Your on-call team pages a human only when a human matters.

The AI agent sits between the monitoring system and the human on-call engineer. When an alert fires, the agent reads the alert context, checks the runbook, queries recent incident history, and produces a triage summary with a recommended action — before paging a human. Known faults with a documented remediation path are closed by the agent. Novel faults, faults involving customer data, and faults outside the agent's confidence threshold are escalated to a human with the triage summary already prepared. The human pages less often and is better prepared when they do.

From $25k/month, covering: EffektAI agent mesh operation (alert triage, known-fault remediation, anomaly detection), monthly eval rig runs against your production system, SRE capacity for incident response and post-mortem, and quarterly architecture reviews. The retainer goes down as the automation rate goes up — as more fault classes are handled without human intervention, the scope of human SRE required decreases and the retainer adjusts. This is a contractual mechanism, not a verbal commitment.

Typically within the first 4–6 weeks of the engagement. The Diagnose phase identifies the top fault classes by volume and maps which have documented remediations. We configure the agent to handle those first — the highest-volume, lowest-novelty faults. In most engagements that accounts for 50–70% of pages. The remaining pages are the genuinely novel faults that benefit from a fresh human eye. Engineers stop being woken for things a script could close.

Three primary gates: triage accuracy (did the agent correctly classify the fault and recommend the right remediation?), mean time to detect (is the agent catching faults faster than the previous monitoring setup?), and false-positive rate (is the agent generating alerts that turn out to be non-issues, wasting engineer time?). The rig runs monthly against production data and the results are included in your quarterly architecture review.

Yes. The Diagnose phase applies the same archaeology to a system we are inheriting as to a modernisation engagement. We map what's running, identify the known fault classes, review the existing runbooks, and assess the monitoring coverage. A typical inherited-system Diagnose takes 2 weeks and produces a gap analysis and a proposed agent configuration. We have inherited systems built by Big-4 firms, offshore teams, and internal development teams.

The agent escalates immediately and the senior Effektiv SRE engineer joins your incident bridge within 15 minutes during business hours (AEDT/AEST), 30 minutes outside business hours. We do not operate a third-party NOC or offshore on-call rotation. The engineer who joins your bridge is the same engineer who configured the agent and knows your system. Post-incident, we update the agent's fault taxonomy so the same class of fault is handled automatically next time.