Across Africa, where a wealth of languages are spoken, too many are excluded from services simply because digital platforms operate primarily in a few global languages. For a large proportion who don’t speak those languages, they can be shut out.
Bolthale AI, a South Africa-based startup that participated in the Google for Startups Accelerator: South Africa, set out to change this. The team is building AI-powered natural language processing tools that empower businesses to connect with their customers in languages that they understand and trust.
We sat down with their founder, Thapelo Nthite, to learn about how they are using AI to level the playing field.
AI, for us, is about inclusion
Our goal is to ensure that people get a chance to properly participate in the economies and societies they are a part of. In the midst of great digital transformation across the continent, there is still the issue of linguistic inclusion. For example, in South Africa you can’t register a business online or open a bank account online if you don’t understand English. You are stuck.
I saw this first hand in an interaction with my grandmother when she asked me to load prepaid airtime on her phone. She asked me in Setswana, which is the language that both she and I speak, but she couldn’t do it herself because the voucher was written in English and her phone was in English. So the barrier was not her ability or articulation of what she needed to do, but the fact language made the process inaccessible to her. It made me realise that there are a lot of people like her who struggle to do very basic things that would enable them to exist properly in digital spaces.
The shift to people collaborating more with GenAI has opened up our business
My co-founders and I were technical from the get-go: we studied in the electrical engineering department at the University of Cape Town and were already doing research into technology that enables people to interact with computer systems using natural spoken language. So the bridge to using AI was natural for us.
What was less smooth was the fact that when we started the business, AI was not really in the mainstream — there was a lot of educating that we had to do to get people to understand that the technology worked and they could trust it. At that point corporations didn’t have the AI strategy that all of them have now and people kind of had to stick their necks out if they wanted to work with us. But that’s shifted a lot since GenAI — people use LLMs more in their day to day life and this mental shift has helped us get our solutions adopted.
From scaling our AI models, to using AI tools at work
At the heart of our business, we’ve got our core AI models and then the customer engagement platforms that sit on top of those models. We’re making the models themselves available to other organizations and innovators to plug into their own solutions because we understand that the technology has use cases far and beyond what we’re focusing on — for example, generating audio books or language learning. So we have the API that’s open to empower others, and then the SaaS platform.
Outside of that — and this one of several ways the relationship with Google for Startups was helpful — there’s our own relationship with other people’s AI tools when it comes to running a business. For example, how I use Gemini. I have a minimum of five Gems that I use regularly for different processes — like one I call ‘Mothusi’ (which means assistant in Setswana), which is basically my personal assistant. We also use coding assistants, which has sped things up and improved our code reviews.
AI is moving fast so we try to make upskilling our teams an ongoing thing. We have this thing called Speaky Wednesday where someone in the team will give a presentation or share something relevant to the business, such as new tools to keep on weaving AI into the fabric of how we work.
And what’s next? More languages. We’ve identified over 20 languages spoken across Sub-Saharan African by over half a billion people…. So watch this space.
- This article was originally published by Google Africa Blog. It is republished by TechFinancials under a Creative Commons Attribution-NoDerivatives 4.0 International Licence. Read the original article
