Leveraging AI to increase the lending universe, with Sabelo Sibanda

More than 200 million formal and informal micro, small, and medium-sized enterprises in emerging markets lack adequate financing to thrive and grow; and the majority cite the lack of collateral and credit history and business informality as barriers to that financing. Today's guest, Sabelo Sibanda, has built CreditAis to solve this by providing explainable deep learning models that use alternative data to accurately score small businesses that don't fit the incumbent scoring criteria.

You can learn more about CreditAIs on their website at https://www.creditais.com (or follow them on LinkedIn)

We also mentioned Tuse, an earlier venture by the same pair. If that sounds interesting, learn more at https://tuseapp.com/

You can learn more about myself, Brendan le Grange, on my LinkedIn page (feel free to connect), my action-adventure novels are on Amazon, some versions even for free, and my work with ConfirmU and our gamified psychometric scores is at https://confirmu.com/ and on episode 24 of this very show https://www.howtolendmoneytostrangers.show/episodes/episode-24

If you have any feedback, questions, or if you would like to participate in the show, please feel free to reach out to me via the contact page on this site.

Regards,

Brendan

The full written transcript, with timestamps, is below:

Sabelo Sibanda 0:00

Those quiet experiments we're running to see whether there are real world use cases for some of this cool tech.

Brendan Le Grange 0:12

Buckle-up kids, it's time for me to relive some of my high school glory days. In my last two years of high school, I played hockey for a team that, under the coaching of Mike McConnachie, had become one of the very best in the country. I was surrounded by all state and national team players, though I was not one myself.

In fact, I was by far the worst player only there because the quirks of age groups and goalkeeping meant that I was pretty much the only choice but, them's the breaks. Anyway, in those two years, we never lost a league game and we never lost a cup game, but in each, our perfect run was spoiled by a single tournament loss. In my first year, we went down to Robin Petersen's Alexander Road team in a pre-season warm up that I'm happy to write off. As I said, I wasn't very good at the time, and I haven't yet expected any better.

But towards the end of my second year, all the practice against top quality strikers had improved my own game, and I was getting a lot more possessive of that unbeaten record. So it still stings a little to think of that loss to Grey High School 25 years later. I mean, they were greatly assisted by the appalling refereeing of Mr. Trott, but the point is grey High School has produced some good hockey players.

And as it turns out, they've produced some great entrepreneurs, too! Sabelo Sibanda, co-founder of CreditAIs today's guest. And who knows, maybe he's a reason for me to give up my grudge against Grey High School. Welcome to How to Lend Money to Strangers with Brendan le Grange.

And in case you're out the loop we're here on a Tuesday because this is the third in my little mini series of South African-themed bonus episodes, which will run throughout my holiday in August. So I'll be seeing you again on Thursday as normal, too.

Sabelo Sibanda, you are a machine learning engineer and co-founder at CreditAIs a FinTech that provides explainable deep learning models through API's to allow lenders to credit score small businesses. You're also a serial entrepreneur. So before we talk about CreditAIs, let's explore that what shaped you into the entrepreneur that you are today?

Sabelo Sibanda 2:41

Well, I've been very fortunate to have been exposed to entrepreneurship from an early age, my dad was in industry, my mom was a teacher. Whilst growing up, I was really exposed to a lot of their friends, colleagues, etc. I mean, tech entrapreneurship was something that was very attractive since high school even though I only got involved some 10 years later.

And yeah, what really did shape the experiences was just exposure. I think being around like-minded people makes it just that much easier to get involved.

Brendan Le Grange 3:09

Yeah, and CreditAIs is not your first business. So although that's the one we're really here to talk about today, let's start with, what is Tuse, and what was the thinking behind that business?

Sabelo Sibanda 3:20

So Tuse was a mesh networking solution. Initially, it was solving the problem of access to connectivity. We realised that mobile network operators didn't have the networks available in some of the most remote areas of the continent. We wanted to solve that, so we built an application that essentially allowed your smart phones - or mobile internet devices, more specifically - the ability to become nodes in this mesh. So instead of just receivers, they could be transmitters of data packets, essentially creating a mesh network through which you can send and receive data packets independently of mobile network operators.

Brendan Le Grange 3:55

And it did remarkably well. I saw you winning the ITU telecom award. What did that mean for you in what what sort of recognition? Have you seen for that business early on?

Sabelo Sibanda 4:05

Yeah, we were incredibly fortunate. So we won that global award and it raised our profile within the industry - unfortunately, it wasn't the most lucrative businesses very, very, very labour intensive, capital intensive, but we were lucky we got into the Founders Space accelerator in San Francisco, raised some VC from Draper Fund very soon thereafter, and ja, we deepen the relationships with the team that we had at Tuse, many of whom are still actually with us at CreditAIs.

So we got to meet some really incredible people as a result of having built that business.

Brendan Le Grange 4:39

And when we were scheduling this interview we had to avoid last week because you're in Dublin, what were you doing there?

