For the complete documentation index, see llms.txt. This page is also available as Markdown.

Recovery (R) — Money you could have saved

The Recovery pillar quantifies waste and surfaces opportunities to reduce spend without cutting capability. It identifies four distinct waste categories:

  • Zombie Sessions — AI agent runs that produced no commits or output

  • Idle Licenses — seats paid for but never used in the period

  • Stale Code — AI-generated lines that were committed but never merged or immediately reverted

  • Model Misselection — sessions where an expensive model was used for tasks a cheaper alternative handles equally well

See total recoverable spend across your organization and drill down to the team and model level. Compare accepted versus rejected AI-generated code to measure output quality and spot teams struggling with AI integration. This pillar transforms abstract "waste" into concrete dollar figures.

Key metrics:

  • Total Recoverable — aggregate savings potential across all waste categories

  • Productive vs. Recoverable Spend — stacked view showing real output versus waste, by team or model

  • AI LOC Accepted vs. Not Accepted — lines of code generated but rejected, with cost visibility

  • Recovery Opportunity Map — treemap visualization showing waste distribution by category and team

Last updated