AI in Banking and Insurance
Introduction
The upcoming technology whether is is not upcoming it is quiet an old technology but from a business relevance perspective it is a new form of enthusiasm for AI. What do you mean by AI ?.
There are three primary sensing technologies computer
vision which talks about seeing what is called as the input coming from images
or videos like we see with a naked eye then nlp or text mining is about input
text data and speech processing is about speech in but it is not that only these kind of inputs
are dealt with in core machine learning of
but it is also about any kind of data beyond these it can be structured
excel data or tabular data so once we
learn about the machine learning it actually talks about creating some king of
pattern from the raw data it takes raw data to be structured data or unstructured can be three types vision ,
nlp, a speech process so basically if look at
machine learning it is a brain of AI which is basically talking about
pattern extraction from raw data it could be structured tabular kind of data or
unstructured data is like images or video revision nlp or text data speech data
processing . So once we have this basic notion of AI lets look at what is machine learning ?.
Machine learning is basically considered as a way by which you take in data and
you kind of create a what we call as
effectively some kind of a pattern extract from the raw data . Pattern is also called as model and the beauty of
this machine learning is that it is purely a function of the data that is
given . The data which give to the
machine the pattern is extracted from the same data . For example lets say you
have created lots of input x-ray data
for cancer detection you label each of the cancer x- ray has having cancer or no cancer
then your machine learning programme should be able to extract the pattern from there so that when a new x-ray is given it
should be able to say whether the x-ray has cancer or no cancer so it is also important for us to understand that whenever we talk about extracting model or
extracting pattern from raw data it is quiet different from conventional programming . Programmic logic is
controlled with the program but in
machine learning we
have no control over the logic of
the model or the patterns in it . It is
purely data in program out or data in model so logic is hidden inside the data
which is extracted in form of the model or pattern so the only way the model
can be changed by changing the data so just before we want to go deeper into applications of AI in specifically
the vertical we talk about
banking and financial services . Machine learning itself is not like one
type fits all there are different kinds of machine learning in the simplest
form where we talk about machine
learning we have both input and output and what we find in the pattern or model is
that relation between input and output
data . That again can be two types the
output is something like you know
cancer where you have only two output
values cancer and no cancer so it is
fixed number of output value it is
called classification whereas if your
output is a continous value like you are predicting astock value
the it is called as regression so
supervised machine learning is the most
common kind of machine learning we work with
which is basically either classification
or regression so machine learning is not
like one size fits all so in some cases
you may have output variable and input
variable and if you have output variable
and you try to find relation between variable you will try to find out if it is a regression is a continous output variable or output is a fixed set of values which called as classification like the
cancer case one other case there could be scenario where you only have data and you are extracting some pattern
that is called unsupervised machine learning . There are several
examples of unsupervised machine learning where you could just extract patterns from data like it could be example like you form groups of
similar data called clusters or you could form paires of things like if you
have bought one thing you are
recommending another think like all of
you when you shp in Amazone . So there
are different examples of
extraction of different kind of
patterns and just raw data no output
only input or you are doing a
recommendation system when you are doing then browse your recommended something so that is a classic example of unsupervised
machine broadly mainly we have these two
kind of machine leaning supervised which is classification or regression
and also provides very just doing pattern extraction like clustering recommendation
systems or no extraction of rules .
AI broadly refers to the notion of human intelligence
being mimicked by machines that is why it is called Artificial Intelligence but
if you look at the over all picture of what you mean by the concept of mimiking
human intelligence there are broadly three aspects to this :
The first one is called Senses so like
human beings have senses like we see with our eyes , we listened with our ears
like that the AI technologies have sensing has a primary component whereby they
take in the data and the data could be coming from something which you are
seeing or something that you are
speaking , something which you are writing a text or it could be data coming
from any other data base or any source . So the sensing part is basically where
you take in the data , then the important part is about making sense of the
data which you got from the sensor meaning extracting patterns from the data
which you have sensed . Pattern is how you can extract a specific kind of what
we call as pattern which is a very very crucial for understanding the relationship
between different parts of the data or understanding how the pattern can be
understood so that it can be repeated when a new kind of data which comes . So
idea of core or brain of AI is primarily this part which is called as machine
learning which is basically about extracting patterns from raw data then
optionally several cases it is not just about analyzing the sensed data or the
data which is being sent and analysed for patterns but we also respond with an
action for example lets take a diverse car in that case we take a sensor which
is the computer vision camera which comes in analyses the patterns in the what
we saw in the camera then based on that we can take an action as to how we
should break which direction should we move or you can also do simply taking
the data in by sensor and then extract patterns from it then a human can take
action so it is not necessary that the autonomic response or automous response
by machine is mandatory so main code is
about taking in the data and analyzing the data patterns so that is what we
call as underlying technologies .
The way the company function for a period of time has been completely tranformed . Traditionally a manager taking the decisions most often was based on their experience . All of them transformed has completelely switched to a data driven decision making . We know that lots of data is processed and finding insights is called as big data analytics . The branch of study is data science . Now a days big data not used by company is very less also this is no longer a trend but a necessity . After 2008 recession financial institutions such as banks so many challenges they have gone through. In that point of time bank started using applications they started taking the advantage of these applications and try to find very different revenue streams to bring back their profit margin .
The Big data , Machine learning and AI how it transform the banking process through 6 use cases ?
1.