Sabelo Sibanda 4:46

So in Dublin, we're part of the Irish Tech Challenge South Africa. I think Impact Amplifier, the Department of Science and Technology as well as a Technology and Innovation Agency (TIA) put together a curated visit to Dublin, very specifically, in Ireland to meet with public and private sector leaders, you know, to explore possible collaboration opportunities.

So that went swimmingly, wish we could have stayed a little bit longer to, you know, properly solidify a lot of the things that we began discussing, but incredibly, incredibly fruitful!

Brendan Le Grange 5:19

And always nice to arrive in a European heatwave to get the best out of a city like Dublin.

And I think that's a great way as well to introduce CreditAIs, which we've mentioned now. But let's maybe take a closer look, you found that again, with Thulisile Volwana, at a high level, what are you bringing to the market?

Sabelo Sibanda 5:36

At a very high level, we are bringing an operating system for lenders to the market.

Very specifically, we go about it through what initially it was an API that collects behavioural data, and using explainable machine learning, we're able to then infer credit risk. This is based off of 10s of 1000s of data points. And we run specific models - actually sort the data first, of course - but we provide an explainable decision making engine.

Brendan Le Grange 6:05

And if we talk about context, what does that microfinance market landscape look like today? How easy is it, generally speaking, for the sort of thin file borrowers that you're helping and the lenders that you're helping to get financed

Sabelo Sibanda 6:21

In the past that it wasn't all that easy, unfortunately, and one very startling statistic that we came across when we're doing our homework is that over 50% of loan applications get declined at the application stage, either due to being incomplete or failing the fica requirements, before we even get a chance to pull that.

So we wanted to find ways to make the process as easy as possible. And yeah, that said, Those that did proceed through to the scoring stage, especially for first time borrowers or small business owners face just terribly large odds to qualify for funding, it's near impossible for many folks that we're trying to serve. So we've opened up that market and we look forward to working with like minded people.

Brendan Le Grange 7:06

You have to wonder, what other industry would you turn away one out of two people queuing outside your door in the morning? If you opened your shop, and you sent every second person away, carrying your advert under their arm, no other business would survive!

But often in credit, we see the situation where we're turning away so many of the people that our marketing budget has attracted. So it really is important for borrowers, of course, to have access to the credit they need to grow their businesses. But this is something lenders want as well. And I like that about your messaging on your website very clearly serving both sides of that market.

And using behavioural data and machine learning to credit score these micro businesses, these individuals who would otherwise be hard to lend to so called thin files, I come from a traditional credit bureau background where we would use months or years even of historical loan repayment data to calculate our scores. How are your scores different to that?

Sabelo Sibanda 8:02

Yes. So what we found is that very often, especially with the thin file borrowers, that quality of data is not readily available. So we had to look at different data points. Of course, with the aid of a lot of published academic papers, you will find that our team is quite strong academically. And I suppose implementation wise as well, we look at these data points that have historically proven to infer really good indicators of one's future creditor behaviour.

So a very practical example, for instance, has GPS data - we found the gating on that is about 0.6 - so the behaviour and movement of people, as far as proximity to place of work is a huge indicator of consistency. And we're looking to sport a lot of variables that we can add together that show a consistent sort of habit pattern, that then we can infer on a behavioural base, whether or not somebody is able to repay a loan.

Brendan Le Grange 8:31

A lot of that traditional data, the cleanest version of it is 'did you pay your loans'. But we've always been looking for habit indicators, always looking for steady repayments, steady performance, to give us that feeling of consumer who're predictable, and that never used to really be available. But now we do have all these other data sources to lean on. And it's great to see that we're finding ways to turn that into a loan decision so that consumers don't get caught in that old cycle of you have to have credit to get credit.

You mentioned earlier about using API's to get the scores into the customers hands again, when I was in the space myself, we used to sell software and it used to be in most cases incredibly complex software, you would have to do a year long implementation and lots of training, you would be selling it to a lender with a team of highly qualified statisticians to operate but how does it work for a small lender? If you are going to provide them with this credit scoring capabilities? What do they need to do?

Sabelo Sibanda 10:00

So initially, we had the assumption that a vast majority of these lenders were already digitised, ie they had, you know, web applications, mobile applications through which they accepted a loan application.

We found ourselves doing our research. And actually, you know, conducting interviews with potential clients that digital transformation is a very recent thing, brick and mortar operations were pretty much the standard. So our assumption that, you know, we could just provide them with an API that they could plug into a pre existing solution was a flawed one.

And we revisited this and have since created white label solutions be the web applications or mobile applications. So basically, what a potential client does is get in touch with us let us know what they need, then we customise the solution accordingly, we realise that if and when they need a solution of this nature, we need to make it as easy as possible for them to adopt it.

Brendan Le Grange 10:51

One of the barriers to entry has in the past been, how complicated it was, and that something like credit scoring would have to follow so many other steps of digitization, so many new projects in the data space to get everything in and stored and cleaned and standardised over the months that it wasn't something you could think of. And now to be able to white label it and kind of go from day one is a good incentive, I think, to reward companies for making these steps and to show these quick returns on investment.

