Artificial intelligence has moved from experiment to infrastructure. It is being built into customer service, credit processes, recruitment, operations, compliance, and strategy. In many businesses, AI is already shaping decisions that affect costs, customers, employees, and trust.
That is why the governance conversation needs to widen. Too many organisations still treat AI risk as a specialist technology issue. They ask whether the model works, whether the data is good enough, whether the system can improve productivity. These are useful questions, but they are not enough. The bigger, more important issue is whether the organisation fully understands the impact of AI on the business and on the society in which it operates. In other words, has it grasped that AI risk is ESG risk.
Every AI decision has an environmental, social and governance footprint. The environmental footprint is often hidden because AI is sold as digital, weightless and efficient. But AI is physically intensive. It depends on data centres, processing power, cooling, electricity, and storage. A business that celebrates AI-driven efficiency without measuring the energy cost of that efficiency is only seeing half the balance sheet.
The social footprint is just as important. AI systems influence how people are treated. They can shape who receives a loan, who gets shortlisted for a job, which customer gets priority and which complaint is escalated. Bias and unfair outcomes become real when a customer cannot challenge a decision, an employee can’t understand a process, or a vulnerable person is pushed into an outcome that was never properly reviewed by a human.
This is why trust is a central issue in effective AI governance. Trust is not achieved simply because the data is clean or the model is accurate. It is built when people can understand a decision, challenge it where necessary and know that someone is accountable for the outcome. AI can optimise processes, but it cannot carry responsibility. That responsibility still belongs to the organisation, its leaders and its governance structures.
This is also where many companies are exposed. They may have an AI strategy, policies and even a governance framework. But a framework that is not properly operationalised is not governance, it is merely documentation. Real governance means clear decision rights, named accountability and controls that work at the speed at which AI is being deployed. The danger, of course, is that AI scales faster than oversight. A human team may make hundreds of decisions a week. An AI-enabled process can make thousands in seconds. If the controls around that process don’t scale at the same pace, the organisation is not managing AI risk, it is creating it.
All of this means that boards and executives need to stop focusing solely on whether AI creates value and look more closely at whether their controls are keeping up with their AI aspirations. That means knowing where AI is being used, what decisions it supports, what level of risk those decisions carry and where human oversight is required.
The idea isn’t to simply try and slow down innovation; it is to introduce friction by design. This means placing deliberate human checkpoints where the consequences of an AI-enabled decision are material. Low-risk automation may not need heavy intervention, but medium- and high-risk decisions should never be left to run autonomously without review, escalation, and accountability.
A rejected insurance claim, a credit decision, a hiring recommendation, a fraud flag, or a customer vulnerability assessment should not disappear into an AI black box. For all these decisions and actions, there must be a named owner, a decision-making body and process, and a clear audit trail. This is not anti-technology; it is mature risk management. Good controls protect AI value and allow organisations to use it confidently because they can explain how decisions are made, who is responsible and what happens when the system is wrong.
The bottom line is that if AI is not appropriately governed, it can create value and erode value at the same time. This happens when, for example, it improves efficiency while weakening accountability or reduces costs in one part of the business while creating hidden costs elsewhere due to rework or governance, talent, and reputational damage.
The fact that AI has moved from the edge of business to being central to the decisions that shape value, risk, trust, and accountability means it now sits firmly inside the ESG agenda. That means the race for AI leadership is no longer about who can deploy AI everywhere the quickest, but rather who can show where they are using it, how they are managing its environmental and social impact, who is accountable, how they are ensuring human judgement remains in control.
