theaicatalog
Cognition ยท Products

Devin adds efficiency-based cost controls for enterprise AI spend

Cognition has launched three enterprise features in Devin that allocate AI spend by how efficiently each engineer uses it, instead of applying uniform usage caps. The Efficiency Score rates each engineer from 0 to 100 using estimated productive engineering hours per unit of spend, and it credits work such as investigations that never produce a pull request. Devin Coach flags common prompt mistakes, such as vague requests or an expensive model for a simple task, before a session runs. Efficiency-based approvals automatically raise limits for engineers graded Healthy or Satisfactory, up to a ceiling set by admins, while other requests go to an admin for review.