Skip to content
How to ThinkIn the Age of AI

Review new agents 7-14 days after deployment — remove…

Schedule agent addition reviews 7-14 days after deployment to verify whether actual coordination cost matches pre-addition estimates, removing the agent if cost exceeds estimate by more than 50%.

When This Fires

  • 7-14 days after deploying any new behavioral agent, process step, or team member
  • When a recently added agent "feels more expensive than expected" — measure rather than assume
  • When the coordination budget (Cap coordination at 15-25% of total hours — new…) is being exceeded after an agent addition
  • As a mandatory post-deployment gate before the new agent becomes permanent

Common Failure Mode

Skipping the review because the agent "seems to be working": the agent may be producing value while its coordination cost is quietly exceeding the budget. Value and cost must both be measured. An agent that produces $100 of value at $80 of coordination cost is a net $20 — but if the estimate was $40 of cost, the 50% overshoot (from $40 to $80) means the value case was built on wrong numbers.

The Protocol

(1) Before deploying a new agent, document the estimate: expected coordination cost per week, expected value per week. (2) Schedule a review at day 7 and day 14 post-deployment. (3) At each review, measure actual coordination cost: time spent on handoffs, synchronization, conflict resolution, context-switching, and communication related to the new agent. (4) Compare: actual cost vs. estimated cost. If actual ≤ 150% of estimate → the agent is operating within acceptable bounds. Continue. If actual > 150% of estimate → the coordination cost exceeds justification. Options: redesign the agent to reduce coordination overhead, or remove the agent entirely. (5) Document the findings for future agent additions: what coordination costs were underestimated? This improves future estimates.