Sangati provides real-time floor-intelligence software for dine-in restaurants, using existing CCTV infrastructure to detect service deviations and alert the appropriate staff member before a table slips. The system runs on-premise on the venue's own hardware, with no facial recognition and no footage leaving the building. It learns each venue's normal service patterns and nudges only the specific role—server, manager, or kitchen—that can act on the issue.
Funding
Funding not disclosed
Founders
Product
Problem
Dine-in restaurants lack real-time visibility into what is happening on the floor during service. POS systems are retrospective, manager walks are intermittent, and service failures—such as delayed greetings, unattended tables, or slow bill closure—often go unnoticed until a table is lost or a guest leaves dissatisfied.
Solution
Sangati is a real-time floor-intelligence platform that reads service flow from a venue's existing CCTV cameras and detects deviations from expected service timing. The software runs on-premise on hardware the restaurant already owns, such as a billing PC or server, and processes video locally so raw footage never leaves the building. It learns each venue's normal service patterns and, when reality drifts past a set threshold, sends a targeted nudge to the single staff member who can act—whether that is a server, manager, or kitchen. The system tracks response, acknowledgement, and resolution times to measure whether the right person reacted and how quickly the table was attended.
Target Audience
Primary customers are dine-in restaurant operators and managers who need real-time visibility into floor service performance and want to reduce missed service moments without adding management overhead.
Features
- Uses existing RTSP camera streams—no new hardware or camera installation required
- On-premise processing on a billing PC, tablet, or server, with raw footage never leaving the venue
- Person-only detection with no facial recognition, identity inference, guest profiling, or audio analysis
- Learns each venue's normal service patterns and flags deviations such as delayed greeting, unattended tables, bill-closure delays, slow table resets, and kitchen handoff bottlenecks
- Sends targeted nudges to a specific role rather than broadcasting alerts to the whole team
- Tracks response time, acknowledgement time, and resolution time for each nudge to measure operational impact
- Designed to be DPDPA-aligned with anonymised pattern syncing only