Fraud Detection to Payments Processing - a Natural Extension for The ai Corporation

As ai Corporation expands its footprint into payments processing, the company is looking at converting its core competency of using ML for fraud detection as its USP. Exploring new geographies, it has partnered with Quatrro, understanding that partnerships are the need of the hour as they strengthen both the parties’ position.

Nidhi Prabhu: It is good to have you with us today, Mark. Could you please tell us something about The ai Corporation, and your journey in the payments space?

Mark Goldspink: The ai Corporation has been around for about 20 years. I used to work for a company called Retail Decisions, which is pretty well known in the payments world. I was Managing Director for their payments and fraud screening business, which was bought by ACI in 2012. I joined ai in 2013 with the responsibility of turning it into something big and great, just like the Quatrro Company. And that’s what I am trying to do now.

The ai Corporation was originally a product-only business. We are now moving into more of a service business, but we do not necessarily want people delivering our services. What we see is technology services, not BPO services. That is why we see the opportunity with Quatrro as being something that we can help build together. Essentially, we see ourselves as a technology company that will build the products and services that will be used, we hope, by some of our large clients over a period of time.

Maybe it is worth giving some context to my background - I come from the oil industry, so I am not originally a ‘payments person.’ However, in 2000 I moved into the payments business. I’ve been in payments, CMT and e-commerce since then; this has been quite an interesting journey. Petrol to payments might sound like a strange transition, but there’s quite a lot of logic there, because petrol payments I believe is one of the fastest moving consumer goods. When you’re filling up a tank of fuel, you have to get it right every time, because if you don't, it's a disaster. Whereas with transactions on the Internet, if you decline someone, it’s not quite as insulting as being stood up in a queue at the counter of a petrol station. A lot of my expertise and knowledge comes from making those transactions as seamless as possible. That's what excites me and interests me most.

Also, in the petroleum sector, the ability to speed up the process has been something that is in a constant demand. So pay at the pump was obviously introduced - I don't know how big pay at the pump is in India, but in other parts of the world, it has become a big part of making the transaction process as quick as possible. We launched it in the UK when I was retail director there, and it was hugely successful.

However, fraud was a real problem. I had to understand it very quickly because we had just rolled out a lot of these pay at the pump terminals in the UK. So, that kind of knowledge and expertise is really about security and how transactions are managed securely. It’s something that I have probably been doing, even though I have never thought about it, a lot longer than since the year 2000. It's an area of constant change, constant interest, and something that we are very passionate about at The ai Corporation.

Nidhi Prabhu: How do you use machine learning in your products?

Mark Goldspink: We started saying let’s build ML so simple that the CEO could use it. We are making it simple for our customers to use by giving them the control, so they can stop fraud immediately. We have self-learning algorithms, which is probably the closest point to essentially where ML starts to really come to its own. To me, it is not about AI or ML, it is about solving a problem that I think my customers have: the problem that real-time data, real-time information and data in its entirety is getting bigger and bigger and they need solutions that can help them prevent fraud quickly and effectively. Operationally, that’s where I see this technology being deployed. To me, it is not a conceptual chase of AI and all those great things, today it’s about ‘how do we help the operational folk in our organizations do their jobs and prevent cybersecurity issues?’

We then bought a payment gateway, which is now integrated into our fraud system. The reason for doing that is that there’s no better way of testing your own system and your own technology than running it yourself. When we run it and manage it ourselves, it means that when we release the software, it is based on R&D to a high standard. That acquiring capability plus fraud functionality is something that we are very proud of.

We’ve just been approved by the FCA in the UK. So we hold an e-money license now and the reason the application went through so quickly, we believe, was because of our fraud background. There are not many people that come with a fraud system first, adding on the issuing side second. It’s usually the other way around. So I am glad that we did that, and that meant our approval was swift.

Nidhi Prabhu: Isn’t Quatrro more of a competitor for you than a collaborator? How did you think of getting into collaboration with them?