You're not talking about three years down the line anymore. So yeah, hopefully it'll be use it away, as well, to kickstart some of this and digitization of the credit process. It helps everybody you talked about if somebody wants to work with you, they just reach out and get the project going. What's the easiest way for them to do that which they go to talk to you at credit eyes?

Sabelo Sibanda 11:40

Yes. So the easiest way is definitely to go to our website, there, there's a link, they can click on just, you know they send us a brief message with their needs, we set up a call and ensure that, you know the type of solution or intervention that we introduce is one that makes sense for the business.

And then we get to deploying, we're working very hard to make it as quick as possible. And also as easy as can be. So the website will be the best place to go.

Brendan Le Grange 12:06

Great creditais.com

I'll certainly put those in the show notes as well. But before I let you go, we're a show that loves credit scores, we are talking often to an audience that loves credit scores. So without giving away too many of your trade secrets here, when you talk about the scores that you're offering, what sort of profile are you doing is that you're talking about SME and individuals, you're using alternative data sources, but in terms of actual performance, and predictiveness? How's that going? Are you seeing that work? Well, are you seeing people respond to it? What's the under the hood of the credit scores

Sabelo Sibanda 12:40

Sure. So I think in the beginning of this business, we realised that we have to put in some real time in order to prove that with potential customers. So we've been at it for just slightly over two years. And we're able to not only back tests from source data, but actually test or run our models on real data.

And the results have been quite impressive, if I say so myself, we're happy with that confidence of 86%. We do provide then a boolean results where you know, it's a Yay or Nay, We realise we don't want to reinvent the wheel with another score that one has to contend with.

And our humble assumption is that it's working and working well at that.

Brendan Le Grange 13:16

And what sort of lenders are you working with? Are you working across the board? Or were you seeing the most daily being generated to lenders were in the market?

Sabelo Sibanda 13:25

So definitely, in microfinance here in South Africa, we've been incredibly fortunate to join the microfinance association of South Africa and gotten exposure to just so many of these companies that are doing some incredible work, not just in SA, but the rest of the continent.

It's above and beyond just our lenders, but also collection agencies, also integration specialists, and just the entire ecosystem is kind of welcomed us quite willingly, which is something I didn't expect, we had tried to look at a different segment and the other reception was quite cool there. And so we've been doing well with the microfinance industry.

Brendan Le Grange 14:03

I have rolled out credit scores in a number of countries around the world. And certainly the approach was the same. And you could certainly recognise what was happening that the score in the Philippines would be dictated by who's reporting to the Philippine Bureau, what products do they have? How do they report it? What's the regulation? Like how is that different to the US and they're not you're bringing in a templated solution that you've got to try and fit everything to or and then then you move to Malaysia, then you move to Thailand. And it's always different products, always different data. And you've got to spend a lot of time trying to find out how the exact same story is being told in each different market because it's filtered through the products and the lenders and all sorts of decisions that have been made independent of you to get to you whereas with the behavioural data, it feels more natural that actually the sort of behavioural data you can pick up from GPS isn't dependent on how Bank A decided to set up its product. It's not dependent on the way the industry got together 20 years ago and agreed to share a standard data template, it's going more to source.

And I think it's got an interesting potential there for future growth where you can perhaps spread those sorts of scores. And of course, the context is going to be different. And you're going to have to adjust them for other markets, but it says have a lot of interest, therefore, a way of doing credit scoring that is less bound by the borders. And it'd be very interesting to follow that as you grow and to see how some of these shared behaviours maybe impact risk differently in different markets and how you bringing credit scoring abilities to more parts of the world.

Sabelo Sibanda 15:36

Yes, absolutely. I mean, one of the things that I'm very, very excited about is that we very intentionally got a diverse team, for instance, our CTO, Dr. Michael Kyazze, he's from Uganda. So having that insight of how things operate there, etc, firsthand is just an incredible advantage. And we're seeing a lot of correlation between behavioural data in South Africa and elsewhere. So we're just working to validate our assumptions and discerning open to experiments every day.

Brendan Le Grange 16:05

Yeah, good luck on that. And the team, as you said, it's a strong team, and you've brought some over from Tuesday, you obviously got a good culture and environment they're going you obviously in lots of recognition. So it's not just talk, you you're delivering great products. And I think it's a space that is going to benefit from this for lenders and borrowers and really take out some of that pointless cost that sits in the middle when the system isn't working efficiently. So I always love to hear a good story about a credit score. also love to hear a good story about innovation to improve access to credit and you're embodying both of those at CreditAIs.

So, Sabelo, thank you so much for your time.

Sabelo Sibanda 16:42

Cheers. Thank you so much, Brendan, for the opportunity. I sincerely appreciate it.

Brendan Le Grange 16:46

And thank you all for listening. If you enjoyed that, please do rate and review on your preferred podcast platform and share widely including on LinkedIn. And while you're there, send me a connection request. The show is written and recorded by myself Brendan le Grange in Brighton England and edited with assistance by Kane Hunter. Show Music is by and weak and you can find full written transcripts, show notes and more content at www.HowtoLendMoneytoStrangers.show

And I'll see you again next Thursday.