The scorecard
What arrives at the end of week one
Before any work is committed. If the number is fine, you have spent nothing and learned that.
Reliability scorecard
support-triage agent
Overall
2.5/10
Verdict
do not ship
Weakest
Escalation calibration
Est. work
8 days
9.4
Drift detection — A threshold rule would have paged 88 times over this window. After FDR correction, 5 are real.
0.0
Escalation calibration — The current threshold delivers 5.9% errors against a stated budget of 2%. It was picked by hand and has no guarantee attached.
4.5
Eval sensitivity — The suite can see a 6.7 point regression. You said you care about 3.
6.1
Test determinism — A suite that passes 61% of the time on unchanged code cannot tell you whether a change broke something.