Commonwealth Bank of Australia (CBA) has flagged surging costs associated with artificial intelligence as the tasks the technology performs become more complex [1].
The warning comes as global corporations transition from initial AI experimentation to full-scale implementation. This shift highlights a growing tension between the promise of automation and the actual financial burden of maintaining sophisticated models.
CEO Matt Comyn said that businesses will tighten their scrutiny of AI-related spending through 2026 [1]. The bank is seeing a trend where the increasing complexity of required AI outputs drives up the necessary investment, challenging previous assumptions about the cost-efficiency of the technology.
Comyn said the emergence of low-quality AI-generated output is "work slop" [1]. This phenomenon occurs when AI produces superficial or inaccurate results that require human intervention to correct, effectively neutralizing the productivity gains the tools were intended to provide.
As the bank navigates these challenges, the focus is shifting toward demanding clearer returns on investment [1]. The rise in costs is attributed to the need for more powerful computing resources and the ability to handle more intricate data sets as the bank attempts to move beyond simple automation.
CBA's experience reflects a broader corporate trend in Australia and beyond. While AI was initially viewed as a way to reduce overhead, the reality of "work slop" and escalating operational costs is forcing executives to re-evaluate their budgets [1, 2].
“Businesses will tighten their scrutiny of AI-related spending through 2026.”
The CBA's public acknowledgment of 'work slop' and rising costs signals a cooling period in the AI hype cycle. It suggests that the 'low-hanging fruit' of AI efficiency has been picked, and the next phase of integration will require significantly more capital and human oversight to ensure quality and accuracy.
