South African Businesses: Data Estate Quality Trumps AI Model Sophistication

• Organisations in South Africa are struggling to achieve business value from AI due to poor-quality data.
• The maturity of the data estate is a more critical determinant of success than the sophistication of the AI model.
• Data quality, governance, and accessibility are the primary causes of failure for AI implementation worldwide.
• Organisations must invest in data readiness before investing in AI experimentation to unlock business value.

The result, he says, is inaccurate outputs, contradictory recommendations, poor decisions and users who quietly stop trusting the system, resulting in stalled adoption and the evaporation of business value.

Organisations in South Africa are facing a harsh reality: despite investing heavily in AI models, many are struggling to achieve tangible business value. According to Gunther Wucherpfennig, Executive: Data Analytics and AI at Mint Group, the root cause of this failure lies not with the sophistication of the model, but rather with the quality and governance of the underlying data estate. This is a stark contrast to the hype surrounding AI adoption, which often focuses on the potential benefits of advanced analytics and machine learning. In reality, Wucherpfennig notes that organisations are discovering that the real determinant of success is not the AI model itself, but rather the maturity of the data estate beneath it.

The impact of poor-quality data on AI adoption cannot be overstated. When organisations deploy AI on top of incomplete, duplicated, outdated or inconsistent data, the results are often disastrous – inaccurate outputs, contradictory recommendations, and users who quietly stop trusting the system. This can lead to stalled adoption and the evaporation of business value, resulting in a significant waste of resources for organisations that have invested heavily in AI initiatives. Wucherpfennig's research suggests that global success rates for AI implementation are around 40%, with data quality, governance, and accessibility being the primary causes of failure.

While there is no specific legislation governing data estates in South Africa, organisations must still comply with various regulations related to data protection and governance. The Protection of Personal Information Act (POPIA) requires businesses to ensure the accuracy, completeness, and relevance of personal information they collect and process. This includes implementing measures to prevent data duplication, ensuring that data is up-to-date, and providing mechanisms for individuals to access and correct their personal information. Organisations must also consider the broader implications of AI adoption on data governance, including issues related to bias, transparency, and accountability.

Organisations in South Africa should start with an honest AI-readiness assessment, focusing on data readiness and business readiness to unlock the real promise of AI, rather than investing solely in AI experimentation.

Source: Original reporting via Mint Group