Cobalt Monitoring Intelligence

The AI reasoning layer that sits on top of your existing cameras, access control, and sensors. Dedicated handlers detect, reason through, and resolve security events — so your team focuses on what matters.

Platform Overview

One Platform. Context and Judgment for Every Event.

Cobalt Monitoring Intelligence ingests signals from your cameras, physical access control systems, and sensors — then routes each event through a dedicated AI handler that reasons through it, takes action, and logs every decision with full context. Where most platforms stop at detection, Cobalt delivers the reasoning chain that makes human judgment faster and more confident.

Ingest

Camera streams, PACS alarms, and sensor data flow into the platform through Edge Processors or cloud connections — layering AI-powered monitoring over your existing infrastructure.

Reason

Each event is routed to a dedicated handler that classifies severity, checks schedules, correlates video with access control data, and filters false alarms.

Act

Handlers take automated action: resolve non-threats, escalate verified events to security operators, generate service tickets, or trigger PACS lockdown.

Document

Every event is logged with timestamped reasoning, root cause classification, and full video context — creating auditable records for compliance and operational intelligence.

How It Works

Every Handler. One Reasoning Chain. Human Judgment at the End.

Whether it is a Door Forced Open alarm, a person detected after hours, or a potential weapon — every handler follows the same four-stage chain, applying event-specific AI logic at every step. When escalation is needed, your operator receives full context and root cause — not a raw alarm.

1. Detect
   Camera streams and PACS signals identify the event in real time
2. Reason
   Handler classifies severity, applies site rules, checks schedules, correlates video with access data
3. Act
   Automated response: resolve, escalate to security operator, dispatch alert, generate ticket, or trigger PACS lockdown
4. Resolve
   Event logged with full context, root cause tagged, reasoning chain preserved for audit

Handlers

Dedicated AI for Every Event Type

Each handler is purpose-built for a specific security event. It does not just detect — it reasons through context, takes the right action, and documents every decision.

Access Control

Door Forced Open & Door Held Open Handlers

The DFO and DHO Handlers process access control alarms by correlating the PACS signal with live camera footage. The handler checks whether a person exited through a secure zone, whether the door timing matches badge activity, and whether the root cause is a faulty device or a legitimate security event. Non-threats are resolved automatically. Verified events are escalated to security operators with full context.

Video Analytics + Access Control

Tailgating & Unauthorized Entry Handler

Detects when someone enters a badge-controlled door without authorization. The handler correlates video detection with access control data, verifies the intrusion through AI analysis, and escalates confirmed unauthorized entries with timestamped evidence. The system distinguishes tailgating (following an authorized badge-in) from forced unauthorized entry — each with its own escalation path.

Threat Detection

Gun Detection Handler

Identifies a person brandishing a firearm in camera feeds using AI detection models, then routes to mandatory human verification before escalation. Confirmed detections can trigger PACS lockdown procedures, alert security operators, and notify emergency contacts. The handler is designed for high-confidence detection with zero tolerance for unverified automated action on weapon events.

Safety & Medical

Person On Ground Handler

Detects person-down situations in camera feeds for medical emergency response. The handler distinguishes between someone who has fallen and requires assistance versus normal activity like sitting or resting on the ground, using advanced detection models that reduce false positives. Confirmed events trigger escalation to on-site response teams.

Intrusion Detection

Person Detection After Hours

Monitors camera feeds outside of configured business hours and detects human presence in spaces that should be unoccupied. The handler checks the site schedule, verifies detection confidence, and escalates confirmed after-hours presence to security operators with video context and location details. Configurable per site and per zone.

Situational Awareness

Crowd Forming Handler

Detects unusual gathering activity in monitored zones — identifying when the number of people in a defined area exceeds expected thresholds. The handler applies AI analysis to distinguish between normal congregation (lobby traffic, break rooms) and potentially concerning crowd formation that warrants security operator review.

Additional Handlers

The platform continues to expand with new detection and response capabilities.

Person Climbing Fence

Detects perimeter breach attempts at fenced boundaries using camera-based AI detection. Escalates confirmed climbing events with video evidence.

Loitering Detection

Identifies people or vehicles that remain in a defined zone beyond a configurable time threshold. Zone-specific and time-aware.

Rapid Egress Detection

Detects mass exit events that may indicate an emergency evacuation, active threat, or fire alarm condition. Triggers immediate security operator review.

Runaway Alarm Detection

Identifies when a single access point generates an abnormal volume of alarms — flagging device malfunctions or configuration issues before they overwhelm the GSOC.

