From Tokenmaxxing to Trusted Throughput — Mingsheng Hong, Ironclad
Token usage dashboards should act as smoke detectors, not leaderboards incentivizing waste
“we don't want to create even indirect incentive to maximize the token usage itself”
Ironclad's VP of Engineering describes a maturing phase of enterprise AI adoption where runaway token costs — including anecdotes of $500M monthly cloud bills — are becoming a serious concern. He argues usage dashboards should flag low-adoption pockets as a warning signal rather than rank high consumers, to avoid perverse incentives. The talk frames AI cost governance as the next challenge after adoption, relevant to roughly half of engineering orgs in the audience.