Leaked Accenture audio pins its AI token spend on non-engineers, not engineers
Leaked Accenture meeting audio, surfaced by 404 Media, has agentic AI strategy lead Justice Kwak saying internal data puts most token consumption on non-engineering staff, with PDF-to-image-to-markdown conversion named among the biggest consumers. No spend figures accompany the claim, but it relocates AI cost control out of engineering budgets and into ordinary document workflows across the business. Simon Willison's read is that the cost pressure may push organisations to accept that PDFs are a poor medium for communicating information.
Source: simonwillison.net ↗
it's actually not our engineers that are driving the token consumption. It's a lot of the non-engineers
Justice Kwak, Accenture's agentic AI strategy lead
Why this matters
- → AI token costs concentrate in business workflows, not engineering—shifting budget accountability across orgs
- → PDF-to-markdown conversion emerges as a major token consumer, exposing document format inefficiencies
- → Cost pressure may force enterprises to rethink information architecture, not just prompt tuning
The PDF problem