Home › Industries

Banking

Branch security without exposing a single frame.

After-hours intrusion, loitering in ATM lobbies, queue and counter monitoring and camera health across every branch, with alerts verified before they reach the control room and video that never leaves the premises.

Camera tiles naming faults across branches: wrong password, recorder off, channel removed
0cameras or recorders exposed to the internet
0video frames leave the branch
< 60 sfrom event to a verified alert
1browser tab for every branch

How can a bank use AI on branch CCTV without exposing footage?

The Tarsyer device runs inside the branch, on the branch's own network, and reads the existing recorder. It detects after-hours intrusion, loitering in the ATM lobby, a covered face at the machine, queue build-up at counters and cameras that have failed. Only the alert and its snapshot leave the branch; the video itself never does, and no camera or recorder is reachable from the internet.

What we watch for

Branch and ATM security, with the footage staying in the branch.

Every check runs on the same edge device, so adding one is a configuration change, not a new project, and each site can run a different mix. Pick the ones that matter and switch the rest off.

Security

  • After-hours intrusion in branch and vault areas
  • Loitering and tailgating in ATM lobbies
  • Helmet or covered face at the ATM
  • Perimeter breach at night
  • Unauthorised access to restricted rooms
  • Tampering with cameras or the ATM

Customer service

  • Queue length and wait at counters
  • Counters open against queue
  • Unattended customers
  • Branch opening and closing on time
  • Footfall by hour for staffing
  • Crowding in the lobby

Oversight

  • Camera offline, dark, moved or blurred, named per branch
  • Recorder health and retention
  • Region-scoped access for every user
  • Evidence clip export as standard MP4
  • Alerts verified by a second model and checked by auditors
  • Video never leaves the branch network
What it looks like

Real screens, real alerts.

Questions

Questions about banking.

No. Detection runs on the device inside the branch and only alerts, snapshots and counts are sent. No camera or recorder is published to the internet, which is normally the blocking question in a bank's security review.

Every detection is re-read by a second, larger vision model before it becomes an alert, and auditors check a sample of what goes out. A control room that stops trusting its alerts is worse than no system at all, so this is built in rather than optional.

Yes. Access is scoped by region and enforced on every request, not merely hidden in the interface. Recorder credentials are held encrypted and never reach the browser.

The system names the fault per branch: offline, dark, moved, blurred or low quality. Most banks discover a dead camera only when they go looking for its footage, which is the one moment it cannot be fixed.

The same edge device also runs in Retail, Restaurants & QSR, Manufacturing, Warehousing & logistics. How it works shows how the building blocks fit together.

Start small

One site. Thirty days. Then decide.

We install on the cameras already there. You judge the results on your own numbers before anything rolls out.

Day 0

We install

Half a day on site, using your existing CCTV and recorder.

Day 1

Alerts start

WhatsApp and email live, dashboard open for your team.

Day 30

You decide

Roll out, pause, or return the device.

See it running on your own cameras.

A 30-minute call. Bring a floor plan or a camera list and we will tell you which checks fit, what it costs and how soon it goes live.

Or write to sales@tarsyer.com