I am convinced that banking as the
we have known in the last 40 years
more or less is about to become extinct
technological innovations are
dramatically changing the way
in which clients are managed and in
that the risks are managed and that is the
the very essence of the banking business that
involves challenges
very novel unsuspected some for
the supervisors but the truth is
that the digital age and developments
based on artificial intelligence
oblige entities to adapt to
these new circumstances or what
otherwise possibly disappear in
the medium term
of artificial intelligence is coming
speaking since the decade of
fiftys
in man there seems to be a kind of
obsession to develop artifacts and
instruments capable of replicating what
does the human being but obviously with
higher speed with greater accuracy
greater precision with better results
with a smaller margin of error remember
that robotics is considered one of
the branches of artificial intelligence
that’s why in the industry of
manufacturing we find many examples
very interesting for a long time
time factories fully robotized
that assemble and make vehicles
same processors that carry the
computers inside
medical diagnoses most recently
there are already robots that make sutures in
in humans until you reach the
3d printers that javier mentioned
able to print bridges buildings
integers printing each one of its
components that are then assembled
but man always dreams of more
always want to create the machine that
I get to learn by myself almost to
think like him and if he could
possibly create some robot that
have feelings
that’s why we speak today
we hear a lot of talk today about what
has come to call machine learning
machines capable of learning by themselves
alone sounds like something from the jets 11
in 1962 but the reality is that that ‘s
something that is already happening machine
learning is the conjunction of
computing and statistics for
prepare computers so
run algorithms so
iteratively and also from
interactions with humans
go enriching and selecting a
increasingly robust model a model
every time you better surely
remember a few years ago when
We called some company and we attended
an answering machine with a
totally flat answer if you
want x mark 0 if you want
mark 1 if you want z mark two
nowadays we can almost dialogue with the
machines that serve us in the call
Many companies’ centers make us
questions say your name
good the recognition of the classroom of
linguistic recognition is
considered also one of the great
milestones of artificial intelligence and it
has reached the point where its
algorithms are able to determine if
the customer is already being inconveniencing
losing patience no longer wants
keep talking to the robot in which case what
what does
the machine is that it transfers the call
to a human operator to finish the
customer relationship so that the
experience is not so difficult
The bench
so
it is a tremendously fertile field for
development of applications based on
artificial intelligence for what he said
cardigan if someone has information is
a bank or someone can have reach
have information is a bank think about
the issuer of the credit card or
the debit card that is with which
you count the issuer of that card
he practically lives in his pocket
pants inside your wallet is capable
to perceive
where you do your shopping every
how much are they going to al molt what kind of food
consumed where they like to vacation
imagine the capacity that has
issuer of credit cards for
determine what your tastes are
trends preferences that attracts them to
you and now think about the impact
that can have that in the marketing
now I want to clarify something this has a
huge impact on the marketing of
entities but also in the management of
risks
a few years ago when they did not buy by
internet some product simply
the cart appears the product was inserted
in the cart and you bought it and not
it was much more than that
now the situation has changed when
you buy on the internet some
product immediately begin to receive
a message look at the people who
they bought that product they also bought
this and they also bought this other or in
place of that product because it does not compare
these others what is behind this is
an algorithm
what are you comparing
studying their consumption habits
inclinations your tastes with those of
others
for the algorithm are similar to
you and then take advantage of that
information to try to increase
the sales
good and what about social networks is
possible to predict patterns of
behavior from the study of
social networks the answer is
affirmative there are already algorithms capable of
analyze what is called information not
structured that is the information that
can be extracted from social networks
imagine that information that
is extracted from the like that is given to
a photo to a comment or a
comment that you write in a
network like facebook or any of these
well there are already algorithms capable of
analyze that kind of information is not
unstructured and derive also
behavioral patterns from
that kind of information
all this subject of the great volumes
of data that is causing this huge
disruption is typically called
views and has already been mentioned by
some some of the exhibitors
and exactly what is sought is to do
mining over all that big data
analyzing structured information and not
structured to predict the
future behavior of people
the implications as he explained them
before they are not only from the optics
