Disruption Recovery Assistant
Outcome
Recover travellers from cancellations and delays quickly by matching disrupted itineraries to feasible re-accommodation options within fare rules, duty-of-care policy, and available inventory, with an agent or traveller accepting the option before it is issued.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Traveler or operations agent |
| Primary aggregate | DisruptionCase |
| Cosmos DB containers | 2 |
| Grounding data | flight operations, airport events, PSS/GDS reservations and inventory, minimum connection times, baggage state, fare rules, passenger-rights policy, loyalty priority, contact-center history, hotel rail and ground partners |
| Development estimate | USD 60-220 monthly dev estimate |
| Runtime options | Python FastAPI and C# ASP.NET Core |
Productivity hypothesis
These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.
| Impact type | Measure | Conservative internal assumption | Customer measurement |
|---|---|---|---|
| Productivity | Disruption-to-protected-itinerary time | 15-30% reduction | (baseline median disruption-to-protection minutes - assisted median minutes) / baseline median minutes * 100 |
| Quality or risk | Entitlement and option correctness | 5-10 percentage-point improvement | assisted percentage meeting the entitlement and feasible-option rubric - baseline percentage |
| Business outcome | Recovery cost per disrupted traveler | 5-10% reduction | (baseline recovery cost per disrupted traveler - assisted cost) / baseline cost * 100 |
Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.
Rationale: Flight, booking, entitlement, inventory, and partner evidence supports faster recovery ranking and a cautious cost hypothesis.
Pilot guardrail: Travelers or operations controllers approve options before issuance; compare equivalent disruption classes.
Customer validation plan
- Baseline the current workflow with representative users and tasks.
- Run the accelerator on the same task mix with human review retained.
- Measure disruption-to-protected-itinerary time and entitlement and option correctness with the documented formulas.
- Replace internal assumptions with customer-observed results before a business case.
Why Azure Cosmos DB
This AI solution needs an operational data plane, not only a vector index:
- Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
- Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
- Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
- Change feed reacts to source and operational updates without polling.
- Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.
Domain workflow

| Stage | Actor | Input → output | Approval |
|---|---|---|---|
| 1. Declare the disruption and scope impact | Operations control | Cancellation or major delay notification → DisruptionCase with affected itineraries | Policy controlled |
| 2. Prioritize travellers by risk and entitlement | Prioritization agent | Connections at risk, vulnerability, and loyalty tier → Ranked recovery queue | Policy controlled |
| 3. Generate feasible re-accommodation options | Recovery agent | Live inventory, fare rules, and duty-of-care policy → ReaccommodationOption set with compliance flags | Policy controlled |
| 4. Communicate options in the traveller's language | Disruption assistant | Ranked options and traveller contact preferences → Clear multilingual offer with entitlements stated | Policy controlled |
| 5. Accept and issue the recovery | Traveller or operations agent | Chosen option with cost and arrival impact → Reissued ticket, hotel voucher, and updated itinerary | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
disruptionCases | /tenantId, /disruptionEventId | DisruptionCase, AffectedItinerary, ReaccommodationOption, TravelerCommunication | Operational fields only |
entitlementPolicyLibrary | /tenantId, /jurisdiction | PolicyRule | /clauseVectorEmbedding, /clauseVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| Disruption recovery queue | disruptionCases | single-partition | Read affected itineraries and their options for the event in one partition query. |
| Missed-connection risk list | disruptionCases | single-partition | Prioritize travellers whose onward connections will be missed. |
| Policy-compliant option filter | disruptionCases | single-partition | Keep only options permitted by fare rules and duty-of-care policy. |
| Jurisdiction entitlement lookup | entitlementPolicyLibrary | single-partition | Knowledge-container query against entitlementPolicyLibrary (partition-aligned to /tenantId + /jurisdiction): return the compensation and care entitlements that apply to the affected journeys, read as a single partition query on the governing jurisdiction. |
Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.
Azure deployment

Clone the source repository, authenticate with Azure, and run:
Set-Location accelerators/travel-hospitality/disruption-recovery-assistant
az login
./Deploy.ps1 -Runtime python -Location eastus2
Change python to csharp to deploy the .NET implementation.
Explore the implementation
- Launch signed-in demo
- Implementation article
- Domain data model
- Workflow and API contract
- Python application
- C# application
- Domain container Bicep
- Platform Bicep
- Cost and configuration
- Source repository