Kagiso Mpa, co-founder and Chief Technical Officer (CTO) of AI startup Seipone.ai talked to BW TechZone at the recent Global Entrepreneurship Week. Seipone.ai is a sentiment analysis tool which allows brands to analyse what consumers are saying about them online. The power of Seipone.ai is in its ability to even analyse sentiments posted in Setswana and slang.
Mpa touches on the current regulatory environment around AI, how to lead a team in an AI startup, and what the future of AI in Botswana looks like.
What motivated you to incorporate AI-based technology into your business model?
I’m a student of AI. I studied statistics and then went on to do my postgrad in Data Analytics, and then my Master’s in Data Science. So it was something that I studied and also motivated me to come back home and start my own thing because I realised there’s an opportunity here in Botswana. When I started, we were one of the first startups focused mainly on AI. The motivation has always been wanting to be the first to do this in Botswana and I know the endless possibilities of AI and I want to implement that in my home country.
Would you say that the landscape has started to change now?
The landscape is starting to change, yes. Conversations are being had. Even in the government, with the new Ministry of Communications and Innovation. There is innovation there and it shows that people are starting to shape up to the whole 4IR movement. I’m hoping to see more being done from the government’s side because we need political will for us to move the AI landscape forward.
How do you approach training and upskilling your team to ensure that they can effectively utilise the AI-based technologies that you’ve implemented?
First, when we recruit our team, we are always looking for people who are willing to learn and are willing to study. Because we know, and I can assure you, the landscape changes so much. The technologies that are used right now won’t be used in the next five or even two years. So we need people who come with that mentality of always learning and upskilling.
We always have these opportunities where entities like Google or AWS give these credits for saying you can take your people on board, and they study whatever course that they are offering. We always register them on those platforms and advise them to go back to school. We don't have much funds to say we can sponsor them at school, so usually, we just depend on the credits that we get from platforms like AWS, which we use in our organisation. If, say, any one of our guys says they want to go back to school, we set them free to go back to school, and they can still come back and work part-time with us.
As an entrepreneur what gaps have you observed in AI expertise that you think is hindering progress in Botswana?
I think a lot of people want to be coders and do not necessarily want to solve problems. They want to code through and through instead of learning the business side of it. We lack those business skills and skills in problem-solving for business purposes where you solve a problem, and you know that you're going to make money out of it. There’s a reward for solving problems in the AI world especially. I've seen a lot of people starting their startups, and they're mostly focused on building a product, and they're not necessarily taking care of the business side of it.
What we have done is that I'm a tech guy, but I'm not too good with business. I'm not too good with people and writing contracts and all that. So that's how I partnered with my co-founder, Nomsa Makgabenyana. She's very bright, she knows business and the legal stuff that needs to be taken care of. She's the face of the company. I don't even have to be on stages or taking up meetings to go in and talk with people. She takes care of that. She’s the business mind of the startup. I think we need that. We need to balance it out; there has to be that tech person, and there has to be that business person. In most cases, people make the mistake of trying to go at it alone because they have the tech skills, but they lack in the business side of things or the soft skills.
How do you ensure that the AI tools that you use align with ethical business practices and data protection laws?
The tools that we've built are tools that have been tested and are being used worldwide in the AI space. They are from Europe, and we know Europe has one of the best AI legal frameworks. So, we trust them. And also, coming down here, looking at ours, we’ve only just had the Data Protection Act not too long ago. The DPA is still lacking in the sense that it doesn't necessarily talk much about the tools themselves. It only talks about data.
But on our side, we never use personal data. It’s a matter of emotional intelligence. We know that we are not supposed to do 1234, we come with that from school, from the university where I was taught. If you are a data scientist, these are the legal frameworks that you're supposed to follow. These are the data protection and data management frameworks that we are supposed to follow. We take our experience and partner it with the legal framework and the guiding framework that we have in our country, and that's how we approach it.
Finally, could you describe a situation where a solution didn't necessarily meet your standards and how you overcame the challenge?
We have never experienced that, except that I would give an example of our competitor, which is an international tool. It doesn't consider our local languages as data. So, in most cases, it disregards our languages or our data in that sense. What I'm trying to get at is that this tool gives out wrong results. Because ours is a sentiment analysis tool, it paints all sentiment tools in the wrong light because it will maybe say 20% of Batswana are not happy with brand X. And then when you go back to see their processes you realize that it has disregarded a lot of Setswana content, and a lot of Setswana opinions.
It picks the perfect English sentences and text and all that. So, it is inefficient and maybe needs a body that will regulate it because these are conclusions that are made out of our opinions, and then it puts them out there onto the international market, and you realize that it's painting the wrong picture about our people because the conclusions are not scientifically correct. We need laws that speak directly to tools, and to the processes that we take into developing these tools because we are only looking at 4IR as an umbrella and we should be looking at it in its separate forms because it's a lot of things that go underneath the whole 4IR umbrella.
Do you have any words of advice for young startup business owners or young people who are trying to get into AI-based technology?
My advice is to just keep on building. Build, build, build. Because at the end of the day, we are dealing with Thomases, people who want to see something before they believe in it. So, let's build and let's solve problems. Once you solve a problem, there’s a reward for that, so let's build solutions that solve problems our African problems.