Sample brief · illustrative data
A week in the life of an agent fleet.
This is the weekly brief for a demo organization running six agents. The numbers are illustrative; the behavior is the product's.
data-extractor needs attention this week.
Spend
$881.68
10% up on the previous week ($802.63).
Agent runs
13,460
3% more than the previous week (13,129).
Production success rate
85.3%
9,302 runs have a known outcome (69% of runs). -0.1 points on the previous week (85.4%).
Agents, most urgent first
AgentLabeled runsSuccessVs last weekSpendVerdict
- data-extractor
- Labeled runs
- 610 of 773
- Success
- 45.2%
- Vs last week
- -8.1 pts
- Spend
- $68.80
- code-reviewer
- Labeled runs
- 1,402 of 1,847
- Success
- 72.4%
- Vs last week
- +0.0 pts
- Spend
- $262.27
- triage-agent
- Labeled runs
- 373 of 373
- Success
- 79.4%
- Vs last week
- -11.9 pts
- Spend
- $74.16
- doc-summarizer
- Labeled runs
- 41 of 2,104
- Success
- —
- Vs last week
- —
- Spend
- $140.97
- lead-qualifier
- Labeled runs
- 3,510 of 4,231
- Success
- 94.2%
- Vs last week
- +1.8 pts
- Spend
- $181.93
- support-router
- Labeled runs
- 3,120 of 3,892
- Success
- 89.1%
- Vs last week
- +0.0 pts
- Spend
- $120.65
Verdicts:
An alert fired this week
Investigation · Latency threshold
seat-booking-agent
Couldn't form a confident hypothesis — here's the raw correlated data.
- p95 latency 6.2s → 9.4s over 4 hours, then recovered on its own without a deploy or config change.
- Slow runs are spread across every tool and prompt path — no single span dominates the regression.
Correlated data is attached; a provider-side latency incident would explain the shape but could not be confirmed from your traces.