonna be a two year trend,

but I would prefer it happens with us rather than to us.

And so we're gonna walk through some frameworks,

we're gonna walk through some ideas so

that you all are educated when you're pulled

into those conversations.

Because the fact is, is we are living in the age

of Moore's law and it's only gonna get faster.

Okay? The state of data

and AI today, this is one of my favorites.

This is what we hear in reality.

This is kind of where we are when it

comes to deploying at scale.

And it supports some of the statistics that we saw earlier.

And I only share this

because I want everyone to not feel FOMO.

I want everyone to thoughtfully move forward regardless

of whatever initiatives you have with data

and AI within your organization.

'cause these things take time.

And so we do hear a lot of stuff out in the market,

but the reality is most

of us are still figuring it together.

Like figuring it out. When it comes to AI,

it's not nearly as sophisticated.

Granted the hyperscalers and the startups, yes,

because they are either AI native

or technically native, they can fundamentally move faster.

Fintech, proptech, adtech, yes.

'cause tech native at its core,

fortune 1000s, not so much.

There's a lot of adjustments that need to be made.

And then when it comes to data ecosystems to be able

to leverage your data, well yeah, it sounds easy.

How on earth do you happen to do it?

They created an entire role called the Chief Data Officer

because we haven't been able to wrangle

or develop a comprehensive data ecosystem in decades.

And so we are stitching things together

as this image indicates.

So why, let's get into the lessons learned.

We are seeing paces of innovation we've never seen before.

The most common household item, like even in the shanties

of Brazil, the communities will have a refrigerator

that took nearly 50 years for there

to be full scale adoption.

Fast forward, fast forward. Notice there's a trend.

Took 25, 24, 10. It's about four to five years.

And we're about two years into the journey.

What we know right now is gonna look

very different in about a year.

All right? In addition to that, it's really hard to keep

up with the pace of change.

How many of you guys out of curiosity are dealing

with an AI initiative or a data initiative?

Show of hands. Okay.

And of those strategic priorities,

keep your hands up please, if you wouldn't mind.

What other strategic priority dropped?

Keep your hands up if nothing dropped,

put your hands down if another strategic priority was

deprioritized to make room for data and AI.

All right, so we're about 50/50

'cause we're really good at just adding stuff to our plate.

But you have to create space and room and time.

'cause this isn't a, we'll just do AI. This isn't a tool.

This isn't buying 400 licenses of copilot.

And there is a difference. And we're gonna go

through the difference between using AI or and and doing AI.

But it is not just a strategic priority,

it's a fundamental operational imperative

that we quite frankly is gonna require

our attention and resources.

And I'm not sure everyone's getting it yet,

which is why we're seeing those numbers.

And in addition to that,

the A in artificial is a little bit misleading.

There's nothing artificial about it.

And I think it's creating unrealistic expectations.

It's not a light switch you turn on, you just work.

I'm not buying a $20 ChatGPT

or perplexity license and there I go.

I can start writing my prompts when you're doing it at scale

because you've actually need it to be production ready

'cause it's external facing

or even internal facing thousands of employees.

It's a very, very different beast.

So I say most of the capabilities

that are succeeding right now in the AI space

are either automated intelligence, so automation,

augmented intelligence, so massive knowledge repositories

for credit scores, customer records,

warranty service, whatever it may be.

And anticipatory.

So a lot of your predictive models, that's where the center

of gravity is right now.

And I have found, and it took me nine years of mistakes

that every time I used artificial intelligence in front

of a board, they're like, oh,

there were just expectations it

was gonna be done in six months.

But when I replaced it with one of these,

for some reason there was a a more

of a normalization in level setting.

Alright? Now the crazy thing is,

is we are also advancing into autonomous intelligence.

And this is where agents kick in.

That is a very different beast,

but I won't overwhelm you guys.

Let's talk a little bit about data.

With that volume, variety

and velocity of data that's being generated

and information being handed to us,

it has gone exponentially more difficult

to build a cohesive data ecosystem.

10 years ago, I knew what the modernized data stack was.

I even had an answer five years ago.

There is actually no modernized data

stack at this point in time.

It's kind of like going to the Cheesecake Factory

and there's too many options in the window

and you always get buyer's remorse no matter what you order.

There's just a lot of options

and they're all amazing at what they do.

And some of them are enterprise grade, some

of them are amazing capabilities from startups.

