Voice AI Alerts & Uptime Monitoring

The problem

"How long has it been broken?" is the worst client conversation.

A Voice AI agent starts giving wrong answers Thursday afternoon. Your client notices Saturday morning. By the time they email you Monday, two days of bookings are gone, the relationship has cooled, and the conversation starts with "how long has this been happening?" — which is the version of "how long have you not been paying attention?"

SuperLedger's alerts are built to make that conversation impossible. The agent goes silent, an alert hits your inbox in minutes. Containment drops below threshold, you know within the hour. A HighLevel connection drops, you hear about it before any real customer does. You fix it. Your client never knew.

Real-time email alertsSilent-agent detectionPer-client routing

Live alerts · last 24h● Live

2

Critical

3

Warning

14

Resolved

Greenfield Legal — agent went silent (0 calls, 9am–12pm)

no_activity · expected-busy window

3m ago

Apex Roofing — containment dropped 12pt

threshold_breach · last 6h vs prior week

22m ago

Metro Dental — transfer pickup at 64%

below configured floor (75%)

1h ago

What triggers an alert

Five trigger types. One inbox.

📉

Containment rate drop

When AI containment falls below the default threshold (80%) over a rolling window, the alert fires. Catches degradation before it shows up in monthly reports.

📞

Transfer pickup rate falling

Measures the percentage of human transfers that actually get answered. If pickup drops, it usually means the staff routing is broken — not the AI. Either way, you want to know.

🔌

Disconnect / OAuth-drop detection

When a HighLevel connection silently drops — a revoked OAuth grant, an expired token, a disconnected sub-account — the alert fires. The outage that would otherwise look like "a quiet day" until your client noticed.

🔇

Silent-agent detection

If an agent receives zero calls in an expected window (e.g., normal business hours for that client), the alert fires. Often the first sign of a broken phone-number routing or a disconnected sub-account.

🚨

Gap-severity spike

When critical or high-severity gaps spike above baseline (often a sign that a knowledge-base entry or prompt change went wrong), the alert fires. Tied directly to Gap Analysis.

Real results

Here's what one SuperLedger client looks like after 30 days.

24,027

Calls handled

57%

AI containment rate

88.8%

Transfer pickup rate

456

Staff hours saved

Real data. One client. 30 days. A national automotive service chain running Voice AI across 50+ locations.

FAQ

Frequently asked about Alerts

What triggers an alert?

Containment rate dropping below threshold, transfer pickup rate falling, an agent going silent (no calls in a window it should be busy), a disconnect or OAuth drop on a HighLevel connection, and gap-analysis severity spiking. Alerts go out via email immediately.

How does SuperLedger know an agent is down?

Two real signals, watching your actual traffic. Silent-agent detection fires when an agent receives zero calls in a window it should be busy (the first sign of broken phone routing or a disconnected sub-account). Disconnect detection fires when a HighLevel OAuth connection silently drops. Either way you hear about the outage before your client does.

Do alerts cover Conversation AI chat agents too?

Yes. Both Voice AI and Conversation AI agents are monitored. Chat-side alerts include resolution-rate drops, escalation spikes, and silent-channel detection (no chats received in an expected window).

Where do alerts go?

Email by default — to one or more agency operators per agent. Alert routing can be scoped per client or per agent. Webhook delivery is on the roadmap for Slack and PagerDuty integration.

Related features