of marketing but they have a lot to
see also with the subject of the management of
risks
for decades bankers have tried to
predict behavior
debtors especially their willingness to pay
Many times a debtor can offer
some good guarantee show good
income but have little or no
willingness to pay how important it would be
then be able to predict in advance which
what is it that in risk management is
called the probability of
failure of each of the
debtors
historically it has been done based on
the past and then the nations
they receive a probability of compliance
the companies receive a rating that are a
a probability in compliance and the
people receive a scoring
that previously I said
historical but now tend to be
prospective and how they tend to be
prospective ie project what is
the most likely scenario
behavior of that good person
because they compare a lot of
variables that characterize that
individual compare it with others
individuals that the algorithm determines
that are similar and then they say this
individual looks like this to me that has
paid well and therefore the
probability of default is very
low that has a huge value
but when I started working in
developments of these algorithms these
scoring we live some experiences very
Interesting
and I remember well that the first time
in the national bank where I was
the last 13 years we developed the
first scoring that pretended to be
predictive of what the
probability of default
obviously there was some resistance to
field to change mostly jumped two
roosters
we have 20 years of experience
granting credits no system that
you put there will overcome
the feeling the experience the smell that
we have developed ourselves throughout
20 years old so when the sporting
say rejected if it seems to me what
approves the credit
what we decided was to make a deal see
for one year
I will control the approved portfolio
for you with score in rejected if at
end of the year the delinquency of the portfolio
that you approved with square in rejected
is relatively high then you
He abandons his nose and makes her case
scoring one year later the default of that
portfolio approved by them with the
scoring rejected was four times the
delinquency of the consumer portfolio of the
office the discussion is over
they can really be extracted
very valuable instruments from the
application of this type of instruments
there were not a few challenges that we had
to start working on this kind of
modeling that has a lot of application
of artificial intelligence about envelope
the on the beta is the first
obviously convince the general manager
from the time of making a large investment
in supercomputers
what was it that we needed to
process all that volume of information
that a bank of that size relatively
large in costa rica may may have
but the most difficult challenge without the least
doubt was to find people
the university of costa rica
graduated at that time three mathematicians
by year
and I told them look, actually I think
that graduating in the middle because there
two and a half what they want is
study some things that I can not
understand of development theory
very complex mathematicians and theorems and
landed who want to work in a
average bank per year that’s pretty much
what they produce
and if we still ask the mathematician
that he knows something about programming that has
become a necessity
to have true scientists from
data I tell you that the thing gets
still much more complicated to me
It seems paradoxical that a country that attracts
the amount of foreign investment
direct that attracts costa rica because
prepare very good engineers very good
technicians prepare so little
mathematicians prepare very few
econometricians and also not give them
that integral formation
adding statistical techniques and
programming techniques that that was the
biggest challenge we had but also
things get complicated when they are already
totally prepared and ready because
try to keep them
the competition immediately begins to
try to take them away if not
resign and put their own company of
sale of models to the other banks that
I also lived that experience
the result can be very valuable very
useful
and get all these banking topics
digital how can you see in this
video offered by a Spanish bank does not
there are Costa Rican banks that have
advanced a lot in the digital issue not
I wanted to put none of them to not
get into conflicts of interest like this
that what I bring to you is this video
prepared by a Spanish bank in which
show how it is already possible to have
practically all services
banking on the wrist in the watch that
they have
and that they carry and that they carry
22 very important challenges that can be
solve with artificial intelligence
the first one we talked about indebtedness
of the homes I think that gerardo
touched on the subject it talks about that people
is over indebted in costa rica bueno
what if we had an assistant
artificial intelligence capable of telling you
the person look and their income is
being overtaken by your expenses no longer
is contributing enough to your fund
of pension is no longer contributing what
enough to your investment fund
you should take action
another important issue development banking
I saw bank executives sit down
four hours with a small micro
businessman to try to work out a flow
projected from here because they do not have the
sufficient accounting systems
can give them to the bank and then take
the bank a successful decision will exist
