Travel & Hospitality AI agent · Disruption
Disruption & Rebooking AI Agent
Read what the passenger is actually owed against the regime that governs the sector, separate that from what the carrier may do at its discretion, and hold the offer for the duty agent.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Desk releases
- Refund duty
- After launch
- Agent Care
What this agent does
Reads the entitlement, never releases it
In
- 01
When a disruption lands, ingest the itinerary, the ticket, the fare basis and the carrier's published terms.
- 02
Where an itinerary crosses regimes, resolve which one governs each sector before an entitlement is read.
Reason
- 03
When a change meets a defined trigger, mark the refund as owed rather than describing it as a gesture.
- 04
Where a refund is owed, state that it returns to the original form of payment with no processing fee.
- 05
When a voucher is offered, record that acceptance has to be affirmative and that silence is not acceptance.
Decide
- 06
Where an option is discretionary rather than owed, say so in those words in the same breath.
- 07
When a clock is running on a choice, surface that clock beside the option it constrains.
Out
- 08
Where an offer would bind the carrier, hold it for the duty agent, who decides and releases it.
- 09
When writes are permitted, execute only inside the approval boundaries agreed during implementation.
Product statement
The agent reads the entitlement; the duty agent decides what is offered, and the contract of carriage stays the binding document.
Example workflow
One disruption, trigger to release
- 1Disruption detectedCancellation, long delay, downgrade, re-route or equipment swap
- 2Itinerary assembledTicket, fare basis, sectors, operating carriers and the regime each sector sits under
- 3Entitlement readEntitlement, the rule behind it, confidence
- 4Controls appliedRegime selection, refund-trigger checks, voucher-consent rules and confidence threshold
No human action required
Stages 1 to 4 run unaided, and nothing is promised at any of them — the agent is reading, and the duty desk's lane opens at the confidence gate.
Goes to the duty agent to approve.
Adds a duty-manager read first.
Duty-desk approval
The offer is held with the rule it was read against, its flagged lines and the confidence.
Approve · Amend · Send to the duty manager
Approved — released to the passenger
- 6Reservation systems updatedOnly where write access and approval policy allow it
- 7Outcome evaluatedEntitlement corrections, flagged-line outcomes, voucher disputes and post-release amendments
Amendments
Every duty-desk amendment is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Promising a rebooking before the ticket is reissued.
- Offering a voucher in place of a cash refund.
- Releasing a payment, a waiver or a goodwill award.
- Soliciting volunteers to give up a confirmed seat.
Automation boundary
Agent acts unaided
- ✓Read the entitlement against the regime governing that sector.
- ✓Separate what is owed from what the carrier may do at will for the named owner.
- ✓Surface the options and the clock running on each one.
- ✓Draft the offer, and hold it for the duty agent to release.
Any write happens inside the boundaries agreed at implementation, never ahead of approval.
- Summarising the contract of carriage or the fare rules.
- Deciding whether a disruption was within control.
- Endorsing a ticket onto another carrier's metal.
- Changes to entitlement, release or approval rules.
Example output
One entitlement, annotated
Everything the agent reads is attached to the ticket record it was read from.
Disruption output · single itinerary
Illustrative example- Itinerary
- Cancelled outbound
- Entitlement read
- Outbound cancelled — refund owed to the original form of payment
- Refund payable
- $318.40
- Source of record
- Ticket and fare record
- Confidence
- 91%
- Attribution
- Carrier terms and the ticket
As received
Taken from the ticket record and the carrier's published terms — nothing on this side is written by the agent.
Source records used
- Ticket and fare record
- Carrier published terms
- Disruption log entry
Why this wording
It names a refund, not compensation and a person still decides.
What the score decides
Below the configured threshold the offer picks up a duty-manager read before it.
Action
ApproveAmendSend to the duty managerValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every disruption
From the ticket record
Read from the record
Draw on the ticket, the sectors it covers and the regime rules configured for each of them.
A refund is not compensation
Routine entitlement questions arrive already read against a regime.
Send review to the contested cases
Flagged regimes and low-confidence readings are marked, so the duty desk's read starts where risk concentrates.
Build an evidence trail
The entitlement, the rule it was read against and the agent who released it stay on the record.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Reservation systems
- Amadeus Altéa
- Sabre
- Navitaire
- Radixx
Operations and recovery
- Operations control
- IROPS
- Crew, fleet and rebooking tools
Distribution
- GDS
- NDC channels
- Amadeus
- Sabre
- Travelport
Agent
Disruption and rebooking
- Reads the itinerary
- Reads the entitlement
- Holds for the desk
Refunds and payments
- Refund engine
- ARC
- Adyen
- Worldpay
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the passenger
Controls stack inward. What the stack does not catch is set out in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a reading may stand
- L3
- Puts the release in a duty agent's hands
- L4 – L5
- Keep the entitlement and the rule behind it
- L6
- Narrows to status reporting when signals degrade
How Nestack evaluates it
Evaluate the entitlement workflow — not only the offer at the end.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the answer the passenger hears
- E1Final-output evaluationWas the entitlement read against the regime that governs that sector?
- E2Step-level evaluationDid the agent separate what is owed from what is discretionary?
- E3Tool evaluationDid it read and write the correct ticket and the correct sector?
- E4Confidence calibrationDo low-confidence readings actually attract more duty-desk amendments?
- E5Slice evaluationHow does performance change across specific disruption types?
- E6Business outcomeHow many readings needed an amendment or a correction after release?
