A recent conference of the Institute of Internal Auditors in Flic-en-Flac, Mauritius, highlighted the challenges organizations face in implementing Artificial Intelligence (AI) at scale. The conference speaker, Marc Israel, noted that not a single hand went up when he asked if any organization in the room was running at least one AI pilot. This response was surprising, given that AI adoption is increasing globally. A McKinsey survey published in November 2025 found that 88% of organizations use AI regularly in at least one function, up from 78% the previous year.
The McKinsey survey, which gathered responses from nearly 2,000 organizations in 105 countries, also found that only about 6% of organizations are considered high performers, where AI contributes at least 5% of operating profit. Similarly, a Boston Consulting Group (BCG) survey of 1,330 senior executives found that only 7.5% of organizations have achieved advanced AI adoption. These findings suggest that while many organizations are experimenting with AI, few have successfully scaled its implementation.
One of the main challenges organizations face in scaling AI is governance. BCG's survey found that nearly half of companies now get some real value from AI, but the binding constraint is no longer whether AI creates value, but whether companies can control and govern it. A common scenario, as described by Israel, is that an organization's board will ask questions about AI implementation, such as who owns the model, how to explain decisions, and how to detect drift. However, these questions often go unanswered, leading to delays or abandonment of AI projects.
Israel notes that boards rarely say no to AI, but rather ask for more information, leading to repeated delays. In one example, a bank built a model to pre-assess SME loan applications, but the board risk committee asked questions that couldn't be answered, and the project stalled. The failure was not in the board's questions, but in not building the necessary answers alongside the model. This highlights the need for organizations to develop effective governance and assurance processes for AI.
Effective governance and assurance processes for AI involve oversight, rollback, audit trails, and cost limits. However, BCG's survey found that only 5% of organizations have the full set of controls in place, while 42% plan to give AI agents genuine decision-making authority by 2030. This gap in governance and control is concerning, as organizations increasingly rely on AI for decision-making.
Another blind spot in AI governance is the lack of visibility into AI spending outside the enterprise IT budget. BCG found that more than 80% of AI spending now sits outside the IT budget, with marketing, HR, and claims teams buying AI on credit cards. This makes it difficult for boards to review AI through the IT budget alone. Israel suggests that organizations should ask three questions before the next AI budget cycle: which AI systems shape decisions in the organization, and who owns each one; what evidence is needed to approve scaling an AI project; and how quickly would the organization know if an AI system got it wrong.
The importance of effective AI governance and assurance cannot be overstated. As AI adoption continues to grow, organizations must prioritize developing the necessary controls and processes to ensure AI is used responsibly and effectively. By doing so, organizations can unlock the full potential of AI and achieve greater value from its implementation.
Key points
- Most organizations globally, including in Mauritius, struggle with governance and scaling AI effectively.
- Effective AI governance involves oversight, rollback, audit trails, and cost limits, but few organizations have these controls in place.
- Organizations must prioritize developing AI governance and assurance processes to ensure responsible and effective AI use.
