Asher's Blog

For years, surveillance in business meant that a person watched a dashboard, reviewed logs, responded to an alert or examined transactions after something had already gone wrong. The company may have collected a great deal of information, but attention remained scarce. A warning could exist inside the system for hours because the right person had not yet seen it or understood what it meant.

AI changes that model. The new kind of surveillance is not primarily about watching employees; it is about allowing intelligent systems to watch the infrastructure on which the company and its customers depend. An AI system can compare normal and unusual behaviour, connect signals from several services, inspect transactions continuously and bring the most important exception to a human before it becomes a crisis.

This is especially useful in financial services, where the cost of delayed attention can be enormous. A manual team can review alerts and reconciliations, but its capacity is limited by time and fatigue. An intelligent system can notice that a withdrawal pattern has changed, that a privileged account is acting unusually, that a deployment altered a critical control or that several apparently unrelated events form one suspicious sequence.

The same principle applies beyond finance. AI can monitor energy equipment for signs of failure, examine road or bridge data for deterioration, help hospitals identify unusual supply shortages and watch digital infrastructure for attacks. It does not eliminate the need for engineers, auditors or security professionals. It helps them direct attention to the place where judgment is most urgently needed.

In our own work, I am comfortable allowing intelligent agents to analyse, propose and prepare small improvements quickly. What I am not prepared to do is give an agent unrestricted authority over a production system. A human must approve a release or any important change, and an agent that attempts to bypass that boundary should immediately lose access. Intelligence is useful, but authority must remain deliberately designed.

That distinction matters because AI can be confidently wrong. It can misunderstand a pattern, inherit a bad rule or recommend an action without appreciating its wider consequences. Bad actors will also use intelligent tools, and no responsible company should pretend that adopting AI makes it impossible to be hacked or defrauded. The purpose is not to create an illusion of perfect security. It is to notice more, respond earlier and learn faster.

Good AI surveillance therefore needs several controls. The system should have the minimum access required for its work, its actions and recommendations should be recorded, and critical decisions should require human approval. Teams should test what happens when the model is unavailable or mistaken, and a person must be able to stop the process without negotiating with the technology. Surveillance without accountability merely creates a more sophisticated blind spot.

Privacy must also be part of the design. A company does not need to collect every piece of information simply because AI can analyse it. Monitoring should be connected to a legitimate risk, limited to relevant data and governed by clear retention and access rules. Protecting a system cannot become an excuse for treating customers or employees as people with no right to dignity.

The opportunity for African companies is significant because many monitoring processes are still manual. We do not need to wait for every institution to develop a large security operations centre before improving its capacity to detect risk. Well-designed intelligent tools can make sophisticated monitoring available to smaller banks, hospitals, energy providers, logistics companies and public institutions.

The new surveillance is continuous, intelligent and selective. It does not replace the last responsible human; it makes that person better informed. The companies that use it well will not be those that hand over every decision to machines. They will be the companies that combine machine attention with human judgment, clear limits and the humility to know that every control must keep improving.