Token exergy is a management analogy for the useful business outcome extracted from AI work. Borrowed from thermodynamics, where exergy means usable work, it shifts attention from raw token consumption to the value a model actually produces. In practice, high token exergy means the system turns prompts, context, and tool use into decisions, summaries, detections, or actions that matter.
In cyber security, the term helps teams judge whether AI is reducing risk or just creating expensive activity. A phishing triage assistant that correctly classifies alerts has high token exergy; a looping agent that repeats answers, retries tools, or burns context without improving the result has low token exergy. Defenders can watch for low exergy as a sign of poor prompt design, runaway automation, over-privileged tools, or prompt injection causing wasteful behavior. Used well, the metric supports tighter budgets, better governance, and stronger control over AI-driven workflows.