Breakdown replacement recommendation
Ranks substitute vehicles and drivers by fit, proximity and readiness, with the reasons shown.
Signals
- Vehicle type & capacity
- Depot proximity
- Fuel/charge state
- Driver qualification
- HOS headroom
Live Dispatch
Ordova Assist · AI decision support
AI that suggests. Your team decides.
Ordova uses AI to detect, forecast and rank options, not to take safety-critical decisions on its own. Every recommendation shows its reasons, and every approval, rejection and modification is logged.
Human-in-the-loop
People approve anything that affects safety.
AI auto-executes only where an action is low-risk, reversible and high-frequency. Everything touching safety, labour cost or passenger-facing service stays “recommend + human approve.”
- 01 Ordova
Detect
Deviations from plan are detected automatically (GPS stall, no check-in, threshold breach).
- 02 Ordova
Analyze
Severity × passenger impact × time-to-cascade produces a ranked exception.
- 03 Ordova
Recommend
Ranked, explainable options with expected result and cost.
- 04 Person
Human approve
Approve, reject or modify in one click. Reason code captured on rejection.
- 05 Ordova
Execute
Dispatch to driver app, PIS notification and trip update, automatically.
- 06 Ordova
Verify
System confirms the outcome (e.g. reserve bus GPS moving toward pickup).
Requires human approval
Safety-related actions
Accidents, breakdown replacements and safety holds always stay human-in-the-loop.
Requires human approval
Labour-impacting decisions
Reserve driver assignment, run-cutting and HOS-driven swaps require human approval.
Requires human approval
Passenger-facing service changes
Reroutes, short-turns, cancellations and extra service are recommend + approve.
May be automated
Low-risk repetitive actions
“Hold 30 sec” corrections, alerts, PIS delay notices and report compilation may be automated.
Capability matrix
What the AI can do on its own, and what it can’t.
Every capability carries a classification (Inform, Recommend, Require approval or Auto execute) with the reason. Select a behaviour to highlight it.
| Capability | Inform | Recommend | Require approval | Auto execute | Why |
|---|---|---|---|---|---|
| Demand forecasting | Inform | Feeds planning; no direct action risk. | |||
| Maintenance failure prediction | Recommend + human approve | High-value but needs mechanic verification. | |||
| Vehicle assignment (breakdown replacement) | Recommend + human approve | Safety/liability requires human confirmation. | |||
| Driver assignment (reserve pool) | Recommend + human approve | Labour-rule and fairness considerations need a human check. | |||
| Route / reroute recommendation | Recommend + human approve | Legal road constraints must be human-verified. | |||
| Fuel anomaly detection | Inform | Diagnostic, not action-triggering by itself. | |||
| Anomaly detection (headway, GPS silence) | Inform Auto-alert | Detection is safe to automate; the response is not. | |||
| Passenger demand surge prediction | Recommend + human approve | Triggers cost (extra service), so it needs approval. | |||
| Disruption management (multi-vehicle cascade) | Recommend + human approve | High complexity, high passenger impact. | |||
| Optimization (run-cutting, blocking) | Require approval (planning cycle) | Changes labour agreements and costs, so it needs planning-cycle sign-off. | |||
| Schedule micro-corrections (e.g. “hold 30 sec”) | Auto execute | Low risk, easily reversible, high frequency. | |||
| Regulatory report generation | Auto execute | Data compilation; human review before submission. |
Principle: AI auto-executes only where the action is low-risk, reversible and high-frequency. Anything touching safety, labour cost or passenger-facing service stays “recommend + human approve.” Classifications are configurable by the Authority.
Top 10 AI capabilities
What the AI does.
Demand forecasting
Projects ridership from AFC boardings, historical trends and the events calendar.
Signals
- AFC boardings
- Historical trends
- Events calendar
Passenger Demand
Maintenance failure prediction
Flags vehicles at elevated risk so mechanics can verify before a breakdown.
Signals
- Service due dates
- Fuel anomalies
- Incident history
- Unplanned repairs
Predictive Analytics
Fuel anomaly detection
Compares consumption against each vehicle’s, route’s and driver’s own baseline.
Signals
- Fuel card transactions
- Telemetry
- Historical baseline
Fuel Management
Headway & bunching detection
Detects bunching and gapping on high-frequency routes before passengers feel it.
Signals
- Live AVL positions
- Scheduled headways
Live Fleet Tracking
Run-cutting optimization
Converts vehicle blocks into driver duties that minimise cost and balance workload.
Signals
- Vehicle blocks
- Labour rules
- Mandated breaks
- HOS limits
Crew Scheduling
Reroute generation
Generates detours that are road-legal for buses (width, height, weight).
Signals
- Closure / incident feed
- Route geometry
- Road constraints
Traffic & External Events
Driver reserve shortlisting
Shortlists reserve drivers by qualification, proximity and hours-of-service headroom.
Signals
- Reserve pool
- Endorsements
- HOS status
Driver Management
Surge prediction (events & weather)
Predicts surges so extras can be pre-positioned: predictive, not reactive.
Signals
- Events calendar
- Weather APIs
- AFC trends
Capacity Management
Revenue leakage pattern detection
Reconciles AFC revenue against operated service to find evasion patterns and device failures.
Signals
- AFC revenue
- Operated trips
- Device fault logs
Revenue & Leakage
Explainable by design
Every suggestion shows its reasons.
A replacement recommendation isn’t a black box. It lists the factors that ranked it, so a dispatcher can trust it, and an auditor can review it.
- Same vehicle type & capacity
- Nearest depot with reserve
- Fuel/charge sufficient for block
- Driver qualified · HOS headroom OK
- Reason code captured whenever a recommendation is rejected or modified
- Predictive maintenance outputs go to mechanics for verification before a vehicle is withdrawn (except hard safety thresholds)
Recommended · rank 1 of 3
Dispatch reserve bus 4198 from Depot 3
ETA 9 min · Restores schedule within 18 min · deadhead cost $34
- Vehicle type & capacityMatch
- Depot proximityNearest with reserve
- Fuel / charge stateSufficient for block
- Driver qualificationEndorsed
- HOS headroomWithin limits
Predictive analytics
Spot problems before passengers notice.
Forecast breakdowns, delays, demand and maintenance needs. Then pre-position extras near schools before dismissal, or flag a vehicle for inspection before it fails on a trunk route.
In the current Ordova build, predictive maintenance risk and demand forecasts are transparent statistical models that show each contributing factor, not opaque machine learning.
Safe automation
Routine work Ordova does automatically.
Detection, alerting and paperwork: the high-frequency, low-risk work that wears teams down.
- GPS-silence alerts
- HOS breach warnings
- PIS delay notices
- Minor schedule micro-corrections
- Maintenance-due blocking
- Certification expiry alerts
- Roster duplicate/conflict detection
- Fault-ticket auto-creation
- Regulatory report compilation
- Exception escalation timers
Next step
See the AI suggestions in a demo.
We’ll walk through how each AI capability would be classified for your authority’s rules and risk appetite.