The integration of Artificial Intelligence in enterprise strategy often sparks debate about the level of autonomy machines should have. A more pressing concern for business leaders is determining where human judgment is crucial. This has led to the concept of human-orchestrated AI, where machines operate at speed and scale, while humans set objectives, boundaries, and accountability. In emerging markets, particularly across the Middle East and Africa, finding the right balance is vital. AI can address various constraints, including skills, service delivery, and financial inclusion.
A risk-based approach to AI deployment is recommended, rather than delaying adoption due to risks or asking governance questions after deployment. A useful principle for leaders is that the greater the consequence of an AI decision, the greater the required human oversight. This leads to practical levels of decision-making: automating low-consequence decisions, supervising material consequence decisions, and having humans make high-consequence decisions. For instance, routine and easily reversible decisions can be automated, while decisions affecting finances, employment, or healthcare require human judgment and accountability.
At the material consequence level, AI can perform much of the work, but human supervision and escalation mechanisms are necessary. In contrast, high-consequence decisions require human judgment and accountability, with AI only informing or supporting the decision. The difference in oversight is crucial, as seen in AI systems categorizing internal documents versus influencing credit decisions or healthcare recommendations. The human role is increasingly focused on designing the loop, establishing thresholds for AI autonomy, and intervening when necessary.
Human orchestration of AI also involves providing context, particularly in emerging markets. AI systems are shaped by data and assumptions, which may not accurately represent local languages, cultures, and socioeconomic conditions. The quality of underlying data is essential, and human oversight cannot compensate for poor information. Organisations must understand the data used by their AI systems, its representativeness, and potential biases. In emerging markets, these questions should be addressed from the outset to ensure effective AI deployment.
Emerging markets should not adopt oversight models developed elsewhere without modification. Instead, they should apply global responsible AI principles in ways that reflect local economic, cultural, and regulatory realities. Local experts should be able to recognise when AI systems miss linguistic or cultural nuances or rely on assumptions that do not translate well across markets. This approach enables the combination of global capability with local intelligence, ensuring AI systems are effective and responsible.
Accountability in AI is a critical aspect that cannot be delegated to an algorithm. As AI becomes more autonomous, organisations need clear answers to basic questions about authorisation, boundaries, monitoring, and responsibility. Good governance can actually facilitate innovation by providing clear expectations about accountability, fairness, and transparency. By being more deliberate about AI autonomy, human judgment, and accountability, emerging markets can adopt AI responsibly and rapidly.
According to Alan Turnley-Jones, CEO of NTT Data Middle East and Africa, responsible AI adoption is not a brake on innovation. Instead, it can give organisations greater confidence in their ability to innovate within established boundaries. By applying responsible AI principles and adapting them to local contexts, emerging markets can harness the benefits of AI while minimising risks. This approach enables the development of smarter oversight models that balance machine autonomy with human judgment and accountability.
Key points
- Emerging markets require a smarter model of AI oversight that balances machine autonomy with human judgment and accountability.
- A risk-based approach to AI deployment is recommended, with human oversight increasing in line with the consequences of AI decisions.
- Accountability in AI cannot be delegated to an algorithm and must be ensured through clear expectations and governance.