Customer
Analytics
When you compare other industries , bank has more data generated . There the big data process what are all the data is there we can once understand.
First one is KYC data Know Your Customer data : When we go to bank every time updated information we fill there we can see that is Know your customer data.
The second one is Banking visits : when we always visit bank that data points completely locked .
The third one is Credit Card History : What all we have transaction through credit card we can understand .
The fourth one is Web History : that means what all we are doing in web that all will be locked .
The next impotant is Call Log : that means when we call all of them is logging . You know that when you call the bank this call will be recorded the message we can see that is unstructured data is is very important data .
Then next more important is Social Media Information : collecting these many data after that sending through the engine we will get that much , lots of analysis we can make such as sentiment analysis , credit analysis so many of these kind of analysis .
Once customer complete data behaviour is being tracked . Tracked meaning understanding the behaviour pattern . So using these insights two examples .
One example will be one dashboard will come one particular customer how much time he is interacting with them or what are the modes of interacting with them the bank . So that dashboard shows this much percentage through mobile , this much through web ,this much through social media or through calls so these kind of all modes we can get in detail .
The
second example is preventing customer churn that means their the track record
when will this customer will leave the bank and go is already understood and
predicted accordingly we can do things required and retain customer . So these
kind of very important examples the big data application given to bank
2. Fraud
Detection and Prevention :
So you all have a basic idea about how theft happens in a bank so it is divided into 3 types. First one is Identity theft , second one is credit card fraud , third one even if it is not related to bank it is related to financial institution Insurance Fraud . So lets see how we can through Big data application prevention.
We can see credit card fraud for example a customer suddenly got a information , where customer is spending 2 lakhs with his card in US in Jewellery and a message came , so these many years through customer pattern he has not gone outside India also his expense , payment , spending category after seeing that his account 50000 not more than that he has paid also not having a habit of buying jewellery so through this transaction suddenly the bank one information will be given to the customer is it seems like a fraudulent activity ? . Thus the customer can automatically block it and say it is not my transaction so they will be saved . So that we has a customer individual a bank no human interaction so these can be monitored by big data application .
The second one is identity theft you all know that todays latest is all about machine learning and artificial intelligence so through are face detection and finger print detection most of these we can lower these issues . So similar to that is the insurance fraud you all know that in US almost all 80 percent automobile insurance fraud is happening so that is an incident happens money will be taken so that all this tracking can be kept and fraud can be prevented .
3. Personalized Marketing :
This where big data applications very important feature that means this allows the analyzed reports of banks to give the right product to the right person at the right time based on their needs so through an example we can understand :
There was a customer John in leading bank has a mortgage loan and has a credit card and also has a normal account . So John planned to renovate his house for example buying furniture through credit card is used thus the bank can analyse and understand the pattern . So the bank keeps an offer when that time the credit limit is got over the bank understand that it is a risk after that the message is given that you can higher your credit limit at this time it is very important . So like this every one bank has their data social media throught that understand what all they want to buy they will keep offers within this percentage this many months you can close given , definetly that customer will buy so like that personalized one by one customer understanding their behaviour one one products can be sold .So added to that different one is the recommendation engine that is when we buy one thing through Amazone ''you have bought this "so you recomment to buy other things so the recommendation engine is also the big data applications for this customer is given .
4. Risk
Management :
We all know that bank risk department is in all banks . So that means the investment that they do where bank is taking all the measures . So that means this predictive machines is very helpful . So that means when the risk comes if the market condition is bad many people has the difficulty to pay and the second one is the competition where if it is a very competitive market there is risk . Third one is customer trust worthiness that means how much the customer is trust worthy so that means a customer updated data we can keep one risk analysis where the big data application analyse and gives data to Bank . In that very clearly it is said about the credit scoring accordingly the bank decides so that means data driven decision making very important . We humans cannot be able to do that . So that means a credit card mortagage loan is it suitable for this customer it is very important information for the bank. So Risk Management is a very important tool by the big data application .
5. HR
and Operations :
Human Resource and Operations in the bank this data science or artificial intelligence products is used we can see so for example a very leading bank in GCC using visual and image based artificial intelligence bots that is using this cheques are validated. This 37 vision image validation is done by these bots . This leading bank almost 20 thousand cheques are received so their total working hours 70% is used for the same but by using this bot within minutes they complete this process so that much efficient this is working . The second one is setting effective staffing levels that is bank always have 60% staff cost as per analysis so that is estimation in that the pattern how many customers per day they are coming their transaction time also the resource planning given to bank so effectively fix the calender by bank how much is the resource required all those information will be given by bank through these tools .
Finally the last so that is when we call the bank customer get the information bots is available so this is all the output of artificial intelligence so likewise some pattern where the customers what all they are facing difficulty from the bank is it through online they the difficulty or bank difficulty so all these information can be given by bank to serve the cutomer the best .These kind of predictive information bank is getting . So likewise for the cost so how many staff can be reduced ,while handling the bots the cost considerably is less . Thus you can understand how this big data application combined with Artificial Intelligence , machine learning how the banking process supported or how the performance is maximum .
So adding to these there are so many other technologies fintech , cryptocurrency this all banking process revolutionize transform different tools but we have seen now what is happening the process how the big data application supports.
Parting Thought
“ The illiterate of the 21st centurry will not be those
who cannot read and write , but those who cannot learn, unlearn and relearn.”
Artificial Intelligence
