As artificial intelligence moves deeper into critical business decisions, organisations face a new governance challenge: proving that innovation is being matched by accountability, transparency and responsible oversight.
Seen Here: Professor Yudhvir Seetharam, Chief Analytics Officer for FNB Commercial, is challenging businesses to rethink how they approach artificial intelligence — warning that as AI becomes embedded in critical business decisions, its risks can no longer be treated as purely technological. From environmental impact and algorithmic bias to accountability, transparency and human oversight, Seetharam argues that AI governance is fast becoming a defining test of ESG maturity — and responsible innovation will be just as important as technological ambition. Photo Credit: Supplied
Artificial intelligence is no longer sitting on the experimental fringes of corporate strategy. It is rapidly becoming part of the infrastructure of modern business, influencing everything from customer service and credit decisions to recruitment, compliance, operations and strategic planning.
But as AI becomes more deeply embedded in organisations, a new question is emerging for boards and executives: Are businesses treating AI risk with the same seriousness as their broader environmental, social and governance (ESG) responsibilities?
Prof Yudhvir Seetharam, Chief Analytics Officer at FNB Business, argues that AI governance can no longer be viewed simply as a technology issue.
The central challenge, he says, is understanding the broader consequences of AI-powered decision-making — and recognising that those consequences have environmental, social and governance dimensions.
AI’s hidden environmental footprint
The perception of AI as an intangible digital technology can obscure the physical infrastructure required to operate it.
Behind every AI-powered service are data centres, processors, electricity, cooling systems and vast quantities of stored data. As businesses scale their use of AI, the environmental cost associated with that infrastructure becomes increasingly relevant.
For companies pursuing ambitious ESG targets, this creates a potential blind spot.
An organisation may achieve greater operational efficiency through AI while simultaneously increasing its consumption of computing resources and energy. Without measuring both sides of the equation, businesses risk celebrating efficiency gains without understanding their full environmental impact.
When algorithms affect people’s lives
The social implications of AI may be even more immediate.
AI systems are increasingly involved in decisions that can materially affect individuals — from determining whether someone qualifies for credit or insurance to influencing recruitment outcomes, identifying potentially fraudulent transactions or determining how customers are prioritised.
When these systems produce biased, inaccurate or poorly understood outcomes, the consequences extend well beyond technology.
For the customer who is denied credit, the job applicant who is screened out or the vulnerable customer whose circumstances are misunderstood, an algorithmic decision can have very real consequences.
This makes trust a fundamental component of responsible AI governance.
According to Seetharam, trust cannot be established simply by demonstrating that data is clean or that an AI model is statistically accurate. People also need to understand how significant decisions are reached, have mechanisms to challenge outcomes where appropriate and know who ultimately carries responsibility.
AI may make a decision, but it cannot assume accountability for that decision.
That responsibility remains with the organisation and its leadership.
Governance must move faster than AI
One of the biggest risks facing businesses is the growing gap between the speed at which AI can be deployed and the speed at which governance structures can respond.
A conventional human-led process may involve hundreds of decisions over a given period. An AI-enabled system can potentially make thousands of decisions in seconds.
If controls remain designed for the pace of traditional business processes, they can quickly become inadequate.
Having an AI policy or governance framework on paper is therefore not enough.
Effective governance requires clearly defined decision rights, identifiable accountability and controls that operate at the same scale and speed as the technology they are intended to oversee.
In this environment, governance cannot simply be a compliance exercise. It must become an operational capability.
Building ‘friction by design’
Rather than attempting to slow technological innovation, businesses should consider introducing what Seetharam describes as “friction by design” — deliberate points of human intervention where AI-driven decisions could have material consequences.
Not every AI application requires the same level of oversight. Low-risk automation can often operate with limited intervention, while decisions involving significant financial, employment, customer or reputational consequences warrant stronger controls.
A rejected insurance claim, credit decision, hiring recommendation, fraud alert or assessment of a customer’s vulnerability should not simply disappear into an opaque algorithmic process.
There should be a clearly identified owner, an appropriate decision-making and escalation process, and an auditable record of what happened.
This approach is not a rejection of technology. Instead, it represents a more mature form of risk management that can ultimately give businesses greater confidence to deploy AI.
The value equation is changing
The business case for AI has traditionally centred on productivity, efficiency, cost reduction and revenue growth.
Those benefits remain important. But as AI becomes embedded in core business processes, companies increasingly need to consider the full value equation.
AI can create value while simultaneously eroding value if it improves efficiency but weakens accountability, reduces costs while creating hidden governance expenses, or accelerates decision-making while increasing reputational and regulatory exposure.
The real cost of poorly governed AI may only emerge later through customer remediation, operational rework, regulatory intervention, reputational damage or loss of trust.
For boards, this means asking more sophisticated questions about AI.
Where is the organisation using AI? Which decisions does it influence? How material are those decisions? What controls are in place? Where is human judgement required? Who is accountable when the system gets it wrong?
From AI ambition to AI accountability
The growing prominence of AI is effectively bringing technology governance into the heart of the ESG conversation.
For companies, the next stage of AI maturity will therefore not simply be measured by how many systems they deploy or how quickly they adopt the technology.
It will increasingly be measured by whether they can explain, govern and take responsibility for how AI is being used.
The organisations best positioned to lead in the AI economy may ultimately be those that understand that innovation and accountability are not opposing forces.
They are complementary.
The race for AI leadership is no longer simply about who can deploy the technology fastest. It is about who can scale it responsibly — understanding its environmental and social impact, maintaining meaningful human oversight and ensuring that someone remains accountable when the algorithm gets it wrong.
That is where AI risk becomes ESG risk — and where responsible governance becomes a competitive advantage.
