Introduction: From Meetups to a Communication Challenge in Tech

I came to this country in 2014.

I did my master's at the University of Guelph.

And then I used to attend a lot of meetups like this.

The Real Barrier: Communicating Your Value

But later I realized that, you know, people who are working in AI, tech, data,

to get into an industry or to get into a bank or anywhere like that,

And the real challenge is not the skill sets.

The real challenge is, like when you go into an interview,

we can't actually present ourselves that well.

So the challenge for the tech people is a non -tech problem.

It's not a tech problem.

It's not a skill problem.

1The problem is we are not being able to communicate to the hiring manager.

In a way, they're trying to find out the right candidate.

Why Presentation Skills Matter for AI/Data Professionals

So presentation, meaning having this connection, getting the message across, is a big challenge.

Why that's a problem?

The Talking Gap in Tech

Because in general, many of the tech folks, they don't like to talk.

But if you don't talk, then it's difficult to get, you know, really communicate what

you need to communicate so i'll come to that point later but let's go to the next person

if you don't mind them okay okay okay interesting you can talk about it anybody else from this side

please go ahead sounds good anybody from hr here not yet not yet okay all right

if you don't mind yes but why are you interested about ai what brings you here i mean it's not a

crime for sure good good very good okay let me go to the like whoever any volunteer please go ahead

fascinating thank you thanks for coming tonight who else please go ahead

Participant Introductions and Perspectives

AI’s Rapid Impact on Specialized Careers

Yeah, my name's Mike, not Dave, but I just came back to Spain and I'm one of the first victims of AI fireworks.

I'm a technical translator.

I've contracted with four in Spain and the progress is so fast in my technical translation

that I'm one of the first to do this, so I'm going to pivot to another field.

but things are happening very fast,

so I'm just used to people who are following benchmarks,

because things are happening month by month,

and people have no main knowledge in IT,

the whole set of stuff.

The things are happening much faster

than most people are aware of.

Exactly.

Absolutely.

Thank you for coming.

Anybody else?

Please go ahead.

Using AI in the Music Industry

My name is Tyrell.

I worked in the music industry as a software engineer.

Music?

Fascinating.

Okay.

I was a lead engineer on the Sounds Unite platform,

so we built tools for musicians to...

How do you use AI in music, though?

So we actually used a service called Luzio

to label music

and then populate our Elasticsearch indices.

So you could search for things

that were a little more interesting,

like BPM or the key of the song.

You could have the feel of the song or genre.

Or you could search, you could take one song and say find songs like this one, which is more how we think when it comes to music, right?

And yeah, I came here because I'm interested in it, and I'm also looking forward to it.

Very interesting.

One of my friends was using ChatGPT to actually generate music.

Basically what he does, first he generates some poems, and then he generates some sort of music and uses it.

but that's you know but but some people are yeah yeah amazing amazing who else

Data Residency and Compliance in Banking

fascinating bunch of people tonight go ahead whoever yeah actually that reminds

me of a time when I used to work for one of the banks over here, one of the big

fives, and then the issue was Europe doesn't allow data to to go beyond its

border.

And then I was working with the compliance team and my team is

responsible for developing, was responsible for developing models that

actually can prevent financial crimes.

So we're developing prototypes that were

being able to actually predict what's coming up next so back in those days say 15 years ago 20

years ago the models were all about forecasting so what's going to be the price two years down

the line five years down the line what's going to be the price of the share or what's going to

be the price of this specific product how my you know products are going to behave in five years

10 years, 20 years down the line.

It was all about forecasting, but now it's a bit different.

How Banks Use AI Today

Different in a sense, AI models where banks are now have, they have become kind of an IT firm

because they are building, developing their own models.

They are having their own IT hubs,

and they're coming up with their own LLM models.

Also, they're working on, you know,

agent AI, Gem AI and whatnot.

So in a sense some of the banks are so big that

you know none of the companies are enough to actually develop one unique

product for them.

Meaning it has to be customized for every different bank.