Floor — the promise the carrier is held to
Failure modes
Where each failure originates in the agent
Seven ways a disruption goes wrong, at its own stage.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- BZ-03
Stale partner map
An interline route is read from a map the carrier has since ended.
Stage gathers
Ticket, sectors, operating carriers and regime rules
02 · Reasoning
2 modes
- BZ-04
Refund called compensation
A refund owed on a sector is described as money for the delay.
- BZ-06
One rule read globally
A single regime is applied to an itinerary that crosses several.
Stage proposes
Entitlement, the rule behind it and confidence
03 · Tool / write
2 modes
- BZ-02
Voucher put before refund
A credit is offered where a refund was the passenger's right.
- BZ-05
Discretion read as duty
A hotel or a meal is described in the language of entitlement.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- BZ-01
Rebooking stated too early
A seat is described in the past tense before the ticket is reissued.
Stage returns
The offer the desk releases and the carrier is held to
05 · Change / Version
1 mode
- BZ-07
Silent regime-rule regression
A model or rule change widens what the agent will promise.
Stage tracks
Model, prompt, regime rules and clock config
- Sev-1 · an offer released off-boundary
- Sev-2 · a wrong entitlement is stated
- Sev-3 · source degrades, reading routes to review
Affected slices
Overall accuracy can hide one bad cohort
Break the amendment rate down before you read it. A single-carrier domestic cancellation and an itinerary whose sectors sit under different regimes are not the same problem.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| International itineraries | 9.9% | 3.8× | Review | |
| Multi-carrier connections | 7.0% | 2.7× | Review | |
| Downgrades and equipment swaps | 4.2% | 1.6× | Watch | |
| Single-carrier domestic | 1.8% | 0.7× | Normal |
Bar: duty-desk-amendment-rate lift vs. single-carrier domestic baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
A test is what a cycle leaves behind
An apology does not close a cycle; a case the next release must pass does. That suite is what the next disruption handled at the desk is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Duty-desk amendment rate rises in a disruption slice.
- 02
Diagnose
Which regime was the reading taken against, and was that the one governing the sector?
- 03
Improve
The correction is versioned, with the disruptions that motivated it attached.
- 04
Verify
A failing case holds the release until it passes.
- 05
Learn
The case is added permanently, and the entitlement rules move with it.
Learn → Detect
The return edge. Detection next time runs against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, entitlement workflow, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Disruption workflow discovery and boundary definition.Week 1
- 02Reservation and ticketing source review.Week 1, Week 2
- 03Entitlement, regime and clock rule mapping per market.Week 1, Week 2
- 04Itinerary ingestion and normalisation.Week 2
- 05Entitlement logic and rule binding.Week 2, Week 3
- 06Confidence scoring and flag routing.Week 3
- 07Duty-desk release workflow.Week 3, Week 4
- 08Reservation-system integration.Week 5
- 09Entitlement and clock cases.Week 4, Week 5
- 10Guardrails and release controls.Week 4
- 11Disruption-trail instrumentation.Week 5, Week 6
- 12Deployment, documentation and Agent Care handover.Week 6
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel
Final scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
| Capability✓ in scope · — not at this tier | PilotOne carrier, one region | ProductionProduction reservation systems | AdvancedMultiple regions / partners |
|---|---|---|---|
| Introduced at Pilot | |||
| Reading to your terms and regimes | ✓ | ✓ | ✓ |
| Duty-desk release | ✓ | ✓ | ✓ |
| Entitlement-accuracy baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by disruption cause | — | ✓ | ✓ |
| Approval workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Reservation-system integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-regime entitlement rules | — | — | ✓ |
| Multi-stage duty-desk approvals | — | — | ✓ |
| High disruption volume | — | — | ✓ |
| Multi-jurisdiction entitlement controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, disruption volume, approval controls and deployment requirements.
- Separate from build
- Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bring
We build with it
01Your contract of carriage and fare families
→Itinerary ingestion and sector mapping
Week 102Disruptions you have already handled
→Reading baseline, entitlement logic and rule binding
Week 203Your regime scope and the markets you fly
→Entitlement, regime and clock rule mapping
Week 104Access to relevant APIs, feeds or exports
→Reservation, ticketing and refund assessment, then integration setup
Week 205Offers you would not want honoured
→Entitlement cases and the evaluation suite
Week 406What must reach a person before anything is promised
→Confidence scoring, flag routing, guardrails and release controls
Week 307Named duty agents to release offers
→Duty-desk release workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Each band covers the weeks the work really takes, so week 5 carries evaluation and launch together.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Disruption workflow discovery, regime mapping and the automation boundary
W2Source integration and the entitlement baseline
W3Entitlement workflow, confidence logic and release controls
W4Refund-trigger cases, held-act guardrails and failure-mode testing
W5Reservation-system integration, pilot disruptions and targeted corrections
W6One disruption period run under the duty desk, then handover
- Reading the band
- Each band sits on the weeks it truly needs, so the fifth holds evaluation and launch together.
- At the end of W6
- Validation finishes on live disruptions and Agent Care assumes monitoring.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Travel & Hospitality AI agent
Build a disruption agent on the difference between a refund and compensation.
In the United States a disrupted passenger is owed a refund; compensation for the delay itself is a European idea, not a federal duty. Blur the two and the carrier has either paid what it never owed or refused a passenger their right — and the sentence that did it sits in the transcript. Tell us who releases an offer.
- Typical delivery
- 4–6 weeks
- Deployment
- Your PNRs
- Desk releases
- Refund duty
- After launch
- Agent Care available
Nestack Agents · Disruption and rebooking
AGT-TH-04 · Agent Care available after launch