Why it matters
It reframes token spend as a function of job role and task, not headcount — finance teams can forecast burn by the kind of work being done, not just the number of seats logging in each month.
Tokenmaxxing read
If marketing managers cost 2.5x editors and Claude Code sessions run 0.37 points more autonomous than chat, routing and budgets should key off task type and autonomy level rather than a flat per-seat token cap.
Source takeaway
A primary dataset straight from the model maker; treat the linked survey of 9,700 workers — a third of whom think AI could handle nearly all their job within a year — as sentiment, not a spend metric.