So

just to give you an idea, I work for five different or six different banks so far.

I started with Standard Chartered Bank, then moved to Royal Bank of Canada, moved to Bank

of Montreal, Scotiabank, HSBC Canada, and then now I'm in Citibank.

In a layman term, bank has basically four pillars.

Banking 101: The Four Pillars

Number one is front line, business, like those who generate the business.

Second is basically the risk.

So many people think the bank is just making money, no.

No.

The core of banking business is basically they are into a business of managing risk.

Say you go to a bank and say, hey, I need a mortgage.

What do they do?

Give me documents.

I want to understand you as a customer, your capabilities, your assets,

the other loans that you have.

They send these documents to a, what do you call it, credit analyst.

they look into it they try to understand that if you would be able to pay back

that loan in time so they were actually calculating the risk associated with

this process and they try to calculate a risk that they are willing to take so

they're in a business of managing the risk so that's called the risk team the

third pillar of a bank is called operations so you go to a bank you open

an account okay somebody open an account the account manager open it for you but there are

a lot of works behind the scene somebody has to print the checks somebody has to print the letter

to send it to you somebody has to maintain you know everything that you do so those are

operations you know many people know um that like if you draw you go to a branch you know give them

a check they process it so a lot of operations around processing and in the

fourth pillar a bank has is called functions what are functions HR is a

function finance is a function marketing is a function like we are audit audit is

a function and then everything else that you see are basically functions so how

how many pillars?

Four.

Early AI Use Cases: Customer Service and Beyond

When we started integrating AI tools into our systems, initially it was

all about how do I provide the best customer service ever.

So that's why probably the first

AI tool banks started using widely is basically in the call center, the customer service center.

right now if you call a customer care center the first person that received

the call is an is a machine basically it's an it's an AI tool and then it

actually decides where to send this call where to forward this call to so that

was initially one of the most popular AI tool banks started using but now not

only just frontline meaning those who are who are into business when I say

a business bank has different businesses you know primarily opening accounts you know savings

checking all the sort of stops and then loans are different and also banks are into capital markets

hedge fund fx whatnot and there are many other products so right now every single product

Every single thing that we try to do, we have been trying to automate it.

From PhDs to Models: Pricing and Analytics Automation

Every single piece of it, like pricing, previously, 10 years ago, 15 years ago, we used to hire

a lot of PhD holders to decide what should be the price of a product.

They used to calculate.

They used to calculate a lot of stuff, but now everything we're trying to do through

using the AI models.

models.

So AI models are being used in pricing.

Enterprise-Wide Adoption: HR, Marketing, Finance

AI models are being used in detecting customer

behavior and forecasting.

AI models are being used even in HR.

Let me give you an example.

So after the COVID, what happened is a lot of banks actually let some of their floors

floor scale meaning um so just picking up a name say rbc had 10 floors in 20 king street

so after all you know after kobe they decided okay i don't need 10 floors i need only five

how did they decide it so they were looking into the can i employ behavior how many employees are

there at the office uh on a given day they saw that at least 10 percent of employees are not

not coming for whatever reason they're sick they're absent they're taking a leave so they

decided that okay let people come to the office three days a week so we're saving two days

and then they calculated the you know given the pattern of people's behavior of coming to the

workplace in person this is the those are the number of desks that we need this is the number

of floors that I need, this is the number of facilities do I need.

So every single

decision we're taking these days are driven by models.

HR is using the models,

data is using the models, pricing team is using the models, customer care team is

using the models, finance team is using the models, even the marketing team is using

the models.

Everyone.

Everything you can think of.

Limits and False Positives: Why Human Oversight Remains Key

Now I work in audit and even

audit is being automated you think of it I mean audit is something we had a hard

belief that it cannot be automated no no way because models makes a lot of

mistakes so I used to work for a team that used to detect sanctions so every

Every single transaction that happens in the bank, it goes to a scanner, and then the model

detects it.

So initially, the number of false positives was like 98 out of 100, then it started reducing.

