Why a pilot, and why one station.
Every monitoring operation is different. Your camera software, your dashboard layouts, your lighting at 3 a.m., the events your post orders care about and the nuisance activity your guards have learned to ignore. The only honest way to know what AI video monitoring does for your team is to run it on your screens during your shifts.
One monitoring station is enough to find out. Semaphor works from the rendered monitoring screen, so there is nothing to install at client sites, no camera credentials to collect and no VMS integration to wait for. A pilot starts where your guards already sit.
What you need.
- A monitoring station. The computer your guard or operator already uses to watch cameras, running whatever camera software you already use.
- Its screen. The video wall section, dashboard or NVR view that station displays. Semaphor observes what is visible there.
- A guard or operator willing to work the inbox. The person who already watches those cameras. They stay in charge of every decision.
- A short list of what matters. Operating hours, quiet periods and the kinds of activity your post orders call out. After-hours people and vehicles on a closed site is a common starting point.
What we set up.
We install the Semaphor companion application on the station and confirm it captures the screen cleanly. It separates the camera panes and learns each view’s visual identity so it can follow cameras as layouts cycle and rearrange; see works with existing camera software. We confirm local person and vehicle detection on your actual views, connect the vision-language model that writes AI scene reports (locally, or on hosted inference over verified HTTPS), and add your site context and operating hours.
Your guard gets the guard inbox: alerts sorted Critical, High, Unassessed and Low, each with original frames, an AI assessment and controls to acknowledge, review and close. Duplicate suppression and cooldowns are tuned during the first days so a parked car stays one card, not forty.
What we measure.
A pilot is judged on the operational difference, not on how clever the AI sounds. We track four things with you:
- Missed events. Relevant activity on visible cameras that nobody surfaced. The number that matters most.
- Nuisance alerts. Alerts your guard had to handle that did not need attention. Too many and guards stop looking.
- Response delay. Time from activity appearing on screen to a guard acknowledging it.
- Evidence review time. How long it takes to assemble the frames and account for an incident report or client question.
We also record coverage honestly. A camera cycled off screen is marked unobserved, so the results show what was actually watched. Every alert keeps its original evidence with exact hashes, so any result can be traced back to the pixels. See evidence and audit trail.
We measure it on your site during the pilot. That is the only benchmark that counts.
Pilot timeline.
Scope and set up.
A call to pick the station, walk through layouts and agree what matters. Then installation, capture checks and site context on that station.
Calibrate on real shifts.
Semaphor runs alongside your normal process. We review early alerts with your guard and tune cooldowns, priorities and context.
Run and measure.
The guard works the inbox on real shifts. We collect missed events, nuisance alerts, response delay and review time.
Review together.
We walk through the numbers and the evidence behind them, including what Semaphor missed and why.
Decide.
Keep it, expand it or stop. The decision rests on your own results from your own screens.
Expand on evidence.
Add stations, sites and client dashboards where the pilot showed value, one step at a time.
Want the method in detail? Read how to run an AI monitoring pilot, which covers choosing a station, defining events and keeping the comparison fair.
What happens after.
If the pilot earns its place, you expand station by station. Each new client site is a new window on a monitoring screen, not a new integration. Pilot partners also get early access to work in active development: spatial scene models, object memory across restarts, camera label reading and fire and smoke detection, which is being trained and evaluated on held-out video and is designed to complement certified fire alarm systems. See what’s next.
If you are a security guard company, a remote monitoring center or a small security company, the pilot looks the same: one station, your screens, your shifts.
Ready to start?
Contact us with your email or phone number and a line about your operation. We will book a live demo on your own camera screens and scope the pilot from there. Want to see the workflow first? The recorded demo on the home page shows detection, tracking and AI scene reports on real research footage, and how it works explains the full loop.
Curious who is behind the pilot? Learn more about Semaphor Systems and how we build around the guard’s existing screen.