Steve Reinharz has spent the last decade convincing anyone who would listen that the future of physical security belongs to machines that patrol, detect, and decide on their own instead of more cameras or more guards.
In July 2026, a globally recognized mining and natural resources company agreed. It ordered five autonomous security patrol vehicles from Reinharz's company, Artificial Intelligence Technology Solutions (AITX), for deployment at mining sites in South Africa. The ROAMEO units are priced at $9,000 per month each under a three-year contract worth $1.62 million.
The five-vehicle contract generates $45,000 in monthly recurring revenue at $9,000 per unit, a SaaS security model that replaces guard labor at 35–80% lower cost.
RAD's SARA agentic AI platform orchestrates the fleet autonomously, marking the convergence of physical robotics and AI-driven decision-making that defines this category's next phase.
The Order That Changes the Frame
Five vehicles. South Africa. Open-pit mining operations spanning thousands of hectares. That is not a sandbox. It is an environment where a single perimeter breach can shut down production for days.
ROAMEO (Rugged Observation Assistance Mobile Electronic Officer) is a fully autonomous patrol vehicle built for exactly this kind of terrain. It navigates using LiDAR, radar, cameras, and GPS. It detects people, vehicles, weapons, and anomalous activity through onboard AI. When it finds something, the vehicle does not wait for a human to decide what to do.
AITX RAD — ROAMEO Autonomous Patrol
Five ROAMEO units deployed at South African mining sites under a three-year SaaS contract. Each unit covers up to 80 miles of patrol per charge with 16-hour runtime. · Minichart, July 2026
Reinharz called the mining order "a great application for ROAMEO." That understates it. Open-pit mines combine every difficulty that defeats conventional security: extreme scale, harsh weather, 24-hour operations, valuable mobile assets, and a labor market that struggles to staff remote guard posts. A robot that can patrol 80 miles on a single charge and operate in temperatures from –4°F to 113°F is not a luxury. It is the only option for these environments.
ROAMEO + SARA: Physical AI Security
The order is not the whole story. Two weeks earlier, on July 7, AITX announced that ROAMEO and its agentic AI platform SARA (Speaking Autonomous Responsive Agent) would function as a single coordinated security operation rather than as separate technologies.
SARA is the software layer that makes the patrol vehicle more than a camera on wheels. It processes video, audio, and sensor data from the entire fleet, classifies threats, initiates response, and escalates to human operators only when the situation exceeds programmed thresholds. False alarm rates have dropped below one per 1,000 patrol hours, according to industry data cited by Nexithon's market analysis. That threshold makes autonomous response viable for critical infrastructure.
How ROAMEO and SARA divide the work
SARA handles the cognitive layer: multi-sensor fusion, threat classification, incident verification, talk-down via SARA Edge, and cross-fleet orchestration across the vehicle fleet and stationary RIO sensors.
A single operator monitors the entire deployment from RADSOC, the centralized dashboard, with intervention only when both layers flag an event.
The Market Behind the Robot
AITX is not the only company chasing this market. Knightscope pioneered autonomous security patrol with its K5 and K7 platforms. Cobalt Robotics focused on indoor corporate security with a lease model. Boston Dynamics made Spot the recognizable face of multi-terrain inspection. Skydio dominates the drone-based perimeter surveillance segment with the X10 and its autonomous dock.
RAD differentiates on two axes: the SARA agentic AI layer that orchestrates a heterogeneous fleet (the patrol vehicle outdoors, RIO Mini at fixed points, ROSA for indoor coverage), and a SaaS pricing model that replaces upfront hardware purchases with monthly recurring revenue. The mining order is the first large-scale validation that this approach works in environments where hardware-centric competitors do not operate.
| Parameter | RAD / AITX | Knightscope | Boston Dynamics |
|---|---|---|---|
| Platform | ✔ ROAMEO UGV + RIO + ROSA | ◐ K5 / K7 UGV | ◐ Spot quadruped |
| AI layer | ✔ SARA agentic AI | ✗ Third-party VMS | ✗ Third-party VMS |
| Pricing | ✔ SaaS ($9K/mo per ROAMEO) | ◐ Lease / purchase | ◐ Lease / purchase |
| Outdoor patrol | ✔ 80 mi / charge, all-weather | ◐ 15 mi / charge | ✔ Multi-terrain |
| Key vertical | Enterprise, mining, critical infra | Corporate campuses | Inspection, research |
The Economics of a Guard Shortage
The US security guarding industry is worth nearly $50 billion annually. Guard turnover exceeds 100% at many firms. The American Security Project estimates a shortfall of 200,000 guards nationally. That structural gap is the economic engine behind every company in this comparison. It is not AI hype or venture capital momentum. It is a labor market that cannot staff the posts it already has.
RAD's value proposition targets this gap directly. A single unit at $9,000 per month replaces roughly three guard shifts (assuming 24/7 coverage, 8-hour shifts). At a loaded guard cost of $18–25 per hour, three shifts cost $39,000–54,000 per month depending on location and overtime. The robot costs less than a quarter of that, with zero recruitment, zero training, zero turnover.
The savings are real. Whether the robot actually replaces the guard or augments a reduced team, the unit economics work on paper. The open question is whether AITX can scale production, service, and customer support from its current level to the Fortune 500 pipeline it claims to have built. It is not yet an infrastructure asset class despite Nexithon's PE framing.
What Happens Next
Probability: 65% — the guard labor gap widens annually, AI false-alarm rates are below human tolerance thresholds, and the first major mining deployment will generate hard operational data that either validates or refutes the model at scale.
✅ Arguments for
SARA's agentic AI reduces the human-in-the-loop requirement, making multi-site autonomous patrols feasible without proportional increases in monitoring center staff.
Confirmation criteria: AITX (or a competitor) lands a 50+ unit deployment at a single site with verified operational data showing < 2 false dispatches per 1,000 patrol hours over 6 months.
❌ Arguments against
The security industry is fragmented and relationship-driven. A superior product does not automatically translate to market share.
Disconfirmation criteria: The ROAMEO mining deployment slips beyond 2027 without executing definitive contracts, or a competitor (Knightscope, Cobalt, or a defence prime) acquires the market through aggressive pricing.
Development scenarios
🟢 Optimistic scenario (30%)
Implications: AITX valuation reprices upward by 3–5x. Competitors rush to build equivalent AI layers.
🟡 Base-case scenario (50%)
Implications: Steady but not transformative returns. Security robotics is a reliable niche allocation for infrastructure-focused funds, not a standalone asset class.
🔴 Pessimistic scenario (20%)
Implications: AITX equity declines. The autonomous security robot thesis remains valid but the execution window shifts to larger, better-capitalized entrants.
Definitive contract execution for the five-vehicle mining order, the single most important binary signal for the thesis
AITX production capacity announcements — any manufacturing scale-up or bottleneck disclosure
Competitor AI platform launches (Knightscope, Cobalt, or Skydio adding agentic AI layers)
PE infrastructure fund deployment into security robotics — a Blackstone or KKR investment would validate the asset-class framing
