How we use AI
AI-assisted analysis. Human-confirmed action.
Most providers say they use AI. We'll tell you exactly what ours does, what it doesn't, and why a technician still reviews every recommendation before acting on it.
Every alert moves through this loop — the AI proposes, the technician disposes, the resolution is documented for next time.
What the AI does
- Reviews each monitoring alert against the site's runbook and incident history. Past issues become pattern data.
- Generates a likely diagnosis with a confidence score. The technician sees both the diagnosis and how confident the AI is in it.
- Suggests a recommended response action — restart this service, dispatch a technician, escalate to vendor support, or wait and observe.
- Flags patterns across multiple sites. Recurring failure modes, predictable component lifespans, and seasonal correlations get surfaced before they cause incidents.
- Drafts initial customer-facing communication for the technician to review and send. Saves the writing time without skipping the human review.
What it doesn't do
- Take autonomous action on your network. Every action is initiated by a human technician. No AI agent restarts services, opens tickets to vendors, or modifies configurations.
- Generate new alerts. The AI analyses existing alerts from the monitoring layer. It can't fabricate alerts, and it can't suppress them either — every alert is logged regardless of AI assessment.
- Replace technician judgement on safety-critical or compliance-sensitive decisions. Anything touching evacuation, emergency dispatch, regulated reporting, or chain-of-custody goes through a human, full stop.
- Train on your data outside our infrastructure. Your site data, runbooks, and incident records stay in our infrastructure (Microsoft Dataverse, hosted in Australia). They are not used to train any third-party model.
Why this matters
- Faster resolutions. A technician with AI-assisted analysis resolves issues in a fraction of the time taken without it. Less time understanding, more time acting.
- Better-documented incidents. Every alert is analysed, the analysis is preserved, and the resolution is logged against it. Institutional knowledge compounds across all our sites over time.
- Fewer false-positive callouts. Most monitoring noise is genuinely noise. AI-assisted triage filters it before it becomes a 2am phone call.
- Honest about the limits. We tell you exactly what's automated and what isn't. There's no “AI-powered” handwaving here — the boundary between AI and human is explicit, by design.
Want the technical detail?
The AI lives inside the Mallen Stack — same infrastructure that runs continuous monitoring and case management. See how the Stack fits together →
Or, if you've got a specific question about how AI fits with your existing security or IT setup, get in touch directly.