AI From The Perspective Of How It Has Evolved In The Banking Context

Middle East Banking AI & Analytics Summit was conducted at The Address, Dubai Mall Hotel, UAE on 18-19 May’2022.  Effiya Technologies as the strategic partner, was excited to associate with this summit, which sought to foster a hybrid collaboration between top industry experts and technology solution providers for viable opportunities in digital landscape of banking industry. Our Managing Director Abhishek Gupta explained on effective AI/Analytics interventions, industry best practices and why innovation is important for resilience and growth. 

 

Abhishek Gupta at Middle East Banking AI & Analytics Summit
Abhishek Gupta at Middle East Banking AI & Analytics Summit

 

Excerpts from Abhishek Gupta’s Presentation :

Let me start with a bit of a discussion around the evolution of AI in business.

 

Evolution of AI

Stage I

This is primarily where the Machine Learning or the insights that we are talking about, they were primarily generated by management consultants.  The data was scarce.  It was not very easily extractable.  It was coming in small chunks of pieces. And in majority of the cases the management consultants are the ones who were taking this data, generating certain size, and then giving them in the boardroom presentations.

There was very little traceability or tracking mechanism for whatever recommendations were made, whether they were successful or not. Of course as the time evolved, Amazon was one of the most famous cases, where it started recommending to you in terms of what is the next possible purchase that you can make, and this is when people started getting exposed to AI.

Similarly in consumer banking, if you look at it, almost in 1980s or so, we were aware of fairly decent work done in terms of credit scoring, a bit of cross propensity models etc., in the United States.  What happened was that there was a significant financial impact that was being delivered in certain parts of the world, not everywhere.  The rest of the world was still using the business insights and judgement etc., to make decisions.

Stage II

Let us come to stage II, this is where we are at today.  Here digital transformation is basically a key. You would have seen this from banking organizations in the last three years or so.  The initiatives on the AI side have been going on, for the last 8 to 10 years.  But atleast a significant focus on digital transformation, that is something which we see happening as of today.  What has also happened is that, along with this there is an interesting part of robotization which has also come in.

 

Robotization and Challenges

There are lot of processes which are today robotized, there is RPA, and there are models which you will see. Given the fact that there has been a little bit of automation that we have been able to bring in, there is a bit of predictive sciences that we have been able to bring in, hence it has also been able to deliver a significant part of the financials.  Now let us talk about how we see the future.

So the majority of the processes that we are talking about from an automation perspective, a lot of them today are plagued by the fact that they are not intelligent enough. 

One of the challenges that we felt is that, digitization or digital transformation from the perspective of whether it is RPA or whether it is from machine learning part, is a good step forward, but even today there is something called a sharp IQ, and there is something called a not so sharp IQ.  So the majority of the processes that we are talking about from an automation perspective, a lot of them today are plagued by the fact that they are not intelligent enough. 

 

Chatbot Behaviour – a Case Study

Let me take a simple example of a chatbot. Majority of the organizations in this part of the world have implemented chatbots.  The question is how many times you have realized that when the customer comes, and he is not exactly asking the same question, how many of the chatbots are able to address that question for that customer, and is able to guide him through.  The problem is very simple, the back-end of that chatbot where the rules were written, they were hardcoded.  So they expected you to behave in a particular way, if it did not happen, then the machine does not know what to do next.

 

Next Wave of Intelligent Processes

This is exactly what we believe will happen going forward. This robotization that you have tried bringing in, that RPA that you are talking about, the machine learning models that you are talking about, the predictive power whether it is great or not so great, what can be done to improve it further; these discussions are not necessarily happening always.  Also the stakeholders who are talking about this or evaluating this, they themselves are not necessarily very knowledgeable on that particular topic. So we believe that the next wave is primarily about making these processes more intelligent.

So we believe that the next wave is primarily about making these processes more intelligent. The second important part is that they will seamlessly start getting embedded into the customer journeys.

The second important part is that they will seamlessly start getting embedded into the customer journeys.  One of the challenges is that I have generally defined my digital transformation, especially the machine learning models, as a stand-alone process.  When these models are developed, we are trying to figure out that where in the process do I fit it in, and then comes the technology, architecture etc. You tried figuring out how the technology will work, and this is how the prediction will happen, and this is how my customer will be treated.

 

Process First and Modelling Later

The point here is, this thought process has always happened after the model is developed, more often than not. Actually you need to take a step back. When you are trying to think about it from a strategy perspective, you should be looking at the process first, and then saying that here are the points where I want that to be more intelligent.  If these are the test points, then what exactly do I want from robotization, or from a modelling perspective, what are your machine learning needs? Once you have defined that, then you go back and say that here are the specific areas where I am going to define one of those cases. You know how they will eventually operate, when the machine learning models or the digital robot that we are talking about has been developed.

This is something which we believe will happen more and more, but it is always an evolution. If you would not have done the first one, then you would not be facing the challenges of what has happened, and then of course you will not ask the right questions as to how to improve this process.

So this is broadly what I would say is how we see the future for AI in banking.  This is just a quick overview of various processes that happens in a bank, and when we say that it has been fairly well implemented, how good is the adoption?

Leave a Reply

Your email address will not be published. Required fields are marked *