Even nowadays, most of the models, there are lots of false positives.

I mean, if there's, if the false positives, we can bring it down to 2%, we're super, super happy.

It can never be 100%.

So, you know, every single thing, what I wanted to say is every single thing is being, trying to, they're trying to make it automate.

Previously, we used to use different techniques, tools, but now AI tools are kind of taking over everything.

but then again it cannot take over everything because it also does mistake

Machine-and-Mind Collaboration

so what do we do so we're using a term called machine and mind collaboration so

I think 20 years down the line in the future this is going to be the I mean

not in 20 years time I mean even earlier it might be the buzzword machine cannot

work alone machine cannot take over everything there has to be a mind

somewhere behind it I give you an example like say you do a transaction of

$10 ,000 every month for whatever reason one money it probably a little bit

higher say 12 ,000 the machine will be saying hey this is an unusual

pattern in there is that a fraud probably not a human mind can actually

actually judge a little bit if they can understand okay what could go different but a machine to a

What Works Well vs. What Needs Improvement

machine it's all about logic but anyway so bank is using AI what's going well as I say customer

care so far going super well marketing strategy going super well pricing going super well it's

working very well what's not working well there are some cases bank is still

trying to use LLM large language models but it's not there yet meaning say I

have to write write a document and say I have to write an audit report so I'm

trying to use LLM but you know every time I use a preset or about every time

I use what do you call it I mean it's giving me different result but it might

not be the way I like so I have to refine it again and again and again and

again so there are some areas machine is not there yet still we have to do a lot

of false positive checks this and that but gradually I think the use of AI

would be way more well now sorry I didn't get to know you you were raising

your hand for yeah go ahead okay if I forget it remind me again if you don't

mind I'll come to that point sorry I'd love to come to you as well yeah go

Go ahead.

Fantastic.

Okay.

Networking and Career Paths into Banking

Anybody else interested to introduce themselves?

Yeah, go ahead, please.

Wow.

Good, good.

Very good.

Thank you for coming.

Anyone else?

Please go ahead.

Hi, I'm Camilo.

I work for a company in the US, like around 11 ,000 people.

I am the AI team lead

that company.

I also have a startup in

South America.

Wow.

And I think

what brought me here was just curiosity

and working a little bit and just being

on top of this set of pre -arts

because in this type of space

this is where we find

the really good opportunities that

are in the oven that

maybe in five years will be a huge

company.

So I guess it's

good to be around this space

for that reason.

Yeah.

Banks Are Becoming Tech Companies

It's very good to see that a lot of people

are here tonight for networking but

When I go to a networking event, first question,

generally people ask these questions all the time,

that, okay, a bank is a financial institution.

What does, like, I'm a tech person.

I work in tech, I work in data, I work in AI.

How can I get into a bank?

What can I do for a bank?

As I was saying, I mean, so the information is key here.

I mean bank is not all about like dealing financials there are a lot of things uh that

has to be done at the back end and that's where we need a lot of lot of tech expertise I was

telling earlier that the banks are becoming tech company these days because they every single tool

they make they're kind of making it in -house most of them so they take something from one of the

third parties but then they modify it they make something of their own so if you're a tech person

Roles in Demand: Coders, Analysts, Modelers, and SMEs

If you're a coder, you know, there are a lot of coders working in each single bank.

Everywhere.

Data analyst.

Well, you might be a data analyst working for many other companies.

Bank use data analyst like anything.

Meaning every single department has their own data analytics team.

Every single team.

All of them.

So you work for HR.

HR has an analytics team marketing has an analytics team so I mean if you are a

modeler so I had a friend who did his PhD in maths mathematics it was like

okay I'm a mathematician I want to I mean what could I do in Canada should I

try to you know get into a college saying hey I want to be a college

professor I want to teach somewhere else what else could I do I was like you can

work in a bank like how because your expertise is in mathematics and you are

you work in modeling see could you work in the pricing teams like the product

pricing teams how because they use a lot of lot of calculations they used to do

it manually and now the mathematicians are and statisticians are actually

helping the AI modelers in there to build up their models.

