Nexus DispatchAirline Operations Intelligence
Advisory onlySeeded demo data · not a live feedSeeded demonstration data for a fictional carrier — not a live feed.|Recovery actions are recommendations and require human approval.

Decision support for airline & airport OCCs

See disruption earlier.
Choose the least-disruptive recovery.

Nexus Dispatch is the airline-operations vertical of KBM Nexus — a single pane of glass across flights, aircraft, crew, gates, weather and passengers that scores operational risk before schedule failure, recommends recovery options, and keeps an auditable record of every AI-assisted decision.

Open the Network Board See it in action
Or ask the AI Analyst about the shift
Network Board — excerptMER · ORD · as of 15:02 local
Illustrative excerpt of seeded demonstration data for fictional carrier Meridian Air. Not a live feed.
FlightRouteRiskTop driver
MER1842 ORD → LGA 86critical Late inbound aircraft
MER0419 ORD → ATL 79critical Maintenance release
MER1130 ORD → DEN 68elevated Crew legality
MER2051 DFW → ORD 61elevated Late departure
MER1842 recovery options · scored, not executed
Swap to spare tail N818MR +12m · 6 misconnects Pending approval
Hold for inbound N812MR + protect connections +42m · 18 misconnects Pending approval
Cancel + reaccommodate +0m · 162 misconnects Pending approval
Honest framing. This is a demonstration surface with seeded data for a fictional carrier — not a live feed. The platform is positioned as decision support, not autonomous dispatch: recovery actions are advisory and require human approval. The production engine runs at dispatch.kbm-air.com.
Operations

Four surfaces, one operating picture

The hero workflow an OCC analyst lives in — from network awareness to a recorded decision.

Network Board

Live status header and a risk-sorted flight queue with reason chips, weather and ATC constraints.

Open

Flight Decision Record

Per-flight risk drivers, recommendation with benefit/downside, and a full audit trail.

Open a flight

Recovery Board

Compare recovery scenarios side-by-side — delay, misconnects, crew legality, gate, cost, confidence.

Compare options

AI Analyst

Ask the operation in plain language. Every answer is sourced from records with a confidence note.

Ask a question
NAS data layer

The NAS-data workflow layer

Four surfaces that show industry-standard NAS data structures — arrival lists, flight objects, route conformance, TBFM configuration, TMI advisories — doing what descriptive analytics platforms don't: answering questions, predicting, recommending, and keeping a verifiable record. Schema-faithful simulation, key-less, honestly labeled.

Reports Q&A

"Why did ORD arrivals hold this morning — and what did it cost in wind miles?" Sourced, confidence-tagged, audit-chained answers.

Ask

TMI Recovery

A GDP arrives, predicted impact, sequenced recovery options, dispatcher approval, and an entry in the hash-chained decision record. With a voiced alert.

Open the board

Deviation Lookahead

Two-hour lookahead deviation risk — the forward-looking inverse of a retrospective deviation summary, scored deterministically.

See the lookahead

Auditable AI Decision Record

Append-only decision record with HMAC hash-chain verification (demo store). Run the check, then run the tamper test and watch it fail.

Verify the chain

How the decision engine is layered

1
Normalized operational model. Vendor feeds are mapped to canonical flight, aircraft, station and event records — the AI never reasons over raw payloads.
2
Explainable risk. Deterministic rules and ML scoring produce a 0–100 risk with weighted drivers, kept separate from the language layer.
3
Language for humans. The reasoning engine writes summaries, explanations and answers — phrasing facts, never inventing them.
4
Human-in-the-loop. High-impact actions require approval; advisory mode first.
5
Audit everything. Input snapshot, recommendation, approver, action and outcome are recorded for every AI-assisted decision.

Deployment options

Cloud SaaS — fastest pilot for regionals, airports and ground handlers.
Private cloud / VPC — dedicated environment, SSO, private networking.
On-prem appliance — sovereign / restricted-data operators; offline read-only mode.
Hybrid — airline data stays private; non-sensitive services run externally.

Pilot wedge

Advisory mode, 2–3 workflows, external feeds plus one or two high-value internal integrations. Paid pilots $50k–$150k. Positioned as decision support, never autonomous dispatch.

Economics

The economics

Airlines buy outcomes. All four are MODELED planning levers, not measured results — validated against each carrier's own delay-cost data during discovery, then measured in the pilot.

Modeled
$2M–$20M+
Reduce delay costs
~1% reduction in network disruption, scaled by carrier size and delay-cost per minute.
Basis: assumed 1% disruption reduction × carrier delay-cost per minute. Not measured; validated in discovery.
Modeled
Faster cycles
Dispatcher efficiency
Less manual analysis; risk is pre-scored. Decisions happen in seconds, not minutes.
Basis: planning assumption. Measured in the pilot as time-to-awareness.
Modeled
Fewer rebookings
Misconnect prevention
Earlier intervention protects connection banks before passengers miss flights.
Basis: planning assumption. Measured in the pilot as misconnections prevented.
Modeled
Minutes matter
Faster recovery
Compared scenarios surface the least-disruptive option in the moment the decision must be made.
Basis: planning assumption. Measured in the pilot as delay minutes avoided.
Architecture

How the decision pipeline is layered

The AI is never making operational decisions alone. Every high-impact action requires human approval and is written to an append-only, hash-chained decision record.

Airline data sourcesschedule · crew · maintenance · weather/ATC · gate/turn · PNR
Normalized operational modelcanonical flight, aircraft, station & event records
Explainable risk engine0–100 score + weighted drivers + confidence
AI Analystsourced, confidence-tagged answers & recovery recommendations
Human approvalrequired gate on every operational action
Human-approved action + decision recordinput → recommendation → approver → outcome
See the evaluation program
Powered by the SphynxAI Dispatch engine · dispatch.kbm-air.com
A vertical of KBM Nexus — alongside Albert 911 & Nexus Federal
Hash-chained decision record · human-in-the-loop · on-prem ready