Seamless Xtra Editor, Mark Dowdall, speaks with global AI & digital transformation advisor and expert, Sudin Baraokar. Sudin has worked with a range of global 500 enterprises, SMBs, startups and Academia and has seen first-hand how AI is transforming the way businesses operate. Find out some of his key insights and strategies below.
Sudin, you are participating at the upcoming Seamless Middle East show in Dubai Wolrd Trade Centre where you’ll be speaking about some very important topics. One of these is identity verification and onboarding. When it comes to AI, where do you think this identity verification and onboarding ranks in terms of areas that could see the most transformative impact?
I think it’s very high up there because when we look at identity verification, we have to look at the entire governance, risk and compliance and the audit. I can tell you from my experience that large banks open more than fifteen to twenty thousand accounts on a daily basis. That’s how big the space is. It’s very important from a GRC perspective and from a RegTech perspective to get this right. It’s important to make sure that the entire onboarding process including KYC details, identity access management and related documents is done properly.
Where AI merges with existing governance system compliance measures and stacks is where we can derive a lot of benefit. And one of the benefits is to reduce fraud. Across the world fraud is running into billions of dollars and this is money lost by banks, money lost by customers.
So it’s very important across the total life cycle life, right from customer onboarding to verification, to the management of that particular ID, to the transaction and settlement of the various account details of the particular customer. I have seen from my experience that this particular governance risk and compliance is very important. Banks need to make a complete correct audit of all this and report it to the authorities and regulators on at least a monthly basis so that they also get a complete view of how the entire GRC and audit process is taking place in the bank for the identification verification.
Sudin, you mentioned fraud. We’ve certainly seen a huge change in identity verification methods over the past few years. One of the big challenges as a result has been mitigating the risks associated with fraud and identity verification. How do you think AI powered identity verification can help with this?
In terms of how to use AI for reducing fraud, we are already implementing multi-factor authentication and adaptive security, which based on the IP of the particular transaction, we can block that particular transaction if it’s a suspicious IP. So that’s already happening. But from my experience, I think hackers and bad state actors are also getting very smart and are maybe even using AI to beat banks so I think this is still an evolving space. There’s a lot more that needs to be done.
I think using AI, machine learning, building up the hyper parameters and getting additional data can all help ensure that ensure fraud is reduced. We can also ensure that our RegTech and cyber security measures are in place do it’s an evolving space. With all of this, a lot more needs to be done to have the correct algorithms to reduce this fraud, specific to what’s happening in the marketplace.
In fact, there’s a project that I did, where I built up an AI model, which could help predict the banking channels where fraud might take place over the next five years? That could be core banking, POS, ATM, customer onboarding, digital payments, or even CRM. So I think this space is requiring more work to be done but we are now far better prepared than what we were previously.
I’m sure lots of businesses are already thinking about how to begin the process of implementing AI in these areas. What are the things you think they most need to consider?
The most important thing is to just start. The other piece of it is to have a proper big data strategy and get a good structure in place. You must know the inventory of all your solar systems, your databases and your different applications. Global banks have something like 400 applications. There are more than 800 solar systems and 2,500 live production databases, and these application ecosystems are very complex so we need to understand them.We need to understand the volume, velocity, variety and veracity of all this data.
We then need to also use the latest technologies like data pipelines in conjunction with ETL extraction, transformation, and loading. One of the benefits of using the latest technologies like data pipelines is that we can merge the data science and statistics with our AI projects so that we get better results in terms of F1 score, loss, accuracy and epoch functions. We can also compare different data pipelines and I think using AI for emerging use cases is the right way to go about it. So getting your data strategy right is very important.
The third thing I would say is use AI for massive transformation. You’re going to spend a lot of money on GPU processing, on the data science skills, on AI experts, on your AI ops teams so make sure it really benefits your organisation. Make sure to get that technical data under control so your AI bill doesn’t go up. Also, cross skill and upskill your existing workforce as well as collaborate with specialists and technology partners.
I think this is how you have to really manage your AI project. I would say that at the start you can start your AI project with at least about 40 to 50 hyperparameters as a level one AI project. Then you start building a couple of algorithms, and for level two, if you want to get further results, you can increase the confidence and accuracy of your model. So, I would say increase it up to 150 to 200 hyperparameters.
And then, finally, as a self-audited system, which is going to be a level three, you can go up to 300 to 400 hyperparameters, covering the length and breadth of your data, of your databases, of your applications and of your source systems. Then go and deploy it, be it either on the cloud or hybrid IT or on prem.
Sudin, thank you very much for sharing your insights with us today. We could go into this in an awful lot more depth, and I’m sure you will at the upcoming Seamless Middle East event.
See Sudin on Day 2 and 3 of Seamless Middle East 2024
Panel: The race for identity verification is on: How will AI make an impact? | Stage 2
Panel: Decentralized exchanges: The frontier of alternative banking with blockchain | Stage 1
