#58 Riding the Third Wave: Kamal Budhabhatti’s Journey from Mobile Banking to AI Transformation in Africa - Lessons for Digital Banking Platforms
After revolutionising Africa’s financial sector with mobile banking and super apps, Kamal Budhabhatti is now leading the charge into AI, defining the future of digital banking on the continent.
Artwork by Mary Mogoi - Website
Hi all - This is the 58th edition of Frontier Fintech. A big thanks to my regular readers and subscribers. To those who are yet to subscribe, hit the subscribe button below and share with your colleagues and friends. Support Frontier Fintech by becoming a paid subscriber🚀
🤝Sponsorships - Frontier Fintech has high engagement rates amongst a readership that includes key decision makers in the African Fintech industry. People read Frontier Fintech with intent and this would be very valuable to your brand. We provide;
Sponsored posts - A banner ad appearing at the top of each email sent out to subscribers;
Sponsored Deep Dive - A deeply researched article that educates the market on your offering and what differentiates you from the market.
Reach out to me on samora@frontierfintech.io to discuss sponsorship opportunities. Spaces are filling up;
🦉🦉 Advisory - Spaces are filling up for advisory work as well. If you’re keen on having a discussion with me about being part of an advisory network for your company, organisation or accelerator, kindly reach out via email. 🦉🦉
One can almost trace the origins of Africa’s digital banking industry to two gentlemen in the late 80s and early 90s when Barry Ryan and Justin Chinyanta founded Fintech International Group, now part of Loita Group. The two men were former Citi Executives and they had seen a gap in the provision of digital banking services. Banks and MicroFinance Institutions were being formed left right and centre and app-server technology was coming into the mainstream due to the growth of the internet and networking technology. Banks needed to modernise their technology from their core banking systems all the way to their customer facing systems Western software was too expensive and therefore local solutions needed to be built. Fintech was indeed a pioneer and built systems across the continent. This company birthed a number of digital banking companies such as Eclectics and many others. It can be considered the Fairchild Semiconductor company for Africa’s digital banking boom.
Whereas Barry and Justin came from the storied halls of Citi and HSBC, Kamal Budhabatti took a more circuitous route. He came to Kenya initially as an immigrant worker, toiling in one of the numerous manufacturing facilities scattered across Nairobi’s Industrial Area, an industrial expanse sandwiched between the Jomo Kenyatta International Airport and the Nairobi Railway Station. This stint didn’t end well and he was soon to be deported back to India by his employer. Whilst in Nairobi as a labourer, he had started dabbling in technology, building primitive banking software. Not to be deterred, in India, he found a way back to Kenya now as an entrepreneur and with the laser-like focus to build something of his own. Kamal got to work building banking and microfinance software. He got a break in the early 2000s, working with Dr. James Mwangi of Equity Bank to build Bankers Realm, the core system that Equity used until they upgraded to Finacle. Dr. Mwangi taught the young Kamal all about banking and particularly how bank ledgers worked, on his part, Kamal taught Dr. Mwangi all about technology. This initial break laid the foundation for what is now Craft Silicon. A sprawling digital banking business with deployments all across Africa and Asia. With over 150 million people using their digital banking products everyday and with the company making revenues of just under US$ 100 million annually. Craft Silicon is a major player in Africa’s core and digital banking market. Their systems are deployed in every African country bar South Africa. It’s a testament to the grit and determination that made Kamal take that second journey back to Kenya.
Source: Craft Silicon
In 2016 together with Safaricom, Kamal launched Little Cab. Little Cab is a super-app modelled around the South East Asian super apps such as GoTo and Grab. It’s built around the idea of daily life services such as mobility, food and entertainment. In his words, the idea of founding Little Cab “stemmed from a concern that banks would eventually develop their own technology in-house, diminishing the need for companies like Craft Silicon”. Little Cab is arguably East Africa’s largest Fintech super-app with over 5 million users. Whilst it competes with Uber and Bolt, it has found a niche in corporate services given its ability to build custom solutions for corporate clients.
