GovernSegment: All traffic

SLAs

Set latency and success rate targets per agent. Monitor compliance and intervene when SLAs are at risk.

Regression

3 SLAs at risk

2 meeting targets. 3 need attention.

Set SLA Investigate latency

Agents5

Healthy2

At risk3

Foxhound suggests

LangGraphWrap the slowest tool call with a 1.5s deadline + one retry.

Expected impact: Modeled p95 latency drops below the configured SLA without affecting success rate.

SLA decomposition

Separate latency pressure from success-rate pressure so the next control change targets the real failure mode.

Latency breaches

warning

Agents currently missing their duration target.

Success breaches

critical

Agents currently missing their minimum success-rate target.

Both dimensions

critical

Agents breaching both reliability dimensions at once.

Scoped view

Fleet-wide

No agent segmentation is active.

AgentLatencySuccess rateStatus

Returns Resolution Copilot

Returns Resolution Copilot breached both latency and success-rate expectations during the regression window.

9.6s

/ 8.0s max

91.8%

/ 97% min

Shipping Delay Resolution

Shipping fallback improvements helped, but the workflow remains near the SLA edge after the timeout cluster.

11.4s

/ 12.0s max

95.2%

/ 96% min

Premium Escalation Triage

Escalation quality improved, but high-value billing cases remain under close monitoring.

4.3s

/ 5.0s max

98.7%

/ 99% min

Billing Operations Guard

Billing dispute handling remained comfortably inside target during the same weekly narrative.

6.4s

/ 7.0s max

97.8%

/ 97% min

Account Recovery Assistant

Account recovery stayed fast and highly reliable even during peak support hours.

3.9s

/ 4.5s max

99.6%

/ 100% min

5 SLAs

GovernSegment: All traffic

SLAs

Set latency and success rate targets per agent. Monitor compliance and intervene when SLAs are at risk.

Regression

3 SLAs at risk

2 meeting targets. 3 need attention.

Set SLA Investigate latency

Agents5

Healthy2

At risk3

Foxhound suggests

LangGraphWrap the slowest tool call with a 1.5s deadline + one retry.

Expected impact: Modeled p95 latency drops below the configured SLA without affecting success rate.

SLA decomposition

Separate latency pressure from success-rate pressure so the next control change targets the real failure mode.

Latency breaches

warning

Agents currently missing their duration target.

Success breaches

critical

Agents currently missing their minimum success-rate target.

Both dimensions

critical

Agents breaching both reliability dimensions at once.

Scoped view

Fleet-wide

No agent segmentation is active.

AgentLatencySuccess rateStatus

Returns Resolution Copilot

Returns Resolution Copilot breached both latency and success-rate expectations during the regression window.

9.6s

/ 8.0s max

91.8%

/ 97% min

Shipping Delay Resolution

Shipping fallback improvements helped, but the workflow remains near the SLA edge after the timeout cluster.

11.4s

/ 12.0s max

95.2%

/ 96% min

Premium Escalation Triage

Escalation quality improved, but high-value billing cases remain under close monitoring.

4.3s

/ 5.0s max

98.7%

/ 99% min

Billing Operations Guard

Billing dispute handling remained comfortably inside target during the same weekly narrative.

6.4s

/ 7.0s max

97.8%

/ 97% min

Account Recovery Assistant

Account recovery stayed fast and highly reliable even during peak support hours.

3.9s

/ 4.5s max

99.6%

/ 100% min

5 SLAs