Most senior executives in South Africa who used artificial intelligence last year reported that it had improved the quality of their work, but many cannot explain how the systems reached those conclusions, an academic has warned. The finding raises questions about whether organisations have the governance, skills and critical evaluation capacity to rely on AI for important strategic and financial decisions.
Survey numbers and a management problem
Dr Shaheen Khan, a senior academic and programme co‑ordinator at a South African business school, wrote that 91% of South African senior executives who used AI in the previous year said it improved their work, citing PwC’s Africa Workforce Hopes and Fears Survey (2025). The article also noted that 64% of employees across Africa reported using AI at work, above the global average of 54%.
However, Khan argues the rapid uptake of AI has outpaced development of AI literacy, governance and critical evaluation skills. That mismatch, the article says, turns what might appear to be a technology problem into a management problem: executives are making operational and strategic choices based on AI recommendations without fully understanding how those recommendations were produced and where they might be faulty.
Why adoption has been so fast
The Technology Acceptance Model (TAM), developed by Davis (1989) and extended by later researchers, offers a framework for the rapid adoption. TAM proposes people adopt technologies more readily when they perceive them to be useful and easy to use. Subsequent extensions by Venkatesh and Davis (2000) and Venkatesh and Bala (2008) showed organisational context, social influence, experience and facilitating conditions also shape behavioural intention to adopt technology.
According to Khan, these theoretical explanations help account for why AI tools — many of which promise quick productivity gains — have been embraced across management layers, even where deep technical understanding is lacking.
Practical gaps and organisational risk
The article highlights specific gaps that increase risk when AI is used for decision-making:
- Limited understanding among managers of how AI models generate recommendations and where they fail.
- Insufficient governance structures to ensure AI outputs are accurate, ethical and accountable.
- Potential overreliance on AI without adequate human oversight or evaluation frameworks.
Those gaps matter because, Khan writes, public debate often focuses on whether AI will replace jobs. While that is important, it can distract from more immediate issues: businesses making financial, strategic and operational decisions on the basis of AI results that may be flawed, biased or contextually inappropriate.
What this means for South African organisations
The piece implies several practical consequences for South African companies and public bodies that are relying on AI:
- Board-level and managerial capability must include not only awareness of AI tools but the ability to critique and validate their outputs.
- Organisations should invest in governance frameworks that set standards for testing, auditing and explaining AI decisions.
- Training and experience — not only vendor sales pitches — are needed to build confidence that AI recommendations are reliable in local operating conditions.
Without these steps, the article warns, the benefits reported by executives may be fragile: perceived productivity gains could mask systemic errors that only surface when decisions go wrong.
Data points at a glance
| Measure | Figure |
|---|---|
| Senior executives in South Africa who said AI improved their work | 91% |
| Employees across Africa using AI at work | 64% |
| Global average of employees using AI | 54% |
The original analysis is by Dr Shaheen Khan, who places the emphasis on improving managerial competencies and governance rather than treating AI adoption as merely a technical rollout. For South African firms operating on tight budgets and in complex regulatory and socio‑economic environments, that means the next phase of AI adoption should focus on transparency, auditability and building critical skills — not only on deploying the latest models.
Reporting by the Technology desk