Kamal has built his companies in such a way that he’s free to focus on his core skill of being a technical innovator. The first time I met him at his office, I was taken aback by not only his diminutive and gentle stature, but also his blood shot eyes, reflecting a week of all-nighters. As the self-described Head of Product Dreams and Designs at Craft Silicon and Little, Kamal spends his time ideating and solving hard technical problems.
Having successfully rode the first technology wave of the early 2000s with Craft Silicon and the second smartphone era wave by building Little Cab, Kamal is best placed to ride this next wave of AI. Who better then, I thought to myself, to walk us through how the digital banking industry should navigate the AI wave. What keeps him up at night? What core advantages does Craft Silicon have and what are the challenges and opportunities that will be thrown up by this AI wave? His insights should be useful for the entire digital banking space.
I sat down with Kamal one Friday afternoon at his office for a long-ranging conversation on AI and Fintech. The article will cover the following;
Why Digital Banking Platforms need to embrace AI;
How Craft Silicon is approaching AI;
Practical guide on executing AI as a digital banking platform;
How to commercialise AI as a digital banking
Some thoughts around the future and the role we play in it;
Why Digital Banking Platforms need to embrace AI
I wrote before about AI in financial services, the first part covered what AI is and I went quite deep. The second article was based on some predictions of what will happen to the banking industry. One of the predictions I made was that Digital Banking platforms are on shaky ground given that banks will have the capabilities to bring everything in-house. It’s a framework that Kamal has shared and maybe not to the same extent as me. Kamal’s move into AI is borne out of both commercial logic and existential fear.
In terms of commercial logic, the fundamental difference between machine learning and neural networks was so big that Craft Silicon realised that this is a complete paradigm shift in terms of technological capabilities. I asked Kamal why Craft is investing in AI and what’s informing this move into AI and he had this to say;
Kamal: So, before ChatGPT came in, we were trying to do a lot of things in machine learning because we have so much data with us, and if today I aggregate all the customers between here and India, we're actually crossing 150 million consumers using Craft Silicon. We have so much of this data, and we then at some point started writing machine learning algorithms for credit scoring. Okay, we tried, but it didn't come up the way we had envisaged. And then at some point, the chatbots came in, ChatGPT came in, Big Data came in. And I think it just changed everything. We looked at our data, and we quickly realised that this is the future. Everyone would require something to do with AI, and AI is not going to be a separate component. It has to be embedded in your day-to-day operation. So it's not like, this is my AI software, it has to blend in with what you are doing.
Samora: Yeah, so I've got two questions there. So machine learning is just putting structured information within an algorithm, whereas what we've seen in Chat GPT is more around unstructured data and deep neural networks. So when you said that the machine learning thing didn't work, was it an architectural thing, or was it more of the business dynamics?
Kamal: No, it didn't work in a way that I think the neural networks were able to deliver better outcomes. So, at that time when we were just doing machine learning for us, even the customers appreciated it was giving good results. But now, when we look back we would say, "Oh my god, that thing we were doing was pretty, you know," in a bad state.
Samora: Exactly.
Kamal: So, on AI, I think conversational AI is going to be also quite big even in the banking industry, not for decision making, but for taking off some of the load from the customer. The best way to interact with AI, I personally think, is going to be voice. So we are working on voice, we are working on text, and we are also working on images, but everything currently is revolving around our existing customers, our banking and fintech customers.
In essence, tinkering with both machine learning and neural networks given his role as the chief scientist at Craft Silicon led Kamal to the natural conclusion that AI is a platform change that should inform the next wave for Craft Silicon. In fact, the discussion focused on how it’s a matter of life and death for digital banking platforms;
Samora: I think that's very important when you're having such a platform change to have someone who's just spending time thinking, learning and reading about this stuff.
Kamal: Yeah.
Kamal: Because for us it's also survival. We are in that mode. If we don't take that step at the moment and the companies that are not taking this seriously, if another company doesn't think of it the way we are thinking, maybe in a few years they will become irrelevant.