But that consistency, that standard has been dissipated

and it's actually made the job of whether it's IT,

the data team, whatever it may be,

anyone who's in the technical role that's fingers

to keyboards and responsible for development,

it's made it a lot more difficult.

And the reasons why, culture

we're just slower to move.

I worked for the music industry

and even though nearly everything was on Spotify

or Apple, a lot of their business decisions were made off

of experience and intuition and gut.

Their marketing strategies were a little bit spray and pray.

It wasn't as data-driven as they had thought.

Culture really matters. Fragmentation too many sources.

And then the pace of change is too fast.

Because by the time you make a decision,

can you make a decision usually in six weeks?

No. You gotta have alignment discussions,

you gotta get everyone on board.

You gotta go through procurement.

It takes three to six months

and by the time you've made the decision

you're ready for deployment.

The architecture's changed.

We cannot sit to make decisions as fast as we used to.

We fundamentally, we use the word agile,

whether it's the methodology

or an adjective, I think it's an understatement, a gross,

gross understatement.

Because the reality is, and I don't mean to scare you,

but this is one of my architectures for one

of my fortune fifties that I worked for, we had

to map out every data source to understand how

to build the migration model.

And this is how my business executive thought the data was.

There is a massive difference, right?

Between the reality of our ecosystems

and then the perception.

Well I don't understand. How hard is it?

Well, because you want to build a dashboard

with net sales, that sounds easy.

Do you know we have 11 definitions of net sales

and I only know that because I had to go through this.

Well, why do we have 11 definitions?

Because the regions could not decide on net sales.

And sometimes some channels are included

and sometimes some channels aren't included.

So we will build this and I could build it for you.

Honestly, it's not that difficult in less than a month.

But which net sales do you want? Well, I want this one.

Okay, but your analyst is using this one,

the data's not gonna reconcile.

And you're gonna assume that

what I built was wrong when in actuality it wasn't wrong.

But you're used to getting PDFs

and Excel spreadsheet from a data analyst

who picked the wrong data source.

So now we've gotta reconcile all of your data

and you can understand how that happens at multiple layers.

Alright, some good tools.

Don't use AI for the sake of AI.

If you're trying to create a competitive advantage and

or fix a problem, there's a lot of advanced tech

that quite frankly does the job.

Sure, put AI-powered on

there and make your executives happy.

It doesn't matter. But the fact is, a lot

of things work really, really well

that you don't necessarily have to experiment with.

Because I always say, if you're gonna cut a piece

of paper, a scissor does a trick.

Why on earth would you introduce a chainsaw into the mix?

And that is the power of AI.

So figure out what that competitive advantage is, figure out

what that problem is that you're really trying to solve.

And if something less complex than AI

can actually solve the problem for you.

Another lesson learned is once you've gone

through the process of yes, we could use AI

to solve this problem and only AI can solve the

problem, you're still not done.

This is a little bit of a secret.

After about my first just under 30 deployments,

I was like, what is happening?

Why are we not getting the results we want?

I created this framework that I now use

to help a lot of my customers.

If AI is the answer, it still doesn't mean use the AI.

Because depending on your use case

or the problem, depending on the complexity to deploy,

and you have to assess your data, your infrastructure

and your talent, and the risk exposure,

external internal fines, regulations,

some things could be AI only.

Some things have to be human only.

But where most people get tripped up is this quadrant.

Some of your use cases should be AI led,

but absolutely need to be human supported.

And there are others that need to be human led

and AI supported.

And that differentiation actually makes a huge difference.

So there's a little bit of a secret

that you guys can use in future meetings.

There is a difference in using AI.

In doing AI and it's mostly the level

of effort you're using.

AI ChatGPT, Perplexity, copilot licenses. How about it?

Go summarize an email.

But doing AI is fundamentally decomposing the

tasks of a workflow.

Figuring out where a machine can automate,

where humans need to decide.

That level of effort is a completely different beast.

And we're confusing the two, which is why a lot

of times there's a delta between the expectations of

how quickly something should happen and the reality of it.

Business value gets tricky with AI.

The challenge with business value is

how is it usually created?

A lot of assumptions. Excel spreadsheets.

I had to stop using business value as the primary marker

of which AI use case to prioritize.

And so this framework, you measure the criticality

of the use case against the complexity

and the one that has the highest criticality

and the lowest complexity for you

to deploy based on your current

environment, talent, and data.

That's the one you go with.

You're not running a marathon if quite frankly,

you just learn to walk after wearing crutches.

