AI Usage in Performance Reviews: Why the Token Counter Broke as a Metric
A large employer put AI usage into its review guidance for 10 months, then took it out. The token counter was an input metric doing an outcome metric's job.
Research, analysis, and commentary on workforce intelligence, AI adoption, and human capital strategy from the Levos AI team.
AI Usage in Performance Reviews: Why the Token Counter Broke as a Metric
A large employer put AI usage into its review guidance for 10 months, then took it out. The token counter was an input metric doing an outcome metric's job.
Why Finance and HR Report Different Headcount Numbers
Finance and HR disagree on headcount because they read different hierarchies. Neither one is wrong, and no agreed external definition exists to settle it.
The AI Productivity Paradox: Perceived Gains Are Outrunning Measured Ones
Fed economists found perceived AI productivity gains exceed measured gains. Four 2026 datasets give four answers, and nearly all of them are self-report.
The CFO's Guide to Workforce ROI
Workforce is the largest line on the P&L and the least measured. A CFO's guide to building a defensible workforce ROI number, from cost to demonstrated output.
80 Percent of Engineers Get a 3 Percent AI Productivity Gain. The Top 20 Percent Get 55.
McKinsey found most engineers gain about 3 percent from AI while the top fifth gain 55 percent. A company-level average describes almost nobody in the building.
Zero of 150 Leaders Achieved AI Cost Reduction. Here Is Why.
A study of 150 leaders found no organization achieved AI cost reduction. Two-thirds said costs rose. Only 43 percent could say where AI was being used.