Talal Abu-Ghazaleh
Technology trends typically come in waves, and the most current wave is the autonomous AI systems that help business execute tasks. This is called ‘agentic AI’, where AI’s are given agency to perform tasks by accessing corporate systems that businesses use, within a set boundary of rules. While this is an exciting development, the underlying technology is still in its infancy, coming with its own set of challenges.
Just recently, there was the example of a software firm in the US that serves private car rental businesses, membership clubs, and independent sales representatives that while in the process of deploying such systems it caused serious damage. A single AI agent erased a production database and its backups in nine seconds, which was clearly a failure of governance.
Such systems now participate in software development, financial processes, logistics, security workflows and administrative functions. They act, decide and execute without waiting for human confirmation. This creates a new category of institutional risk, where it is no longer enough to regulate data privacy or algorithmic fairness. The autonomous behavior itself must be governed, without which these systems can compromise years of intellectual capital and disrupt essential services in moments.
My experience with the United Nations and the World Trade Organization has shown me that technological progress must be matched with responsible governance. The global economy depends on trust in systems, continuity of operations and the protection of assets:  Agentic AI challenges all of them. It introduces a level of operational autonomy that existing frameworks were never designed to manage. This is why we need to define how autonomous systems behave, what they are allowed to do and what boundaries they cannot cross.
These standards must be clear and enforceable, including non-negotiable safety rules, transparent audit trails and ethical constraints that cannot be overridden by the systems themselves. They must also ensure that human supervisors retain the authority to intervene at any moment so that firms can protect their business assets from avoidable risks created by rogue automation, and prevention measures in place within the systems they access. Here, a layered governance approach is a useful methodology to adopt.
The transition to an AI-driven global economy is happening right now. The question is whether we will enter it with strength or with fragility. Innovation without governance creates instability and governance without innovation creates stagnation. Sustainable progress requires both. This is a warning, but it is also an opportunity. It shows us where the gaps are, what must be corrected and that we must exercise great caution when deploying such autonomous systems. My advice is for the business world to implement AI systems with enthusiasm, but with diligence and detailed preparation.
I would urge business executives to treat the governance of agentic AI governance as essential infrastructure. It is as important as cybersecurity, financial oversight and any other business activity. AI has the power to transform, but only if businesses act with discipline and foresight. If this is left as an afterthought, we risk allowing a transformative technology to undermine the very business activities it is meant to advance.