You gotta think big, but start small

and then you decide how to scale quickly.

All right, A few more things.

The reality of where we are today,

and we're gonna go through these a little bit quicker.

AI has happened, it is happening

and there has been a pivot in our perspective.

27% of Americans are now using AI.

There's a double-edged sword. This is good and this is bad.

The bad part is I think we're starting

to outsource our thinking versus just outsourcing our tasks.

The good is those that understand

how to actually leverage it.

We're being able to amplify our creativity and ingenuity.

But the amazing thing about it,

in case you guys didn't know, it took Instagram two years

to build a consumer base of a hundred million individuals.

It took OpenAI two months

and right now 2.5 billion prompts a day.

And most of it comes from free subscribers,

which is why they're raising the gajillions

of dollars to keep running.

Oh, and here's a little fun fact.

How much energy does one prompt take?

'cause no one's talking about the energy shortage.

One prompt is equivalent to recycling 47 water bottles.

That's why we're buying up data centers,

nuclear power plants.

Energy is gonna be the next efficiency game

'cause we can't support the scale so far.

Data surveillance. How many of you guys are

around during GDPR?

Oh yeah, I call it gosh darn pain in my rear. Yeah.

Oh that took up so many years of my life. And for what?

Nothing. Because back then consumers had the right to opt in

for their data to be used.

Now the default is your data's always being used

and you have to actively opt out.

Two years ago, Allie Miller, Cassie and myself

and a few of us in this space, we posted, we had no idea

that LinkedIn was training

its foundational models and our likes, our comments,

our stories, our clicks, everything you were doing,

there's no notification.

I didn't opt into that. I didn't download the app

and click the toggle button said yes.

You now have to proactively opt out.

Which is why anytime you get the orange circle on your phone

or on your watch, it means you've given one

of your app's permissions

to track your voice, track your texts.

So I would go through in your settings and opt out.

That has to be active.

Alright, so how does this all come together

and what does a data healthy data ecosystem look like?

So you can leverage data as a competitive advantage.

'cause now we're talking about how the inner workings work.

It's not glamorous, it's not sexy,

but know these terms so

that when people are making the suggestions,

you don't automatically dismiss 'em.

The first is data classification.

Our old data security and governance methodologies cannot

and do not apply to how we're operating business today.

Putting policies and procedures on a SharePoint drive

and hoping everyone pays attention is just not the way

we're gonna do business anymore.

The first step is you actually have to go through

and classify your data.

How much do you have? How much is sensitive?

How much is confidential? How much is dark data?

And you can't do it manually.

I've tried as a CDO, you actually have to use AI,

which is the predominant agitator

and use AI native capabilities

to auto classify your information that way.

Anytime you wanna use it or anytime something leaks

or you're in risk mitigation, you know what data, where,

who, and how classification is a non-negotiable

second data quality.

Fix it at the business process

of actually is creating it all.

Too often we try and fix it at the end.

You've gotta fix it at the source risk tolerance.

You don't need to be perfect, but on the spectrum of

use it or tighten it, where are you?

And make your decisions accordingly.

But you gotta get some sort of consensus

and normalized opinion within the organization.

Otherwise you're always gonna have the struggle.

Not all data's created equal. You have to tier your data.

Which ones are the ones that we are gonna ringfence protect

and make sure and ensure that there's the highest quality

because there's regulations and fine

and consumers expect it.

Put your energy and resources towards that.

And then the last, we are great at collecting data.

We are not very good at connecting the data.

We are data hoarders, so there's no shortage of information,

but how we can use it to create a competitive advantage,

that's the misstep because we have not spent the time

to connect the information in a meaningful way.

And that is an activity that you must, if you wanna be able

to leverage it, it's not a step that you can skip anymore.

And this is kind of going back to tiering your data.

Not all data is created equal, so you have to pick the ones

that are important and you have to make it accessible.

I have battles with my compliance, legal

and cyber team all the time

because as a chief data officer, chief AI officer,

I was hired to leverage this information.

But every time I have to go back

and ask dad for permission

to use the data like that won't work.

Not in this pace. The CISO,

CDO, compliance

and legal have to come together into this holy trinity

and decide what is

and is not usable so

that not everything is permission based,

otherwise you can't go as fast as you fundamentally need to.

Okay, so with that said, I know we went a few minutes over.

I hope you enjoyed this presentation

and took some nuggets and frameworks.

I saw some snapshots. And thank you for your time.