Because when you

generate a model, you need a lot of variables.

A coder is a coder, but he

needs to know what variables you need to use for your model.

And for those,

they need SMEs, experts.

Those experts are coming from many, many different

industries.

Coming from tech, maths, you know, even language.

Even language.

Like

Non-Technical Skills Matter: Writing and Communication

somebody has to write a document you know you have to write it in a way

something that goes out to public you have to write it in a way it's easily

understandable so they are you know banks are hiring people who are really

good in writing so if you're into tech AI data analytics look for you know

opportunities like we often don't pay attention to the details in the job

description and the reason behind is oftentimes these JDs are kind of good to

have I mean if I have all the skills that has been mentioned in the JD it'd

Why Networking Beats Reading Job Descriptions

be a dream job for you know anyone most important part is this networking why

Why?

What a role needs in terms of skill sets.

What are the requirements?

You have to talk to people.

Go talk to people.

Go to the networking session.

Go talk to them.

What do you need for this role?

What type of roles do you have?

How can I help you?

So that's where you actually get a lot of information.

That's why networking is super important.

Q&A: Adoption, Training, and Risk in AI

So with that, I would like to open the floor for questions

because I know it kind of took more than what I was allocated for in terms of time.

Go ahead, please.

What are your outputs from that?

Human Learning vs. AI Tools in Organizations

So you're asking how do we retain information?

I mean, learning in an organization, often it's successful organizations are very skilled

and the boss talking to the floor manager is a very fluid and transparent and fast knowledge culture.

Correct.

And the new capabilities, we have a new feedback cycle with AI.

How will it affect your organization?

Yeah, I mean, AI is nothing but a tool.

So we are using it for designing our trainings.

So it's being used for designing the trainings,

designing the you know how to how to educate or train the stuffs but as I say

it's not there yet it's not self -sufficient yet it's like any other

tool you might have like you are having a calculator you know how to calculate

but the calculator has limitations so the AI tools that we have right now it

has limitations it cannot it's not there yet to replace the human to human

communication or transfer of knowledge from human to human so that's why in

person classrooms even in any organization in -person trainings are

still super super super important they're investing a lot of money for

in -person trainings compared to like on zoom or teams or you know virtually why

because you know anyone who is sitting in an in -person session they're paying

way more attention compared to when they're on zoom because when you are having a training

on zoom they're you know they're going to they're doing a lot of stuff together they're

doing multitasking so their attention is being divided which is impacting the quality of

work and the quality of training so person to person what you were asking before if i

understood correctly the person -to -person knowledge transfer is still irreplaceable

it's still super super needed so where ai is helping is helping me to design the training

in a different way it's helping me to gather information it helping me to prepare the slides

it helped me to prepare a speech but the speech that i have to deliver that i have to deliver

i cannot do it even if it does it's not as appealing as talking to a human being

Driving Adoption in Sensitive Sectors

so human touch is still irreplaceable as of now go ahead i would like to hear your opinion on

like how to increase the ai adoption in like more conservative fields such as banks or health

because it's not the same mislabeling an email versus accepting a fraudulent transaction so i

I know that there's like a different gap in those fields,

but what can be done to increase that adoption in your opinion?

Since you work in a bank, you have seen how it has evolved over time.

So your question is how the adoption could be increased?

Yeah, in those specific fields where AI can be more sensitive with the outcome.

Risk Tiering and the Three Lines of Defense

Yeah, so previously if we have 10 alerts and we detect it manually, AI is actually kind

of levelling, the levelling is being used, I mean AI is being used for levelling it,

then at the end of the day it's like mind decision.

so in simple example say stamps are coming in through the emails so now what

the AI is doing it's leveling it hey this is a high -risk high -risk email be

cautious but then again there will be links and things like that you will

click on the links or not it's a mind decision so what we did basically the

leveling has been I mean AI has been used to level everything and then there

There are three lines of defense.

