Legora CTO Criticizes Tokenmaxxing as Ineffective Incentive

Business Insider reports that Legora chief technology officer Jacob Lauritzen criticised what he called "tokenmaxxing" on an episode of the "20VC" podcast. Business Insider quotes Lauritzen saying, "A lot of people, say, and bring up token usage at performance reviews... That leads to tokenmaxing, which is people just burn tokens just to look good," and, "That's a really stupid way to do anything." The article reports Lauritzen joined Legora in 2024 and that he recommended hack days and demos as better mechanisms to surface productive AI work.

Reporting frames tokenmaxxing as the practice of inflating AI-tool usage metrics by heavy, often low-value calls to models such as Claude, Codex, and Cursor. Observed patterns in organizations that track raw token consumption show these dashboards can encourage superficial behavior: users optimize for the metric rather than outcome. That dynamic raises two operational issues for practitioners: monitoring token volume alone is a weak signal for productivity, and unchecked token growth increases cloud and API costs without proving value.

Industry reporting places Lauritzen's comments amid broader debates over how to measure AI adoption inside firms. Industry observers have increasingly shifted attention from raw usage metrics to outcome-oriented KPIs such as time saved, defect reduction, or deliverable throughput. For companies with high opportunity cost per engineer-hour, the calculus for exploratory token spend will differ from lower-cost environments; Lauritzen's on-record quote about trade-offs illustrates that tension without prescribing a single approach.

For practitioners and managers, signals to monitor include whether internal dashboards evolve from token counts toward outcome-linked metrics, adoption of structured experimentation practices such as hack days, and tools that correlate model calls with business KPIs. Editorial analysis: Observers should also watch cost-control tooling that attributes token spend to projects and windows of experimentation, since attribution is necessary to decide when exploratory spend produced measurable efficiency gains.