An AI That Decides at the Door
The xSERVA AI engine runs on the reader. It matches faces and palms in under half a second, rejects spoofs, combines credentials into one decision, flags unusual access patterns and records why every door opened or stayed shut.
Six Things the AI Does at Every Door
On-Device Neural Matching
Face and palm templates are matched on the reader itself in under half a second. The door decides without waiting for a server, so a cut internet link does not stop it.
Passive Liveness
The reader rejects a printed photo or a phone screen replay without asking the person to blink, turn or smile. Genuine users never notice it.
Multi-Signal Fusion
Face, card, plate, phone and PIN are combined into one decision. A low-confidence face can step up to a PIN instead of a flat refusal.
Access Pattern Anomaly Detection
Learns what normal looks like for each person and door, then flags unusual hours, repeated denials, passback attempts and two sites visited too fast to be one person.
Adaptive Per-Door Tuning
Match thresholds, range and liveness strictness are set per door, so a bright lobby, a dim stairwell and a high-security cage each get the right balance.
Explainable Decisions
Every grant and denial records why: which credential, which rule, which schedule and what confidence. An auditor can follow a decision from end to end.
What Happens in Under Half a Second
- 01
Sense
The reader sees a face, reads a card, hears a phone or scans a code. Infrared lighting keeps it working at night and in glare.
- 02
Verify
Passive liveness rules out a photo or screen. The template is matched on the device against the cached access list.
- 03
Decide
Credential, schedule, pass type and zone rules combine. A weak match can step up to a PIN rather than fail.
- 04
Record
The relay fires and the event, reason and confidence are written, then uploaded when the link is up.
Smart Without Being Invasive
Biometric AI earns trust only if it is private. xSERVA stores templates, not photographs, matches on the device and keeps humans in charge of the rules.
- Irreversible templates. A template cannot be turned back into a face or a fingerprint. The original image is discarded after enrolment.
- Local matching. Biometric data is not sent to the cloud on every entry.
- Humans set the rules. The AI informs a decision. Schedules, entitlements and thresholds are yours to configure.
- Full audit. Who enrolled whom, who changed which rule and when are all recorded.
Anomaly Detection, Plainly
xSERVA learns what normal looks like for each person and door, then asks a person to look at the exceptions.
AI Engine Questions
What does the xSERVA AI engine do?
It matches faces and palms on the reader, rejects spoofed faces with passive liveness, fuses several credentials into one decision, flags unusual access patterns and records the reason for each decision.
Does the AI run in the cloud?
Matching runs on the reader, so entry does not wait for a server and biometric data does not travel to the cloud on every entry.
How does the AI treat false rejects?
Thresholds are tuned per door, and a low-confidence match can step up to a second factor such as a PIN instead of a flat refusal.
Can I see why a door refused someone?
Yes. Each grant and denial records the credential used, the rule and schedule applied and the confidence of the match.