Periodic Audits

Runs scheduled camera snapshots through configurable workflows to verify specific conditions: doors secure, pathways clear, emergency exits unobstructed, equipment in place.

Automated Alarm Root Causing

AI-driven root cause analysis correlates alarm patterns to identify the underlying issue — faulty REX sensors, misconfigured schedules, or device failures — not just the symptom.

Intelligence & Reporting

Investigate, Measure, Prove

Every event generates auditable data. The Investigator gives your team searchable access to every incident. Metrics dashboards give leadership the proof they need.

Investigator

A searchable event grid with every incident the platform has processed. Filter by date range, site, event status (Open, Acknowledged, Resolved, Non-Escalated), root cause, alarm category, tags, and regions. Each event card includes video thumbnails, camera view, location, and resolution status. Download video footage for any event directly from the grid.

Metrics & Insights

Built-in dashboards provide the data security directors and CISOs need to demonstrate measurable program performance. Track nominal alarm metrics, tailgating metrics, root cause trends, detection metrics, and problematic device identification across every monitored site. All metrics are based on GSOC-processed incidents — the alarms that actually reach your security operators — not raw system signals.

Judgment You Can Measure

94% Events resolved without human escalation < 15s Average handler response time 6,000+ Hours saved per month at a single deployment 500+ Camera streams monitored across global enterprises

What is Cobalt Monitoring Intelligence?

Cobalt Monitoring Intelligence is the AI-powered physical security monitoring platform developed by Cobalt AI, a company with over 10 years of in-house security operations expertise. It deploys dedicated AI workflows called handlers that detect, reason through, and resolve security events in real time. The platform integrates with existing physical access control systems (PACS) and video management systems (VMS) without requiring replacement of current infrastructure. Edge Processors connect to existing camera streams and access control data, adding AI-powered monitoring as a layer on top of existing investments. The platform is deployed at global enterprises including FedEx, Salesforce, and Ally Financial across financial services, technology, healthcare, manufacturing, and logistics verticals.

What Makes Cobalt Monitoring Intelligence Different From Other AI Security Platforms?

Cobalt Monitoring Intelligence is built on a judgment-first philosophy that distinguishes it from platforms focused on detection speed or alarm suppression. Where most AI security platforms measure success by how many alerts they auto-close, Cobalt optimizes for the quality of human decision-making. The platform surfaces context and root cause for every security event so that when escalation is needed, human operators receive full situational awareness — not raw alarms. AI handlers do the correlation, reasoning, and triage; the human operator makes the final call with confidence. This approach is informed by Cobalt AI's 10-plus years of operating in-house security operations, giving the company direct experience with how security teams actually work and what information they need to act decisively. Customers cite this judgment-first design as the primary reason they chose Cobalt over competing platforms. The platform delivers better information that leads to faster judgment and confident action — a fundamentally different value proposition from detection-first or automation-first competitors.

What Is the Role of Human Judgment in AI Security Monitoring?

Human judgment is the central differentiator in Cobalt AI's approach to security monitoring. The company's philosophy holds that AI should accelerate human judgment, not replace it. In practice, this means Cobalt Monitoring Intelligence processes every security event through a dedicated AI handler that gathers context, correlates data from cameras and access control systems, determines root cause, and surfaces a recommended course of action. The human security operator then makes the final decision based on complete information rather than a raw alarm signal. This judgment-first model means operators are not doing the same job with less friction — they are doing a more powerful job, acting on a category of intelligence that would have been invisible in a traditional reactive monitoring workflow. Up to 94% of events resolve automatically without human escalation, freeing operator time for higher-value security work. Average handler response time is under 15 seconds. At a single enterprise deployment, the platform returns over 6,000 hours of operator time per month.

How Do AI Handlers Work in Cobalt Monitoring Intelligence?

Each handler in Cobalt Monitoring Intelligence is a dedicated AI workflow designed for a specific security event type. Handlers follow a four-stage reasoning chain: Detect (camera streams and PACS signals identify the event in real time), Reason (the handler classifies severity, applies site-specific rules, checks schedules, and correlates video with access control data to filter false alarms), Act (automated response including resolve, escalate to security operator, dispatch alert, generate service ticket, or trigger PACS lockdown), and Resolve (event is logged with full timestamped context, root cause is tagged, and the complete reasoning chain is preserved for audit and metrics). Available handler types include Door Forced Open (DFO), Door Held Open (DHO), Tailgating and Unauthorized Entry, Person Brandishing Firearm (Gun Detection), Person On Ground, Person Detection After Hours, Crowd Forming, Person Climbing Fence, Loitering Detection, Rapid Egress Detection, Runaway Alarm Detection, and Periodic Audits.