KPMG names AI audit limits: can't run thousands of audits, client data too dirty
krishnan · x · 2026-11-18
Commenting on KPMG's AI audit push, krishnan notes KPMG itself named two limits: the tech can't yet run thousands of audits, and few companies have data clean enough for agents—so the first bottleneck is the client's ledger, not the model. Since audit is the economy's evaluation layer and evaluation gets automated first, once 100 agents do fieldwork staff hours stop being the cost. KPMG sees "clear time-saving opportunities" but won't say if fees fall—the signature and liability become the product. Who keeps the hours your auditor no longer spends?
More from AGI Musings
- Are women more open about AI companionship, or just less stigmatized? — mark_k · 2026-11-18
- Supply chains, not technology, decide what products can exist — yunta_tsai · 2026-11-18
- Is it over for game designers? Reddit debates frontier models replacing computer work — Ok-Scarcity-7875 · 2026-11-18
- Audit is the economy's evaluation layer; signature and liability become the product — krishnan · 2026-11-18
- AI detectors and literary gatekeepers: "a symbiotic pincer movement of cowards" — aiamblichus · 2026-11-18
- Are open-source models just linear combinations of Claude and GPT minds? — amplifiedamp · 2026-11-18