an artificial intelligence assistant
able to solve those dilemmas
the answer is yes javier mentioned to
apple siri or vix pink that is the
samsung are called artificial
smart assistants good how about
one developed by a bank
I present erika
of the bank of america that I suppose for
that is named
erika is an intelligence assistant
artificial that lives inside the app of
bank of america
what I loved about erika is that it’s a
extraordinary example of machine
learning because it interacts with the
users and when one of them
surprised with a question that I did not have
programmed that I no longer expected is going to
human support seeks an answer and
when the client is totally
satisfied stores the question and the
answer in case they do it again
in the future in other words learn
of the interaction enriches itself
it gets stronger and stays better
ready to continue attending
customers in the future
as a supervisor I can tell you
the first thing I discovered now that
I went back to the superintendence is that
Colombia is a member of the supervisory group
regional which gave me the idea of
ask them what they were doing
on the subject of supervision of fintech
about which I really have no idea
I had an idea yet and with what I
they went out was what the sandboxes
is that mentioned Andalusia and the first thing
what I imagined was a cat doing the
fats thanks in the sandbox
but they explained to me later that
they call regulatory sandboxes and I think
brutal the idea of saying is about
test environments where the
banks and the shf intec to the
superintendence and in that environment of
proof they have the
sufficient cybersecurity protection
to the consumer enough and not until
they receive the approval of the
superintendence then we do not
allow to put it into production and
offer it to customers it turns out that the
people who are more interested
in entering the sandboxes as they told me
the Colombians are the same fintech
because once they pass by they have
the seal of approval of the
superintendence and that generates a
great trust to the users I close with
two comments in my opinion how they
it said at the beginning or the real banks
they really do digital really
they develop their own technologies
syntek or are associated with nascent
offer them I think they have their days
counted and the second as supervisor what
What can I tell you is that the
Supervisor must avoid innovation
It is very good to worry about health
of consumers is very good
worry about innovations
be favorable for those who
they use but the most important thing is never
prevent that innovation and rather
contrary to that, take advantage of it in the
as much as possible so that the same
supervision is more effective and more
efficient thank you very much
we have known in the last 40 years
more or less is about to become extinct
technological innovations are
dramatically changing the way
in which clients are managed and in
that the risks are managed and that is the
the very essence of the banking business that
involves challenges
very novel unsuspected some for
the supervisors but the truth is
that the digital age and developments
based on artificial intelligence
oblige entities to adapt to
these new circumstances or what
otherwise possibly disappear in
the medium term
of artificial intelligence is coming
speaking since the decade of
fiftys
in man there seems to be a kind of
obsession to develop artifacts and
instruments capable of replicating what
does the human being but obviously with
higher speed with greater accuracy
greater precision with better results
with a smaller margin of error remember
that robotics is considered one of
the branches of artificial intelligence
that’s why in the industry of
manufacturing we find many examples
very interesting for a long time
time factories fully robotized
that assemble and make vehicles
same processors that carry the
computers inside
medical diagnoses most recently
there are already robots that make sutures in
in humans until you reach the
3d printers that javier mentioned
able to print bridges buildings
integers printing each one of its
components that are then assembled
but man always dreams of more
always want to create the machine that
I get to learn by myself almost to
think like him and if he could
possibly create some robot that
have feelings
that’s why we speak today
we hear a lot of talk today about what
has come to call machine learning
machines capable of learning by themselves
alone sounds like something from the jets 11
in 1962 but the reality is that that ‘s
something that is already happening machine
learning is the conjunction of
computing and statistics for
prepare computers so
run algorithms so
iteratively and also from
interactions with humans
go enriching and selecting a
increasingly robust model a model
every time you better surely
remember a few years ago when
We called some company and we attended
an answering machine with a
totally flat answer if you
want x mark 0 if you want
mark 1 if you want z mark two
nowadays we can almost dialogue with the
machines that serve us in the call
Many companies’ centers make us
questions say your name
good the recognition of the classroom of
linguistic recognition is
considered also one of the great
milestones of artificial intelligence and it
has reached the point where its
algorithms are able to determine if
the customer is already being inconveniencing
losing patience no longer wants
keep talking to the robot in which case what
what does