Samora: Yeah, a lot of people in that industry are going to become completely irrelevant because it's a 100 times improvement. So if you have a bank that has fully functioning AI and it's helping customers in all sorts of ways and you have a bank that's still using plain mobile banking, it's not even a comparison, even the shareholders will tell that bank what the hell are you doing?
Kamal: Yeah.
Kamal: And so it's a big change for us and I think many of these companies. We need to take this very seriously. This shift has to start happening as soon as possible otherwise at the last moment it can fall apart.
The AI opportunity for digital banking platforms is based on a technical realisation that AI is a platform shift and therefore, there needs to be renewed focus on AI enablement. This renewed focus will cause further disruption in the industry and will be an existential risk for many existing players if they don’t pivot towards AI enablement.
The obvious next discussion therefore is, what approach should be taken to pivot from traditional digital banking platforms and towards AI enablement?
How Craft Silicon is approaching its AI Journey
While there may not be a sure fire way of ensuring success in becoming an AI enabler for the banking ecosystem, Kamal’s approach is based on logic and hard earned commercial experience. It reflects a mix of caution and boldness that comes with decades of building scalable banking systems across the continent.
Experimentation & The Need to Build Trust
AI is full of unknown unknowns given the state of constant innovation that is happening at the edge of AI and the uncertainty as to how the final end state for consumer financial services will be. Given this, Craft Silicon is focusing on both building trust and constantly experimenting. Kamal started off this section by asking my about my views on something they’re working on;
Kamal: We are doing something, I want to validate that with you. So, what we want to do, as the first thing, is, on your mobile banking, because that's what we want to pilot with. Before even a customer logs in, there is a chatbot. And in both voice and text, a customer can ask for any information about the bank. Once he logs in, then the customer is already authenticated. We put the data for that customer, and let the customer interact with that data. So, it's not going to make any decision on his behalf, but the retrieval of the information is going to be on a separate level. I need to see my account balance. I don't need to go to this app and navigate numerous buttons. I could just ask or I could just speak to it, and it gets me the info. That's the first thing that we want to roll out, very basic.
Samora: You made some good points around it being like a multi-step thing. You won't start at the end. You won't start with this advanced AI that does everything for you. In so far as it's a first step, that's great. The way I think about AI, especially in financial services, is that it should enable things that weren't possible before, in a much more dramatic fashion. When I look at myself, I've worked in a bank. And then when you leave, you look at yourself as a customer, not a banker, right? You leave that industry and start looking at banking apps the same way you look at them as a customer, not as a producer. And the thing that I always think about is that I have financial objectives. I have financial goals, and I need this thing to help me reach my financial goals. I think that's one of the main things, so I think when you're launching, that that's fine in so far as it enables people to start interacting with their bank like they would with an AI.
This idea of experimenting and starting with an MVP if you will is built on the idea of trust. Kamal expounds on this;
Kamal: Yeah, I think the reason we approached it like that is I feel that a customer needs to build trust with the AI components. From day one, if you let the AI start advising you, I think confusion will start happening because it might advise something, and I might not take that advice, or I might blindly act on the advice, and fail. I would then never come back to it ever again. So, I think this journey of giving, in, in slow doses, is to gain the trust of a customer. Okay, fine, today you do this. Come tomorrow, maybe he'll say, "Okay, send airtime to my wife, my daughter," and the system will be able to do some financial transaction of a smaller level then grow from there.
Samora: And if you think about it, that's how ChatGPT has done it, right? They just launched ChatGPT. The whole idea was, "What is the form factor for AI?" Their innovation was that the form factor should be a chatbot, and then slowly they added voice. Slowly they've added a number of things, and a lot of people now discuss really personal things with ChatGPT. As a starting point, get someone used to AI, right? And I think it's going to be like a real game changer, in the sense that you now have a relationship with your bank.
