Artificial intelligence (AI) is gaining momentum among small and medium enterprises (SMEs) in South Africa. Businesses are adopting and developing AI to improve efficiency, expand services and reach new markets. However, many SMEs face persistent barriers related to how data is accessed, managed and governed. Access to reliable, well-governed data remains uneven, and weak data practices are emerging as a major constraint on scalable and responsible AI adoption.
The importance of addressing these barriers is increasing as AI systems become more integrated into SME products, services and operations. Gaps in data quality, oversight and accountability create significant risks, including unreliable outputs, biased outcomes and loss of trust in AI systems. Strengthening data governance is becoming essential to support compliance, competitiveness and sustainable participation in South Africa’s digital economy. It also enables AI solutions to deliver positive social and economic impact.
AI adoption among SMEs in South Africa
AI is increasingly recognised as a driver of productivity and innovation across low and middle income countries, with South Africa emerging as a regional leader. For SMEs, AI creates opportunities to reduce costs, improve operational efficiency, expand market reach and develop new products and services.
Despite this potential, AI adoption among SMEs in South Africa remains uneven. Many businesses are still at an early stage due to skills gaps, infrastructure constraints and challenges related to data access and quality. These barriers reinforce the importance of strengthening data governance as a foundation for scalable and responsible AI adoption.
This variation in adoption also reflects the diversity of SMEs themselves. Firms engage with AI in different ways depending on their digital maturity, capabilities and role in the AI value chain. Rather than following a single, linear pathway, SMEs tend to cluster into distinct modes of AI engagement, ranging from the use of AI-enabled tools to the development of AI-driven products and services. Recognising this diversity is important for designing policies and support mechanisms that address the needs of different SME segments.
South Africa’s Draft National Artificial Intelligence Policy Framework (2024) reflects a national commitment to using AI to support economic growth and social well-being through a human-centred and ethical approach. Achieving these ambitions will require enabling SMEs across different modes of AI engagement and strengthening data governance as a foundation for responsible adoption and innovation.
Understanding data governance for AI
Data governance for AI refers to the policies, processes and controls that determine how data is collected, managed, shared and used throughout the AI lifecycle, from model development to deployment and monitoring. It helps ensure AI systems are reliable, secure and aligned with legal and ethical requirements. Weak governance can undermine performance, amplify bias and reduce trust in AI-driven outcomes.
AI introduces governance challenges beyond traditional data management. Systems are highly sensitive to the quality and representativeness of training data and can create risks of biased or unfair outcomes. They also increase expectations around transparency, explainability and accountability in automated decisions.
For SMEs, these challenges are often intensified by limited resources and reliance on third-party platforms, cloud services and off-the-shelf tools. This can reduce visibility over data processing and limit control over risks. Strong data governance helps ensure data quality, protect sensitive information and support accountability. It can also reduce compliance risks, improve AI outcomes and build confidence among customers and partners, forming a foundation for responsible and scalable AI adoption.
Why data governance for AI matters now
As SMEs move from experimenting with AI to integrating it into products, services and decision-making processes, the importance of strong data governance is increasing. Responsible AI depends on reliable, well-managed data. Weak governance can lead to biased outcomes, privacy risks and unreliable decisions, particularly when AI systems are deployed at scale or used in customer-facing contexts.
Regulatory and market expectations are also evolving. South Africa’s Protection of Personal Information Act (POPIA) establishes requirements for lawful data processing, accountability and security safeguards that are directly relevant to AI systems. Provisions related to automated decision-making and cross-border data transfers create new compliance considerations for SMEs adopting or developing AI.
At the same time, broader national policy developments in South Africa, including the National Policy on Data and Cloud and national digital transformation initiatives associated with the Presidential Commission on the Fourth Industrial Revolution, are signalling stronger expectations for ethical and responsible AI development and use.
For SMEs, navigating these requirements can be challenging due to limited resources and reliance on external platforms and service providers. However, embedding robust data governance early can help manage compliance risks, build trust with customers and partners, and support credible participation in digital value chains. Strong governance also helps ensure AI solutions deliver positive economic and social outcomes.
Data challenges shaping AI development and adoption
AI adoption continues to grow, and results in data-related challenges becoming a central barrier for SMEs seeking to deploy AI responsibly and at scale. These challenges relate to data access, data quality, dataset bias and the availability of local-language data which is shaping how data governance for AI is emerging and being adopted.
Access to data
Access to data is a foundational enabler for AI, however it remains uneven across South Africa’s ecosystem. Public initiatives such as Open Data South Africa and Statistics South Africa are expanding access to government and public-interest datasets. The GIZ FAIR Forward AI Maturity Assessment Framework, demonstrates that South Africa continues to show low maturity especially for data availability for SMEs. A low maturity level signifies foundational challenges which are characterised by limited data availability, weak infrastructure, lack of innovation/R&D, and challenges in implementing responsible AI practices. This indicates a significant gap in national AI readiness and highlighting the need for inclusive and ethical AI adoption to avoid digital divides.
Data quality
Even where data is accessible, challenges related to completeness, accuracy and timeliness persist. Poor quality data undermines AI performance and limits the viability of solutions in particular sectors such as healthcare, agriculture and disaster preparedness, where reliable insights are critical.
Biases in datasets
Data quality challenges are closely linked to bias and representativity. For example, UNESCO’s AI Readiness Assessment for South Africa highlights limited access to datasets that adequately reflect women, minority and marginalised groups, increasing the risk that AI systems reinforce existing inequalities.
Local language datasets
The shortage of local language datasets further limits AI relevance and trust. Many AI systems are trained on non-local data, which does not reflect South Africa’s linguistic diversity. Initiatives such as Lelapa.AI are addressing this gap by developing African language models that enable more inclusive voice and text-based AI applications.
Limited access to high-value datasets and data-sharing mechanisms
High-value datasets are often concentrated within large corporations or public institutions. Restrictive licensing, unclear data-sharing frameworks and trust concerns limit SME and startup access, reducing opportunities for collaboration and innovation.
Skills, capacity and organisational gaps
These challenges are further intensified by persistent gaps in skills and organisational capacity. Many SMEs lack data governance expertise, AI literacy and formal structures to manage data responsibly, therefore resulting in fragmented and reactive data practices.
Data governance is increasingly foundational for SMEs in South Africa seeking to adopt AI responsibly and at scale. This blog outlines key data-related challenges, but there remains a lack of SME level evidence on how data is accessed, governed and regulated in practice. In addition, what governance approaches would be most enabling.
To address this gap, GSMA Mobile for Development’s Central Insights Unit is undertaking new research on data governance for AI among SMEs in South Africa, to generate practical evidence and recommendations to inform policy and ecosystem action, which will be publicly available mid 2026.
To learn more about this research or to contribute, please get in touch with us at centralinsights@gsma.com.
This project is funded by the UK Foreign, Commonwealth & Development Office and supported by the GSMA and its members.
