Intrusion AI
Intrusion AI helps agriculture operators continuously understand intrusion conditions across safety-critical areas. Instead of relying only on manual supervision or retrospective reporting, the agent analyses existing CCTV streams against site-specific policy and detects the observable conditions that precede incidents. When something matters, GuardianOS evaluates the zone policy, assigns a severity, routes the alert to the responsible owner, captures time-bounded evidence and updates the safety dashboard — without replacing a single camera.
The problem this agent solves
Today, intrusion issues in agriculture operations are found by whoever happens to be looking. Coverage is inconsistent between shifts, evidence is reconstructed after the fact, and the same conditions repeat until they cause harm, downtime or a failed audit.
Why it matters: Without Intrusion AI, intrusion risk is only visible when someone reports it — usually after the event. With it, the same cameras you already own become a continuous, policy-aware observer that never blinks, never changes between shifts and always produces the record your auditors ask for.
What the agent does
- Detects intrusion conditions on selected cameras and zones in real time
- Applies your configured policy, severity model and operating schedule
- Suppresses repeat and low-confidence events to protect operator attention
- Escalates to the correct owner when the condition persists or worsens
- Creates a complete evidence package for every confirmed event
How it works
- Connect existing RTSP/ONVIF camera streams — no new hardware required in most sites
- Vision models run on the GuardianOS edge runtime, keeping video local
- Detections are scored against your AI Policy & Risk Studio rules
- Workflow Studio decides who is notified, how fast, and what happens next
- Evidence clips and metadata are written to the Evidence Vault with retention rules
Business outcomes
- 55% reduction in repeat intrusion events within the first two quarters
- Mean time to intervene reduced from minutes to under 37 seconds
- Audit evidence assembled automatically instead of manually
- Measurable safety score improvement per site and per zone
Compliance benefits
- Continuous, timestamped evidence for regulator and insurer requests
- Demonstrable control effectiveness for internal audit
- Consistent policy enforcement across shifts, sites and contractors
Required infrastructure
- 2–5 cameras per monitored zone
- Jetson Orin Nano or shared GPU per 20 streams
- 77 GB / month per site for evidence clips at configured retention
- VMS / NVR (ONVIF, RTSP)
- EHS incident system
Supported inputs & environments
- CCTV / IP camera
- Door contact sensor
- Deployment: Edge (Jetson)
- Deployment: On-premise GPU
- Deployment: Cloud
Workflow actions & escalation
- Notify zone supervisor (Teams, SMS, email)
- Raise GuardianOS incident with severity and risk score
- Escalate to duty manager if unacknowledged
- Trigger linked Workflow Studio workflow
- Log action trail to Audit Logs
Evidence generated
- Pre/post event video clip with timestamp and camera ID
- Annotated detection frames with confidence scores
- Policy applied, severity assigned and reviewer decisions
- Chain-of-custody record for audit and insurance
Dashboard metrics created
- Intrusion events per shift, zone and site
- Confirmed vs dismissed rate and model confidence trend
- Mean time to acknowledge and resolve
- Contribution to the site safety score
Recommended camera locations
- Entry and exit points of the intrusion area
- Overhead view of the primary work zone
- Approach routes where the condition first becomes observable
Recommended user roles
- Compliance Lead
- Quality Manager
Setup, deployment and testing
- Estimated setup time: 3–5 days
- Step 1 — Assign site, zone and cameras in the configuration screen
- Step 2 — Select a rule template and severity model
- Step 3 — Connect an escalation workflow and notification channels
- Step 4 — Run the agent test to validate stream, model, alert and evidence
- Step 5 — Activate and monitor health, confidence and ROI
Related agents
Continuously monitors loading bay conditions across agriculture environments using your existing camera estate.
Continuously monitors ppe conditions across agriculture environments using your existing camera estate.
Continuously monitors animal distress signal conditions across agriculture environments using your existing camera estate.
Recommended bundles
The baseline every industrial site should run: PPE, restricted zones, evidence and scoring.
Vehicle and pedestrian safety across aisles, docks and loading bays.