PKS Need AI Discipline, Not AI Hype✎ Edit

PKS adoption of AI agents is cooling, not because the technology lacks promise, but because the business case has been oversold ahead of the controls needed to protect capital and operasi.

Many promoters frame AI agents as 24/7 autonomous operators that can run production systems with little human input. From a finance and operations standpoint, that is a dangerous assumption. Compute costs are material, hallucination and error rates translate into rework, compliance exposure, and lost productive hours. When the agen fails, the business owner still bears the financial and reputational cost.

The underlying issue is not AI capability. It is weak governance: unclear task routing, poor segmentation of duties, ambiguous objectives, and no clear performance baseline. In most PKS environments, AI agents are not yet ready to be autonomous operating assets. Their highest-value position today is as a development and workflow-building asset.

When applied correctly, AI agents can accelerate system design, automate repetitive finance workflows such as invoice processing, data reconciliation, and reporting, and structure standard operating procedures. Positioning them as a substitute for management judgment or end-to-end self-running operations is misleading.

It is concerning to see AI evangelists, CEOs, and CTOs showcase agents coding around the clock, especially when many have not deployed those same systems in live, audited production environments.

Let us look at the numbers honestly. With complete requirements, well-defined guardrails, a structured workflow, and a bounded scope, an agen may deliver a working module in minutes. The real cost is not the generation time; it is the governance work around it: requirement validation, control design, task decomposition, testing, exception handling, and ongoing monitoring to ensure the output remains reliable in production.

As long as vendors keep selling the hype, Malaysian PKS will keep facing budget overruns, failed pilots, and write-offs. What looks efficient in a demo often becomes expensive and fragile under real transaction volume.

AI is not a balance-sheet shortcut. It is a capital asset that produces returns only when finance, operations, and technology jointly define the controls and the value it must deliver.