Mark Goldspink: To me, partnerships are the only way to expand your business. A partnership works on two parties being able to work together. I think culturally, Quatrro and I have always got on very well. Our partnership with Quatrro works very well because we’ve got a common goal, and I think common goals and cultures are the things that drive good partnerships.

Nidhi Prabhu: What about your existing customers, who may be the payment processors or the payment gateways, which are using your anti-fraud services? Do you foresee continuing that relationship as a problem, given that now you are also moving into the same space?

Mark Goldspink: I think we will white label some of the technology that we have for customers. So, if Quatrro wanted their own payment gateway, we would white label it for them and give them that capability. We believe that there is a segment that we can service where there isn't necessarily the competition, and that's what we've talked to Quatrro about in terms of how we get that.

Markets like India, Malaysia, Indonesia, offer opportunities to help provide that capability and give those franchisees and licensees, the ability to service local accounts or to help them attract customers to the facilities. That is something that we see an opportunity in.

Nidhi Prabhu: Some professionals suggest that you have to bring humans into the picture at some point. Now, one part of it, is that human cannot make consistent decisions. But on the other hand, humans can make a judgement call, which is something a machine, as of now, cannot do on a very reliable basis. Do you agree with the point that once the red flags have been raised, it can be better handled by humans?

Mark Goldspink: I agree 100%. Essentially, we are talking about man + machine. We are looking at ways to increase effectiveness and efficiency of managing tedious tasks, like managing 300 alerts per day by a fraud team. It’s essentially getting at the problem of attacking the false declines that go on in the industry.

MasterCard produced the number last year - out of $120 billion dollars in declines, the fraud problem was only $25 billion. People now are almost going too far in attempts to stop fraud. I believe that’s a processing issue, and it is about the number of people that can specifically work on those problems. So what we want to do is to take away the tedious activities and release these clever people to work for more value-added activities.

In the world of fraud what we do is a customer segmentation: we segment out ‘bad people’ using very clever technology. Segmenting out ‘good people’ is very easy really, but there is absolutely no reason we can't use our technology in two ways.

Customer segmentation and being able to understand the customer is the whole point of being able to actually segment down around an individual. That is something that I believe will play an active role in the payments space. The reason for that is because you want to know what people are buying, when they are buying it, and where they are buying it. So the payment transactions are a very, very interesting source of information from which you can essentially pull together a 360-degree view of your customers. We, actually, do it in reverse - we look for bad people. So what we really sell are ML decision engines.

To me, what is very evident is that decisions in the future have to be more data-driven decisions. And as we look at the rate at which data is being generated globally - it is just phenomenal. So the ability to screen that data and get that good, relevant information, I believe, is possible only with the use of technology.

I don’t think it is possible for the humans to be able to actually go through and consume that type of information - machines have to do more and more of that. The ability to continuously authenticate the individual will come as a by-product of a lot of the things that are being undertaken now.

Nidhi Prabhu: Given all data that a firm like yours will have in the future, how do you balance data analytics and the opportunity to turn it into a smooth user experience, with the privacy concerns around this data?

Mark Goldspink: You just described the challenge that we face as an industry.

I think there are several things that we can do better at and work at, but assembling the data to have the information you need in order to make decisions, to me, is probably the first point of the journey. I see a lot of organizations that have disparate systems. Their ability to actually pull data quickly and relevantly is very difficult. So, being able to centralize that data is number one.

Step number two is to be able to understand that data and perform the analytics. And then step number three has got to be predictive.

All of those things I see coming in the future (some of them I have seen happen in other sectors). None of it is technically not possible, it's just the way that the sectors move and apply themselves. Maintaining security around it is a huge, huge issue that has to be managed and carefully controlled.

However, standards that exist today, will not be fit for us in the future. They will need to improve and for that, we will have to understand how the deep web works and how information is passed around there.

Nidhi Prabhu: Mark, I’d like to thank you for sharing your knowledge, passion and providing rich insights into what you do at The ai Corporation in collaboration with Quatrro. Has been a pleasure to speak with you.