the machine is that it transfers the call
to a human operator to finish the
customer relationship so that the
experience is not so difficult
The bench
so
it is a tremendously fertile field for
development of applications based on
artificial intelligence for what he said
cardigan if someone has information is
a bank or someone can have reach
have information is a bank think about
the issuer of the credit card or
the debit card that is with which
you count the issuer of that card
he practically lives in his pocket
pants inside your wallet is capable
to perceive
where you do your shopping every
how much are they going to al molt what kind of food
consumed where they like to vacation
imagine the capacity that has
issuer of credit cards for
determine what your tastes are
trends preferences that attracts them to
you and now think about the impact
that can have that in the marketing
now I want to clarify something this has a
huge impact on the marketing of
entities but also in the management of
risks
a few years ago when they did not buy by
internet some product simply
the cart appears the product was inserted
in the cart and you bought it and not
it was much more than that
now the situation has changed when
you buy on the internet some
product immediately begin to receive
a message look at the people who
they bought that product they also bought
this and they also bought this other or in
place of that product because it does not compare
these others what is behind this is
an algorithm
what are you comparing
studying their consumption habits
inclinations your tastes with those of
others
for the algorithm are similar to
you and then take advantage of that
information to try to increase
the sales
good and what about social networks is
possible to predict patterns of
behavior from the study of
social networks the answer is
affirmative there are already algorithms capable of
analyze what is called information not
structured that is the information that
can be extracted from social networks
imagine that information that
is extracted from the like that is given to
a photo to a comment or a
comment that you write in a
network like facebook or any of these
well there are already algorithms capable of
analyze that kind of information is not
unstructured and derive also
behavioral patterns from
that kind of information
all this subject of the great volumes
of data that is causing this huge
disruption is typically called
views and has already been mentioned by
some some of the exhibitors
and exactly what is sought is to do
mining over all that big data
analyzing structured information and not
structured to predict the
future behavior of people
the implications as he explained them
before they are not only from the optics
of marketing but they have a lot to
see also with the subject of the management of
risks
for decades bankers have tried to
predict behavior
debtors especially their willingness to pay
Many times a debtor can offer
some good guarantee show good
income but have little or no
willingness to pay how important it would be
then be able to predict in advance which
what is it that in risk management is
called the probability of
failure of each of the
debtors
historically it has been done based on
the past and then the nations
they receive a probability of compliance
the companies receive a rating that are a
a probability in compliance and the
people receive a scoring
that previously I said
historical but now tend to be
prospective and how they tend to be
prospective ie project what is
the most likely scenario
behavior of that good person
because they compare a lot of
variables that characterize that
individual compare it with others
individuals that the algorithm determines
that are similar and then they say this
individual looks like this to me that has
paid well and therefore the
probability of default is very
low that has a huge value
but when I started working in
developments of these algorithms these
scoring we live some experiences very
Interesting
and I remember well that the first time
in the national bank where I was
the last 13 years we developed the
first scoring that pretended to be
predictive of what the
probability of default
obviously there was some resistance to
field to change mostly jumped two
roosters
we have 20 years of experience
granting credits no system that
you put there will overcome
the feeling the experience the smell that
we have developed ourselves throughout
20 years old so when the sporting
say rejected if it seems to me what
approves the credit
what we decided was to make a deal see
for one year
I will control the approved portfolio
for you with score in rejected if at
end of the year the delinquency of the portfolio
that you approved with square in rejected
is relatively high then you
He abandons his nose and makes her case
scoring one year later the default of that
portfolio approved by them with the
scoring rejected was four times the
delinquency of the consumer portfolio of the
office the discussion is over
they can really be extracted
very valuable instruments from the
application of this type of instruments
there were not a few challenges that we had
to start working on this kind of
modeling that has a lot of application
of artificial intelligence about envelope
the on the beta is the first
obviously convince the general manager
from the time of making a large investment
in supercomputers
what was it that we needed to
process all that volume of information
that a bank of that size relatively
