African languages, phone calls and local context are at the centre of a new approach to artificial intelligence that aims to bring conversational agents to people who lack reliable internet access. The company, founded by Dr Olubayo Adekanmbi and its team, has repurposed earlier work on building AI skills into a commercial effort to create voice‑first AI agents supporting more than 75 African languages, the report said.
Designing for phone networks, not just smartphones
Rather than assuming users will interact through apps or web browsers, the technology is built to work over a regular phone call. That makes it useful in places where mobile coverage is widespread but data access is patchy or expensive. The approach aims to allow users to call an AI agent, converse in a local language and receive meaningful, context‑aware responses even when they are not connected to the internet.
The company has positioned the product to address practical social needs by focusing on four priority sectors: healthcare, agriculture, education and financial inclusion. These are areas where voice interactions — in a familiar language and with cultural understanding — can have an immediate effect on decision‑making and access to services.
“AI must be able to speak like me, reason like me, talk like me, act like me,” the CEO said, stressing that culture and context are as important as literal translation.
How the technology differs from standard translation models
The core argument is that translation alone is insufficient. The startup frames its work as producing agentic AI — systems that can act, reason and maintain a conversational role — rather than simply converting text between languages. This requires models that grasp not just lexical meaning but also cultural norms, common phrasing, tone and context-specific reasoning.
Technically, delivering voice‑first agents over regular phone networks means building systems that can cope with variable audio quality, interruptions and limited latency budgets. The company also prioritises cost constraints so the service can be affordable for users and partners operating in low‑income contexts.
- Local language coverage: More than 75 languages and dialects.
- Access mode: Ordinary phone calls (works with or without internet).
- Target sectors: Healthcare, agriculture, education, financial inclusion.
Origins and practical aims
The venture evolved from an earlier initiative established in 2016 to develop AI talent and capability. That prior effort focused on capacity building; the current company is commercialising that experience into products intended for real‑world impact. The founders say their mission is to ensure the digital revolution does not exclude populations who lack internet connectivity, are underserved by major global AI providers, or speak languages that are poorly represented in mainstream models.
By making the technology reachable via voice calls, the team hopes to bridge a gap between advanced language models and everyday users who interact primarily through basic mobile handsets. For policymakers, telcos and civil society in South Africa and across the region, the development raises a number of practical questions about regulation, privacy, content moderation and how public services might integrate conversational agents.
Where this could matter in South Africa
South Africa has high mobile penetration but persistent disparities in data affordability and uptake. Voice‑first AI that operates over standard phone networks could be relevant for public healthcare hotlines, agricultural advisory services for smallholder farmers, or basic financial services in peri‑urban and rural communities. Integrations with existing call centres and helplines would be a natural route to scale.
Practical adoption will depend on trust, language coverage and how well the agents handle culturally specific queries. The team emphasises that technology must be grounded and hyper‑local to avoid the one‑size‑fits‑all pitfalls of some global AI deployments.
| Sector | Use case |
|---|---|
| Healthcare | Symptom triage, appointment support in local languages |
| Agriculture | Crop advice, weather‑adjusted guidance |
| Education | Learning support and tutoring by voice |
| Financial inclusion | Basic banking guidance, literacy and account support |
The project is an example of AI design that starts from infrastructural realities rather than idealised user scenarios. For South African technology planners and NGOs, the approach offers a reminder that language diversity, cost sensitivity and offline access remain central to meaningful digital inclusion.
