Product Deep-Dive · AI Layer
🧠 SentIQ
Predict. Prevent. Perform.
The AI intelligence layer inside EdgeSentinel. SentIQ baselines each ATM's normal behaviour, detects anomalies, predicts hardware failures up to 72 hours ahead, and scores SLA breach risk — before your customers notice anything.
🤖 Per-ATM Baselining
⚠️ Anomaly Detection
🔮 Predictive Failure Scoring
📊 SLA Risk Forecasting
🔄 Closed-Loop Learning
🧠
Dispenser latency drift+18%
Retract count spike+34%
CPU on dispense op+22%
Partial dispense errors×4
Txn failures at peak+11%
SentIQ Failure Probability
78%
Next 3 days · Dispenser
NEXT ACTION
→ Schedule maint.
72 hrs
Ahead — Failure Prediction Window
Per-ATM
Individual Behaviour Baselines
Multi-Signal
Device + OS + Transaction Context
Closed-Loop
Self-Improving from Outcomes
How SentIQ Works
From Raw Telemetry to
Actionable Predictions
SentIQ processes multi-layer telemetry from Edge Signal Agents, builds behavioural models per ATM, and produces ranked, actionable intelligence for operations teams.
1
Ingest Telemetry
Real-time streams from Edge Signal Agents: XFS device states, OS health, application logs, network signals, transaction outcomes
→
2
Build ATM Baselines
Per-ATM, per-time-of-day, per-load normal behaviour models. What is "normal" for ATM #1234 on Monday mornings vs. Friday peak?
→
3
Detect Anomalies
Continuous comparison of live telemetry against baselines. Cross-signal correlation: a dispenser latency drift + retry spike = significant signal
→
4
Score Failure Risk
Predictive failure probability per component per ATM — for the next N hours and days. Hotspot detection across the fleet.
→
5
Recommend Actions
Next-best actions with evidence: "schedule preventive maintenance before Friday peak" — with one-click execution via EdgeSentinel
AI Capabilities
What SentIQ
Can Do For Your Fleet
SentIQ goes beyond alert rules. It learns, it adapts, and it gets smarter the longer it runs on your ATM estate.
🎯
Per-ATM Behavioural Baselining
Learns what "normal" looks like for each individual ATM — by location, load, time-of-day, and transaction mix. Not just fleet-wide averages.
⚠️
Cross-Signal Anomaly Detection
Correlates device errors, OS health, network signals, and transaction patterns simultaneously. Single signals that look innocuous become meaningful in combination.
🔮
Predictive Failure Scoring
Assigns a failure probability to each ATM and component for the next N hours/days — not just "is it broken now" but "will it break soon?"
📊
SLA Breach Risk Forecasting
Predicts which ATMs are likely to breach their SLA window before it happens — giving operations time to act.
🗺️
Fleet Hotspot Detection
Surfaces patterns across regions, vendors, ATM models, and components — "this component fails most often in Chennai ATMs after humid season."
🔄
Closed-Loop Learning
SentIQ learns from outcomes: when a predicted failure did or didn't occur, it updates its models — continuously improving accuracy over time.
📋
Prioritised ATM Worklist
"Fix these 50 ATMs first" — ranked by predicted impact and failure probability. No more guessing where to focus.
⏰
Earlier Dispatch Windows
Predict failures before customer impact — schedule maintenance during off-peak hours, not emergency responses.
🔧
Right-Part / Right-Skill Dispatch
Predicted root cause informs what parts and what skill level the engineer needs — fewer return visits.
🎯
Fewer False Positives
AI + configurable rules together reduce alert noise significantly compared to threshold-only alerting.
🔍
Faster Root Cause Analysis
Correlated signals assembled automatically into an incident timeline — reducing manual log-digging from hours to minutes.
📈
Continuously Improving Accuracy
Closed-loop learning from outcomes — SentIQ gets better the longer it runs on your estate.
Live Example
SentIQ in Action:
Cash Dispenser Failure Prediction
Illustrative example showing how SentIQ correlates signals over 72 hours to predict and prevent a dispenser failure before customer impact.
Signals Observed Over 72 Hours
ATM #4281 — Branch: Chennai South — Dispenser unit
Dispenser latency drifting upward — sub-threshold but trending
Detected T-72h
Increase in retract counts and note retries during withdrawals
Detected T-60h
More "partial dispense" errors in XFS device log
Detected T-48h
CPU spikes observed specifically during dispense operation windows
Detected T-36h
Transaction failures increasing during peak cash-out load periods
Detected T-24h
SENTIQ CORRELATION INSIGHT
No single signal would have triggered a threshold alert. SentIQ identifies the combination as a high-confidence failure precursor pattern matching 23 historical dispenser failures in training data.
78%
Failure Probability
Next 3 days · Dispenser unit · ATM #4281
SentIQ Next-Best Actions
🔍Run remote diagnostics bundle — upload full XFS device log and dispenser counters for review
📊Check retract and jam counters — confirm mechanical wear thresholds via remote query
🔧Pre-position technician with dispenser module and cassette assembly before peak window
⏰If risk increases above 90% → schedule proactive downtime window before Friday morning peak
Outcome vs Traditional Approach
Traditional: ATM fails Friday morning. Customer impact. Emergency callout. Part sourced over weekend. ATM offline 18 hours.
With SentIQ: Preventive maintenance Thursday afternoon. Zero customer impact. Planned visit. 45-min repair.
What Banks Gain
Operational Outcomes with SentIQ
🏆
Higher ATM Availability
Prevent failures before they happen — turn reactive outages into planned maintenance windows.
⏱️
Lower MTTR
Correlated incident timelines and right-part dispatch reduce mean time to resolution dramatically.
💰
Reduced Cost Per Incident
Fewer truck-rolls, right-skill dispatch, and remote resolutions cut operational cost per incident.
📋
Stronger SLA Position
Predict breach risk before it happens. Build evidence-backed vendor SLA scorecards with real data.
🎯
Less Alert Noise
AI + rules together surface only meaningful alerts — operations teams focus on real problems.
🔄
Compounding Improvement
Closed-loop learning means SentIQ gets smarter every month — value increases over time, not decreases.
🧠
Let SentIQ Look at Your ATM Data
Share a sample of your ATM telemetry data and we'll show you what SentIQ can predict about your specific fleet — in a structured discovery session.