It’s a very logical approach and I brought out the parallels with ChatGPT given the way people are now using ChatGPT. The video below shows the kind of trust people can build with their ChatGPT given the advanced voice features. It’s this trust that Kamal thinks is critical to build from the start and that can only start with small, timed releases;
Ever the instigator, my thinking was that, if Craft Silicon has access to all this data and has built expertise in AI; why not build a customer proposition that sits in between banks and truly empowers the customer. Ultimately, with their visibility, they can add more value to the customer this way and dis-intermediate banks to a level. His response reflected the caution that is required when there are so many unknowns;
Kamal: It’s something that you know we had also thought of it that should we go out as an independent player working across you know all the financial institutions? or pick up somebody and start from there? We felt that to start with we need to hold hands with somebody and then that's what we are doing. What you are saying though I think is completely something that will happen at some point because a customer needs to get more and more choices where everyone will start communicating with each other. The Customer is least bothered about what is coming from where he just needs financial services delivered to him.
This caution is warranted, so how do you choose the bank you will work with and how do you work with them?
Choice of Bank
Given the need to experiment and the need for speed, Kamal thinks that wasting time with politics is a drag. Rather, it’s important to find a commercially minded bank that wants to get started on these experiments.
Kamal: The big banks are all our clients or we deal with every bank in this country and every bank in East Africa. I think we deal with most and several other banks outside East Africa. Picking up the right bank is very critical. I don't want to go to Bank K, I don't want to go to Bank E for this. However a bank like Bank N for example and Bank S, you know these are the banks are big enough but also agile enough.
Samora: Yeah. My experience has always been that owner operated banks of a specific size are the ones that really innovate.
Kamal: Yeah. And then a lot of politics comes into the bank itself. There are some people who want to support us, some people who have somebody else… You also want to make sure that you have your focus on what you want to do. You don't want to waste your time fighting other battles and the main battle is lost.
Samora: Yeah. And if experimentation is the key factor, then the size of the bank doesn't matter.
How to work with the bank you choose - Again, experimentation
It’s clear that you shouldn’t work with a bank that’s mired in politics. This was my experience at Sote as well. If you’re bringing in a new innovation, it’s important to work with a bank that’s willing to co-create.
Kamal: And actually that's what is happening. We have a customer who we are working with them to create some AI strategy, but still at a very infancy level because AI is, as much as everyone is talking about AI and say, "Oh my god, this is going to do wonders" as we move forward, but I think nobody has really created a core, both competency and the control centre for AI. The MD talks about AI, the CIO is talking about it, somebody else is talking about it, but they've not sat down together and said, "Let's have a clear strategy." It's a very vague strategy. We are now coming into the picture and saying, "Look, allow us to try and don't take this as it's going to start generating returns and profits from day one. Allow us to experiment. Allocate some budget, we try 10 things. They might look fancy. Maybe from 10, eight will fail, two will pass, but allow us to do that because if you will not do that, then it may become very difficult at some point when you want to launch. So, get into that day-to-day operation of using and experiencing the AI. So that’s where we are with Blue Beetle which has been formed by Craft Silicon specifically for AI experimentation.
Samora: Simply, what you're saying is that you're looking at the first step and saying, '”Guys, it may not make sense to have a hard-coded strategy right now. It's good to have a strategy, but it shouldn't be hard-coded. Rather, let's run a series of experiments. Let this team set up like some credit analysis, let this team do custom onboarding. let this team do an app that's just pure AI that is voice-based, and let's see what works, right?"
Kamal: Yeah.
Samora: But that's not how banks traditionally operate. They operate on business cases, so, how are you navigating that?
Kamal: And so that's where players like us come into the picture, who have worked with these banks for a very long time, we need to convince them. So, as a new player, who is good at AI but doesn’t have relationships, it will be difficult. When we walk in, they have worked with us for several years. I think we have more convincing power in the banking industry.
How to actually execute AI as a digital banking platform
There are some technical considerations that every digital banking platform provider will have to think about. To start with, you have to select which one of the foundational models you will work with. Craft is using Meta because it’s open source but if they find that OpenAI works, they’ve set themselves up to easily shift. Kamal walks us through, how you set up at a very basic level;
Samora: You've got Meta building these open source models but then now you have to train your own models to now help your customers. What is the journey like? What are the main things you have to do? Like let's say it's me. I come to you and say come I've got this idea. I've got these resources. I've got some technical capabilities and I want to help Bank X. What is the journey from zero to one?
