Token Leaderboards
The article introduces Token Leaderboards, a feature by Didon that tracks employee usage of AI tools by monitoring tokens consumed. It aims to provide visibility into AI adoption across teams, helping organizations manage costs and understand usage patterns.
Background
- The article discusses "token leaderboards" inside AI-powered workplaces — dashboards that track how many tokens (the basic text units AI models process) individual employees or teams consume when using tools like ChatGPT or Copilot.
- Companies are starting to monitor AI usage the way they track time or keystrokes, raising questions about surveillance, productivity pressure, and whether token counts are a meaningful metric for output quality.
- Tokens are a technical cost unit: each ~0.75 word costs a fraction of a cent. High token usage might mean heavy AI assistance — or inefficient prompting — not necessarily good work.
- This sits in a wider debate: as AI tools become embedded in knowledge work, employers are struggling to measure "AI productivity" without resorting to crude counters that could encourage gaming the system or privacy erosion.