As artificial intelligence transitions from passive tools to "agentic" systems—AI capable of autonomously breaking down complex goals into actionable steps—the conversation around safety is shifting. We are no longer just asking if a model can write an essay; we are asking if it can manage a budget, execute trade agreements, or modify infrastructure without causing unintended systemic failures. This autonomy, while powerful, creates significant operational risk if left unchecked.

Benjamin Palacio, an IT strategist, argues that we shouldn’t reinvent the wheel to manage these risks. Instead, we should look to Financial Information Systems (FIS) as a model. For decades, these systems have utilized rigorous internal controls to maintain stability. By applying similar logic—such as strict segregation of duties, where a system cannot simultaneously set, execute, and validate its own tasks—we can prevent the "closed-loop autonomy" that often leads to unpredictable behavior.

The goal isn't to stifle innovation, but to create a safer scaffolding for it. Implementing boundaries like time limits, resource caps, and mandatory human approval for high-stakes decisions ensures that AI acts as an extension of human intent rather than an independent operator. By treating agentic AI with the same disciplined, regulatory rigor used in global finance, we move toward a future where powerful automation remains inherently safe and accountable.

As artificial intelligence transitions from passive tools to "agentic" systems—AI capable of autonomously breaking down complex goals into actionable steps—the conversation around safety is shifting. We are no longer just asking if a model can write an essay; we are asking if it can manage a budget, execute trade agreements, or modify infrastructure without causing unintended systemic failures. This autonomy, while powerful, creates significant operational risk if left unchecked.

Benjamin Palacio, an IT strategist, argues that we shouldn’t reinvent the wheel to manage these risks. Instead, we should look to Financial Information Systems (FIS) as a model. For decades, these systems have utilized rigorous internal controls to maintain stability. By applying similar logic—such as strict segregation of duties, where a system cannot simultaneously set, execute, and validate its own tasks—we can prevent the "closed-loop autonomy" that often leads to unpredictable behavior.

The goal isn't to stifle innovation, but to create a safer scaffolding for it. Implementing boundaries like time limits, resource caps, and mandatory human approval for high-stakes decisions ensures that AI acts as an extension of human intent rather than an independent operator. By treating agentic AI with the same disciplined, regulatory rigor used in global finance, we move toward a future where powerful automation remains inherently safe and accountable.