Kamal: So the first thing is banks would be very conservative in terms of putting the data outside their data centres. So you'll have to bring the model from wherever it is to down there, then try to train it. You know we even thought that okay maybe we'll bring the model down but let's take the data up there, train the model and then bring it down there but some banks have their own views of not allowing the data to go outside.
Samora: Mhm.
Kamal: So when you are training something in-house I think the costs are very high. So that's something that I think banks would have to really figure out. But now I think the cloud providers have started giving out a private cloud within the public cloud. So you know like in India for example now they have a data centre which is only for these kind of services where as much as it is on Microsoft Azure and it's on on internet somewhere but you get your own private container and that does not just go out and it makes it easier for you to train, to pick all your data from here take it up there get it trained generate the model and then just bring the model down and let it run on the premise so those kind of things are all happening.
Samora: And when you talk about training right, the foundational models are already pre-trained on some things, right? And then you're coming in and saying, I've got, let's say, transaction history for the last 10 years for these 10 million customers.
Kamal: Yeah.
Samora: So, practically, are you taking data from this 10 million customers and this from a very rudimentary perspective putting it on top of what Llama has already done?
Kamal: Absolutely.
Samora: And then now running doing some tests, right?
Kamal: Yes.
Samora: And then you run some tests, see if it's bringing good results. Yeah. and then doing some kind of reinforcement learning on that. And then I eventually say okay now we have something that we think can work. Is that what's happening?
Kamal: That's what is happening.
Some questions definitely arose, particularly around how to get the data right. I wrote in a previous article about how data will play a key role in your AI strategy. The question is how do you aggregate all data points to feed into these AI models; Kamal had some interesting thoughts about this;
Samora: And one one one thing I had in mind was this whole thing of even as you're experimenting, databases, the way information is taken in within a bank. Where customer accounts are kept, where account opening forms are maintained. It's not really built towards getting into some kind of unstructured database and producing AI results. And one of the things I had in mind was even within the interim, there's a role to be played by helping banks, even without launching any product, get their databases to now support AI. And that's even from how you maintain meetings with your clients. Do you record them somewhere and put them in? Because all that information will be useful for the AI. What do you think about that? As you're experimenting, how do you think about helping banks get their data right?
Kamal: Yeah.
Kamal: When mobile banking started, the core banking systems were not capable of doing so many things. Players like us and many others came up with something called a middleware that would sit between a core banking system and the mobile and USSD platforms. This was to try and build an integration with the rigid core banking platforms to something more flexible that is required by the mobile banking and the customer experience on the side of the customer. I think the same thing is going to happen now where players like us will have to come in and build some sort of a middleware where, on the co-banking, or the meetings are here. Something else is happening, try to consolidate everything, transform it, and then let AI run on top of that middleware. So, some sort of an integration layer would come in just like what came in as ESB in those days.
The idea is rather than disrupt the entire industry, build tooling that will enable banks to be AI ready from the get go. It’s this idea that is shaping Backbase’s AI strategy with their “Intelligence Fabric”, an AI layer that enables banks to get started on AI quickly.
Of course, my experience with middle-wares and workarounds is that they eventually create too much technical debt. I put this to him and Kamal’s idea is that ultimately the industry will evolve into an AI driven industry and everyone will have to play their part, from the core banking providers to the digital banking systems.
Samora: I remember even from the ESB (Enterprise Service Bus) days, so you had the ESB days, then APIs came in. Core banking systems were not built for APIs, so there were still workarounds upon workarounds, but at some point things started breaking. In the sense that, and I remember I was running a bank, you do your first wave of innovation, you do all those ESBs and all that stuff, then you introduce your mobile banking, then you introduce another product, then at some point innovation becomes hard because you're dealing with all the workarounds that you've been building for the last 5, 6 years, right? And I feel like even in the African banking system, that's what's happening right now. I think part of the pace of innovation we are seeing is because people are dealing with problems from 10 years ago. Would you agree with me?
Kamal: Yeah, I agree.
