73%
reduction in human pages — ASX-200 retailer, 200 pages a week to 54
8 min
Black Friday CDN fault caught before it reached production — precursor in telemetry, zero human pages
$1.2M
annual run-cost saved at the ASX-200 retail engagement in year one
≤60s
mean time to detect for known fault patterns, versus 4–12 minutes before the triage agents
The structural problem with traditional SRE
Your on-call team is spending 70% of their nights on alerts a machine could close.
Read 12 months of incident logs for a mid-market stack and the same five fault patterns account for the majority of out-of-hours pages. The same alert fires. The same runbook gets pulled. The same engineer types the same commands at 3am. An ASX-200 retailer ran 200 pages a week into a small SRE team. Effektiv read twelve months of their incident logs, extracted the real fix steps from actual resolution data — not from runbooks — and built triage agents around those patterns.
Human pages dropped 73%. Eight minutes before a Black Friday CDN failure would have reached production, the triage agent caught the precursor in telemetry and closed the incident without waking anyone. A vendor priced on seat count has no incentive to reduce the volume of incidents a human handles. Effektiv's retainer is written the opposite way: the bill goes down as the automation rate goes up.
What changes
The same challenge. Two very different outcomes.
Without Effektiv
- 200 pages a week into a small SRE team
- Same five fault patterns re-paged every week
- Engineers type the same commands at 3am, every week
- Post-mortems in a shared drive nobody re-reads
- Mean time to detect 4–12 minutes from precursor to alert
- Vendor priced by seat — no incentive to reduce volume
With Effektiv
- 27% of pages reach a human — the rest auto-close
- Triage agents read alerts against 12 months of actual fix data
- Known faults close behind a rollback gate without a person in the loop
- Resolution database queryable and extendable by your team
- Mean time to detect under 60 seconds for known patterns
- Retainer goes down as the automation rate goes up — incentives aligned
Why incentive alignment matters
Three on-call vendor models.
| Dimension | Effektiv agent triage | Vendor priced by seat | Alert-to-jira automation |
|---|---|---|---|
| Incentive alignment | Bill goes down as automation rises | Bill rises with seats | Per-event pricing |
| Human-page reduction | 50–70% | 0–10% | 15–25% |
| Rollback gate per step | Yes, named in Design | None | Manual rollback |
| Mean time to detect | ≤60s for known patterns | 4–12 minutes | 1–3 minutes |
How we deliver
Diagnose. Design. Deliver.
Two weeks of listening before a line of code. The price is fixed at the end of Design — not at kick-off.
Phase 1 · 1–2 weeks
Diagnose
We map your incident log, runbooks, and cost telemetry. We read 12 months of actual incident history — the commands engineers actually ran to resolve each fault, not the runbooks people meant to follow. We identify which patterns are candidates for automation and which need a human in the loop by design.
Phase 2 · 1–2 weeks
Design
Triage rig spec, rollback rules, and eval gates. Human-in-the-loop requirements documented. Any fault pattern touching a money write or a record of truth stays gated. All model inference on AWS Bedrock in AU regions, inheriting VPC, IAM, PrivateLink, CloudTrail, and KMS controls.
Phase 3 · 4–8 weeks
Deliver
Triage agents built and tuned in a parallel run alongside your existing on-call process. The switch-over is incremental, not a single cut-over. The outcome contract names the deflect rate and MTTR targets — both measured and reported weekly.
Quality gates
What the quality checks measures.
Every output passes a multi-gate evaluation before it merges or ships. Outputs that fail do not proceed. The quality checks and all gate code are yours at exit.
- Triage accuracy — correct routing as a percentage of total alerts, threshold agreed in Design
- Mean time to detect for known fault patterns — target under 60 seconds
- False-positive rate on AI triage decisions — any drift triggers an eval refresh and a paused-automation period
- Human-page reduction vs the prior baseline — measured weekly against the contract target
- Incident review completion rate — AI agent contributes diagnostics on every paged incident
Eval rig · sample run
Eval rig source code shipped to your repo at exit.
Sample engagement
An ASX-200 retailer ran a peak-trade stack with a small SRE team and 200 pages a week. Effektiv read twelve months of incident logs, pulled the real fix steps from resolution data, and built triage agents from those patterns over six weeks. Human pages dropped 73%. A CDN precursor was caught eight minutes before it would have reached production on Black Friday. Annual run-cost saved: $1.2M.
Read the full case →
Compliance posture
ISO 27001 in progress (Q3 2026) ISO 42001 aligned NIST AI RMF mapped IRAP path Q4 2026 Full governance posture →Other services
Other ways we work with you.
Service
Modernisation
AI archaeology decodes what the documentation missed. 11 weeks median.
Read more →Service
Operations & Integration
RPA and ESB out, connected AI systems in. Run cost down 50–70%.
Read more →Service
AI Adoption
With-you mode. Your team ships AI without us in 90 days.
Read more →Service
Customer Experience
AI helpers that draft for human review. Brand-voice eval gated.
Read more →Service
Software Build
Custom products and greenfield software, shipped in 8–14 weeks.
Read more →Common questions
Frequently asked questions.
The retainer goes down as automation goes up
See what your on-call stack looks like with AI in the alert path.
Show us 12 months of incident logs or your current on-call setup. We diagnose which fault patterns are candidates for automation and price the triage rig on outcomes — your page reduction is the benchmark.