tech
Australia’s biggest bank says corporate AI is racking up bigger bills and producing ‘work slop’
CBA chief executive Matt Comyn used the phrase ‘work slop’ to describe the low-quality AI output now flowing through corporate workflows, as token-billed AI costs scale with task complexity.
TL;DR
- Corporate AI adoption faces two main problems: rising costs and "work slop" (low-quality output).
- The cost of running generative AI is increasing substantially faster than budgeted due to task complexity driving up token consumption.
- "Work slop" refers to low-quality AI-generated text, code, and analysis that degrades internal workflows when used without sufficient quality control.
- Token-based pricing models are leading to significant operational expenses as task complexity increases non-linearly.
- This "work slop" problem is analogous to the "AI slop" seen with image generation tools, impacting corporate knowledge work.
- The cost-benefit ratio for AI is tightening for large institutions, leading to increased scrutiny of AI spending and pressure to demonstrate ROI.
- Falling per-token prices have been offset by rising per-task token consumption as companies move from pilot to production AI use cases.
- A "procurement-discipline phase" for AI is predicted to continue through 2026.