A smarter approach to AI cost controls: allocating spend by efficiency
Summary
- Cognition introduced three new enterprise features for Devin to manage AI spend via engineering efficiency rather than blanket usage caps.
- The Efficiency Score grades engineers from 0 to 100 based on productive engineering hours per Agent Compute Unit (ACU) alongside volume.
- Devin Coach analyzes prompts prior to execution, offering inline suggestions to correct vague requests, expensive model choices, or unrelated tasks.
- Efficiency-based approvals automatically raise usage limits for engineers with Healthy or Satisfactory grades, reducing administrative review overhead.
Why it matters
Moving from blunt AI usage caps to efficiency-based allocation helps enterprises scale developer productivity without runaway agent costs. It automates limit approvals for power users while guiding less efficient users toward better prompt engineering habits.