CBA Chief Warns of Non-Linear AI Costs and Corporate 'Work Slop'

Artificial intelligence expenses are set to spiral as companies transition from simple automation to complex reasoning tasks, according to Commonwealth Bank of Australia CEO Matt Comyn. He warned that corporate spending on AI will face intense scrutiny through 2026 as businesses struggle to justify the mounting costs of model tokens.

Unlike flat-fee consumer services, corporate AI deployment is billed by the token, a metric that becomes increasingly expensive as models require more context and data processing. Comyn noted that these costs do not scale linearly, creating a significant management hurdle for firms attempting to integrate advanced reasoning into their daily operations. As the technology demands more computing power, energy, and water, the financial burden is beginning to outpace initial projections.

Despite these headwinds, the bank remains an aggressive adopter. CBA recently appointed the country’s first bank-based chief AI scientist and hosted OpenAI CEO Sam Altman to discuss future strategy. Comyn argues that the rising price of tokens may actually serve a functional purpose by discouraging the proliferation of low-value automated output, which he dubbed 'work slop.' He suggested that the ease of generating endless reports and presentations has led to an abundance of corporate noise that firms must now learn to curb.