AI Expo Africa 2025: Reflections from the Way With Words Team

Attending AI Expo Africa in Johannesburg in 2025 was a valuable experience for the Way With Words team. We joined as visitors, spending most of our time exploring conversations, meeting new people, and listening closely to how the African AI landscape is evolving — especially around speech data. One idea kept coming up again and again: the future of AI on the continent depends on locally sourced, context-rich data.

Throughout the event, we spoke with startups, enterprise teams, and researchers who were building exciting AI solutions. Many of them had strong models and clear ambitions, but there was a common challenge beneath the surface — access to reliable, representative speech datasets. It became clear to us that while model development is accelerating, high-quality African speech data is still a critical missing piece for many organisations.

Local Language Representation

One of the strongest themes we encountered was the need for African languages to be properly represented in AI systems. Conversations highlighted how linguistic diversity should be treated as a foundation rather than an afterthought.

There was a shared acknowledgement that many global models still struggle with African accents, dialects, and multilingual environments. As a result, interest in well-annotated, carefully curated speech datasets is growing rapidly. We heard more than once that localisation starts with data — not just algorithms — and that insight resonated strongly with our work.

Data Quality Over Data Quantity

Another shift we noticed was the move away from focusing purely on volume. Instead of simply collecting more recordings, people were talking about collecting better data.

Balanced accents, varied recording environments, and ethical sourcing came up frequently. Enterprise buyers were asking thoughtful questions about quality control, validation processes, and the role of detailed metadata in improving speech technologies. For us, it felt like a sign that the market is maturing and beginning to value depth and reliability over scale alone.

Responsible & Ethical Data Collection

Responsible and ethical data practices were also a central part of many discussions. We saw genuine interest in building transparent consent frameworks, ensuring contributors are fairly compensated, and protecting the sovereignty of African datasets.

Because speech data carries biometric characteristics tied to identity, there was a strong awareness that it needs to be handled with care. These conversations reflected a growing sense of responsibility across the ecosystem — something we believe is essential for sustainable growth.

Looking Back

Looking back, the expo felt less like a traditional tech event and more like a meeting point for people trying to solve similar challenges from different angles. For the Way With Words team, it reinforced the idea that speech data is no longer a background component of AI — it’s becoming a central building block.

We left with a sense that collaboration across the African AI community is deepening, and that the path forward will be shaped by organisations willing to invest in high-quality, locally rooted datasets.