Samora: So then even on the AI you can imagine the problems are going to be 10 times bigger. If you build a middleware. I understand that a middleware is like a good first step to get things going, but assuming that you could go from zero to one, what would you do? What would you advise a bank to do?
Kamal: So, I don't know what advice would be there, but I think the growth would come in that manner. You know, earlier the mobile bankings had no API, so you had to build something. Then mobile banking providers said, "No, you know everyone is doing that? Why don't I add these functionalities in my core banking platform?" It became a bit easier. I think the same thing would happen here today. You know the core banking guys may not have functionality to consolidate the AI. Somebody will come and do something. Gradually, the core banking guys will also realise that, "Why should everyone do that?" A lot of pressure will come on the core banking, and then they would start embedding the AI on top.
He continues to give me an example of how Craft Silicon has built its new Core Banking System “Nimble” with AI embedded from the start. This is primarily experienced through a chat bot that helps the user to navigate through every function such as account opening, card maintenance and reports.
Despite this, there are plenty of challenges that come up and that need to be solved. The biggest is people, I wanted to understand some of the constraints and this is what Kamal had to say;
Samora: Clearly the constraints are the amount of compute that you have on this private cloud.
Kamal: Yeah.
Samora: Right. That's the number one constraint. And the second constraint is the amount of data you have. Is there any other constraint?
Kamal: Knowledge is a constraint.
Samora: Got it.
Kamal: Of course. And just the ability you know, there's good knowledge sitting in the US sitting in China some in India but these guys who are experts in that area are extremely expensive. The knowledge is out there but it’s very difficult to attract those talents because all the AI experts are now all with these bigger players getting paid huge salaries. So yes, that's a constraint so players like us who are smaller players compared to those larger players we'll have to do some trial error, we'll have to pretend to be an expert to convince the existing customers to spend some dollars with us. That is a big challenge. Knowledge is the biggest problem… and one we also fear that some knowledgeable guys would come in and possibly wipe us out completely.
It raises the bigger point of Africa’s role in the AI revolution. Will we be mere consumers of AI apps and producers of data that feed these frontier models? Kamal fears for a future in which platform providers become mere resellers for some of these bigger models.
Samora: From what you've told me then it's a risky place to be at and why I say it's a risky place is simple. If there are three constraints right, the number one is your data, number two is the infrastructure and then number three is knowledge. So you could argue that number one and two are easily solvable problems. Anyone who's resourceful enough can solve them. Number three is a difficult problem to solve and number three has already been solved by players in San Francisco and China.
Samora: So basically those are the two major players (San Francisco and China) everyone else is not really a player.
Kamal: Completely and so it's not worth their while. They've got bigger fish to fry.
Samora: But they can build their own AI that is knowledgeable enough to train models on its own right, an AGI or something and when they do that then they can dominate because they solve the hardest problem.
Kamal: Yeah and then the fear that I have and maybe it's a legit fear is most of the smaller companies will just become an agent or a distributor of somebody else's ideas and concepts. So these bigger players would have done something and then now a bank like KCB wants this and we’re just installing something. He sells it to me for $1,000, I sell it to him for $1,200 and with $200 I'm happy.The creativity of a Kenyan company is now gone.
It’s a legit fear. African digital companies are at risk of becoming re-sellers if AI goes to its natural conclusion. This is made worse by the talent shortage in the market and these are some of the problems that are keeping many leaders awake at night. Africa lacks the resources and focus needed to win in AI primarily money and talent. However, Africa is not alone. Most countries simply won’t be able to compete with the US and China. As much as these are challenges, Kamal sees them as opportunities when it comes to commercialisation.
Commercialising AI as a Digital Banking Platform
The commercial model for AI at least when it comes to digital banking companies is yet to be fully determined. What’s working in the market’s favour is the hype that is leading many banks to want to experiment. It’s an area that I really pushed hard for an answer to get clarity on the way forward. Understanding the commercial side needed us to flesh out why Craft is approaching bank partnerships, its trump card with Little and some of the factors that are working in its favour when it comes to working with banks in their AI strategies.
