AI is no longer just reshaping industries — it’s rewriting the rules of business. Competitive advantage now depends on how well companies integrate AI into their operations. Yet while AI adoption accelerates, governance remains an afterthought.
Organisations are still figuring out what AI means for strategy, risk, ethics and workforce transformation. Unlike cybersecurity or financial oversight — domains with structured governance — AI is evolving too quickly for traditional models to keep up. The trajectory is clear: from Narrow AI (task-specific systems such as fraud detection) to General AI, where models reason, adapt and self-improve across domains. Super AI, where systems surpass human intelligence, looms on the horizon. The game is no longer about scale, but speed, adaptability, and control.
Companies such as DeepSeek-V2 and Mistral AI are proving that huge computational power is no longer a prerequisite for AI dominance. DeepSeek developed its AI model for just $6m — a fraction of the $100m-plus spent on models such as ChatGPT and Gemini — by shifting towards smaller, highly optimised models instead of monolithic ones.
What happens when AI-driven strategies outperform human intuition? Who takes responsibility when AI makes a costly mistake? More importantly, how do companies govern something they don’t fully understand?
AI isn’t perfect. Hallucinations — its tendency to generate false information — pose big risks, particularly in finance, legal and healthcare industries. A misguided AI-generated market forecast could cost millions. There’s also an emerging governance gap. Many assume AI oversight belongs to IT or compliance teams, but that’s a narrow approach. AI governance must sit at the intersection of risk management, strategy, and execution — not just to prevent downside risks, but to unlock upside potential.
Risk and compliance teams should not just be enforcers but enablers. Instead of banning AI outright, they should provide structured frameworks that balance innovation with security. AI is already transforming industries. In education, AI now provides real-time tutoring and personalised learning at scale. A cleaning company could implement AI-powered robotics and predictive maintenance to optimise operations. The lesson? Companies that don’t redesign their processes around AI will wake up obsolete.
Intelligence ceiling
Despite its rapid progress AI has hard limitations. Shannon’s Information Theory quantifies this: the formula shows that AI can only extract knowledge up to a certain entropy limit, meaning at some point there simply isn’t enough structured information left for AI to improve beyond incremental refinements. That’s why AI development is shifting from large general-purpose models to smaller, domain-specific AI models.
DeepSeek’s $6m model outperforms systems costing 16 times more, and Mistral AI’s compact models rival those requiring enormous server farms. The takeaway? AI governance must prioritise adaptability and precision, not brute force scalability.
AI adoption isn’t optional, but how companies integrate AI will determine their survival. Forward-thinking organisations build resilience by ensuring they can function if AI systems fail. Take Discovery Health, SA’s largest private health insurer. Its AI system processes 92% of claims automatically, but human oversight remains for complex cases. When AI flagged unusual patterns during the pandemic, Discovery adjusted its models in real time rather than suffering disruptions. The most resilient companies don’t just implement AI, they design AI-dependent processes strategically.
Leadership is no longer just about managing people — it’s about orchestrating AI-human collaboration. AI governance starts by mapping core business processes to identify where AI can optimise or automate tasks. Organisations must redesign workflows to integrate AI-driven efficiencies and predictive analytics while defining the human role in AI-augmented operations.
This isn’t theoretical. The SA Revenue Service faced a budget cut from R1bn to a few hundred million rand, forcing it to turn to AI for fraud detection and tax collection. The result? R210bn recovered in undisputed tax debt within 11 months and R21bn in fraudulent tax claims blocked, despite a reduced workforce. The lesson? AI isn’t just about cost-cutting — it amplifies human decision-making at scale.
New roles for leadership
AI is redefining leadership at all levels, from frontline managers to board directors. Leaders must:
- Ensure AI literacy across teams — if employees don’t understand AI, they won’t use it effectively.
- Redefine KPIs — success is no longer about individual output but how well humans and AI collaborate.
- Develop uniquely human skills — creativity, emotional intelligence and ethical judgment are now a leader’s most valuable assets.
Put simply, AI takes care of routine tasks — leaders take care of everything AI can’t. AI readiness needs to be embedded at every level of an organisation. If AI is moving this fast, can leadership confidently answer these questions?
- What does AI mean for our company in the next three years — not 10, but three?
- How will AI redefine customer engagement, decision-making and operations?
- Are our employees prepared for this shift?
Governance must balance oversight and innovation. AI-driven predictive maintenance could save millions, but restrictive compliance could kill innovation before it starts. Instead of blocking AI adoption, risk teams should develop structured, clear policies for AI use.
Without a structured AI framework companies will either become too restrictive and fall behind, or too reckless and suffer the consequences. The question isn’t whether businesses will use AI. The companies that survive and thrive will be the ones that govern AI rather than react to it.
• Mafinyani is risk advisory & financial modelling partner at DiSeFu, a specialised financial technology and risk advisory firm operating in the Sub-Saharan Africa region.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.