We call it first line of defense.

There are firewalls and things like that, scanners being used as first line of defense.

And then the second line of defense.

The third is basically, in compliance with the first line is the people who are facing

the client.

The second line is the compliance team who look after the policies, what could be done

better, what could be improved, things like that.

third line of defense is the audit they look into if the policies are working

well or not so these three lines of defense model is everywhere so even in

terms of filtering out suspicious emails spams

Molleber whatever you know so AI being AI is using being used for first first

filters but at the end of the day it's coming to you anyway some emails will be

be coming to you but then again it's a mind decision whether how consciously are you going

to look into that email you might make mistakes but you know so automation is happening um more

and more um so it saves time it saves money where the you know any corporate is interested in

and anyhow if i can save time and save money corporate will be investing in it and that's

That's what's happening.

So they're investing heavily into automation, if that makes sense.

Upskilling for the AI Era

But then how do we adopt with the situation?

So we need to be skilled.

So there was a time when you used a typewriter.

Those days are gone.

Those who knows typewriter, good, but if you haven't upgraded your skill into computer

literacy or something else, you'll be out of the market.

So what do we have to do?

We need to upgrade our skills with time.

Now the AI tools are coming up.

Training is easy.

It's not so difficult.

But it's up to us if we want to kind of adopt that as quick as possible or not.

So try to master at least one or two, three AI tools, which is super easy.

You know, and that will actually help you a lot.

And what I personally believe, it's always important to, you know,

have an eye on what's happening in the market.

what's going on what new technologies are coming up how companies are changing

themselves what can be the need of the next 10 years or 5 years to 20 years and

adopting you know investing in new skills all is paid off for sure

Regulation, Security, and Responsible AI Adoption in Banks

What is the adoption of AI and how banks are putting new religions to adopt AI without risking, like you've got a bank being a sensitive institution, you cannot have AI for your business.

security first I have to think about what regulators talk about it what are

the regulations around it if there's no regulations around it I'll be careful I

mean give you an example if you remember Canada legalized marijuana at some point

we prior to that it was illegal so bank would never accept any transaction

related to marijuana prior to legalization after legalization you know

it has started receiving those transactions accepting those transactions right now if i

kind of like prior to that you know we were developing at some point we were developing a

model that would help me to detect um a legal marijuana transaction versus an illegal marijuana

transaction how does that happen so some places marijuana transactions are illegal like some in

some states of the US in like illegal in Canada it's legal so we all have to have

some sort of assistance from AI where is it legal where is it illegal so this is

where sometimes often mind comes to play you know a model generate triggers for

everything somebody has to pay attention okay is this coming from a legal

jurisdiction or non -legal jurisdictions because they are there are false

positives so they have to pay attention to the false positive in the

kind of investigate okay this is a true true match and a false match so I mean

that's why the banks are slow because they have to go with the regulator space

if the regulators are quick they're quick because banks always they kind of

have to whatever they do they have to be within the regulatory expectations so

that's why at the beginning they were very conservative about you

cryptocurrency why because there were no regulations at the beginning so as the regulations came in

they gradually okay you know incorporated everything within their system but before that

they happen so most of the banks are kind of you know adopting ai but at a slow pace because of all

those other factors compliance regulations risk data they have to ensure the data security first

of all that's the you know top -notch priority like I mean this is something

non -negotiable at any point of time so I mean chat GPT was introduced a couple of

years ago banks haven't adopted it right away why they were looking into it okay

is my data secure with this AI tool when they became you know kind of okay it is

it is secure I can use it within my own organization they started adopting it

But prior to that, they haven't.

So they took time to better understand

how is it going to comply with my data security,

my compliance, my regulations, and everything else.

Does that answer your question?

Okay.

No problem.

Closing Remarks

Thank you so much.

Anyone else?

If not, I'd really like to thank you for coming tonight.

I'm really happy that you were all patient and listened to me.

Thanks for listening to me.

Thank you so much.

Appreciate it.