First, I really wanted to understand why Craft either through Little or another app doesn’t intermediate banks directly. The question was based on the insight that there is a resource challenge that Craft can solve as well as a consumer engagement opportunity since they’re building the platforms directly. If indeed a company doesn’t want to be a reseller of AI, why not go directly and own the customer relationship?
Samora: So we've discussed AI. We've said that you need to experiment with AI and they need to try out all these things. But then we've also come to the point where now if you push the logic to its end then you have a situation where you're either providing a service to a client or you are swapped out by Open AI or Llama and become a reseller. The next thing to add to that logic equation is, I'm unsure how quickly banks will move towards improving their services with AI, I think they'll set up committees and all those kind of things and it may take long for them to launch anything meaningful and even if they do they don't have any unique advantage to provide something world class from a customer perspective that's my view and I may be wrong right?
Kamal: Yeah
Samora: Where this question is leading is why doesn't Little now become a proper financial service provider. You've got all the tools you need. You know if you build any capabilities you've got everything you need to level the playing field. Why doesn’t Little start with its own drivers, offer them financial services, layer on top of that, market it to your customers and everything and then you build from there.
Kamal: So we want to do that. We want to be that devil but without being called a devil. We don't want to become a bank because once we do that it becomes very difficult for Craft Silicon to sell the products to our clients because on one side you own a financial institution and on the other side you are selling services to banks. We are actually providing financial services to our customers just that we don't want to be labelled as a financial institution. We do that under Little and we are learning a lot from there and some of those experiences we will take them to our bank clients
Samora: Okay, but the logic here is simple. You're a businessman, you build fantastic financial products. Your drivers start using Little Fintech and they love it and you start getting these kind of people, right? What's the incentive for you to not take this and be a bank?
Kamal: Because being a businessman, I know the business is bigger on this side (of selling financial services), you know if I just take Little and make it into Little bank the amount of profit and the revenues I'll generate maybe will be much smaller than what I’d make by selling these services across Africa and Asia. I’d scale to the size of say Equity Bank and that’s it.
Kamal: No, you're very right. Little is your experimentation platform.
Kamal: Oh yes, absolutely.
Samora: Absolutely. And you get to try out all these things. You run the core service of cars, logistics and everything, but you get to test out all these things and it could be even your first AI testing, right? But then once you prove out all these things, then there's much more money to be made in enabling all these banks that have 150 million customers to now make that money.
Whereas there’s the question of whether large frontier models will destroy the digital banking industry; Kamal sees the opportunity not in providing financial services directly but by digging in, experimenting and coming up with a better product for the banking industry.
There are some areas that Kamal sees as providing a massive advantage for Craft Silicon as they ride the AI wave;
Craft has built relationships over 20 years with hundreds of banks across Africa and Asia. In a sensitive area such as AI, these relationships would prove an advantage;
Local presence will become critical. Given that models not only need to be trained, they need to be retrained and fine-tuned. Having the technical expertise and local presence can help to get these models working appropriately. There won’t be a one-size fits all model;
Data is a big advantage, Craft with over 150 million customers under their platforms across two continents has plenty of transactional data that should help them fine tune models. This is an advantage that many players such as Backbase have, but it’s critical;
Balance sheets will matter - The entry fees for a seat at the AI table are not cheap. In fact, according to Kamal; So the days for starting something from garage I think will not be there. Some point in the earlier days you would just start something from a garage with just a laptop that you had but now I think it's slightly difficult.
These factors have a combinatorial effect, you don’t need to get one right, you need to get all of them right.
A trump card that Craft Silicon has and that we have already touched upon is Little Cab and their ability to run experiments using Little Cab and later using their learnings to inform their AI go-to market with Craft Silicon. The future of digital banking is very interesting. The conversation went on and touched on a number of other issues. As always a conversation with Kamal is always insightful.
As always thanks for reading and drop the comments below and let’s drive this conversation.
If you want a more detailed conversation on the above, kindly get in touch on samora@frontierfintech.io
I’m open for advisory roles but there are only a few spots left. Kindly reach out on samora@frontierfintech.io if you’re keen on having a discussion about advisory work.