large in costa rica may may have
but the most difficult challenge without the least
doubt was to find people
the university of costa rica
graduated at that time three mathematicians
by year
and I told them look, actually I think
that graduating in the middle because there
two and a half what they want is
study some things that I can not
understand of development theory
very complex mathematicians and theorems and
landed who want to work in a
average bank per year that’s pretty much
what they produce
and if we still ask the mathematician
that he knows something about programming that has
become a necessity
to have true scientists from
data I tell you that the thing gets
still much more complicated to me
It seems paradoxical that a country that attracts
the amount of foreign investment
direct that attracts costa rica because
prepare very good engineers very good
technicians prepare so little
mathematicians prepare very few
econometricians and also not give them
that integral formation
adding statistical techniques and
programming techniques that that was the
biggest challenge we had but also
things get complicated when they are already
totally prepared and ready because
try to keep them
the competition immediately begins to
try to take them away if not
resign and put their own company of
sale of models to the other banks that
I also lived that experience
the result can be very valuable very
useful
and get all these banking topics
digital how can you see in this
video offered by a Spanish bank does not
there are Costa Rican banks that have
advanced a lot in the digital issue not
I wanted to put none of them to not
get into conflicts of interest like this
that what I bring to you is this video
prepared by a Spanish bank in which
show how it is already possible to have
practically all services
banking on the wrist in the watch that
they have
and that they carry and that they carry
22 very important challenges that can be
solve with artificial intelligence
the first one we talked about indebtedness
of the homes I think that gerardo
touched on the subject it talks about that people
is over indebted in costa rica bueno
what if we had an assistant
artificial intelligence capable of telling you
the person look and their income is
being overtaken by your expenses no longer
is contributing enough to your fund
of pension is no longer contributing what
enough to your investment fund
you should take action
another important issue development banking
I saw bank executives sit down
four hours with a small micro
businessman to try to work out a flow
projected from here because they do not have the
sufficient accounting systems
can give them to the bank and then take
the bank a successful decision will exist
an artificial intelligence assistant
able to solve those dilemmas
the answer is yes javier mentioned to
apple siri or vix pink that is the
samsung are called artificial
smart assistants good how about
one developed by a bank
I present erika
of the bank of america that I suppose for
that is named
erika is an intelligence assistant
artificial that lives inside the app of
bank of america
what I loved about erika is that it’s a
extraordinary example of machine
learning because it interacts with the
users and when one of them
surprised with a question that I did not have
programmed that I no longer expected is going to
human support seeks an answer and
when the client is totally
satisfied stores the question and the
answer in case they do it again
in the future in other words learn
of the interaction enriches itself
it gets stronger and stays better
ready to continue attending
customers in the future
as a supervisor I can tell you
the first thing I discovered now that
I went back to the superintendence is that
Colombia is a member of the supervisory group
regional which gave me the idea of
ask them what they were doing
on the subject of supervision of fintech
about which I really have no idea
I had an idea yet and with what I
they went out was what the sandboxes
is that mentioned Andalusia and the first thing
what I imagined was a cat doing the
fats thanks in the sandbox
but they explained to me later that
they call regulatory sandboxes and I think
brutal the idea of saying is about
test environments where the
banks and the shf intec to the
superintendence and in that environment of
proof they have the
sufficient cybersecurity protection
to the consumer enough and not until
they receive the approval of the
superintendence then we do not
allow to put it into production and
offer it to customers it turns out that the
people who are more interested
in entering the sandboxes as they told me
the Colombians are the same fintech
because once they pass by they have
the seal of approval of the
superintendence and that generates a
great trust to the users I close with
two comments in my opinion how they
it said at the beginning or the real banks
they really do digital really
they develop their own technologies
syntek or are associated with nascent
offer them I think they have their days
counted and the second as supervisor what
What can I tell you is that the
Supervisor must avoid innovation
It is very good to worry about health
of consumers is very good
worry about innovations
be favorable for those who
they use but the most important thing is never
prevent that innovation and rather
contrary to that, take advantage of it in the
as much as possible so that the same
supervision is more effective and more
efficient thank you very much
Please follow and like us: