Travel and hospitality
How airlines, cruise lines and travel firms use Microsoft technology: in-house customer service agents on Azure OpenAI, modernized booking and operations systems, and the payment card and biometric rules that apply.
The problem
Passenger growth can overwhelm support channels. Air India says it doubled its passenger count after early 2022, and millions of customer queries were raising costs, slowing responses and frustrating customers and staff.
Requests are varied and often touch several systems: Air India's assistant covers more than 1,300 question types, from booking changes to refund requests.
Travel companies take payment cards for bookings, and biometric identification, increasingly discussed for airports, is a high-risk area under the EU AI Act.1234
The approach
Microsoft places travel under Microsoft for Mobility. Its travel and transportation page centers on connected data, frontline staff, intelligent operations and customer loyalty, and links a reference architecture for airlines and airports. Customer examples on that page include American Airlines (Azure SQL Managed Instance, Azure App Service, Azure Data Factory) and Carnival Corporation (Azure Kubernetes Service, Azure DevOps). The page does not specifically address hotels.
Air India built its AI.g assistant with its own developers on Azure OpenAI, now Azure OpenAI in Foundry Models, with services such as Azure AI Search, Azure Cosmos DB and Azure SQL Database listed in its 2024 story. Microsoft's 2026 story says AI.g handles about 40,000 queries a day and has resolved more than 13 million conversations with a 97 percent success rate, that customers can still choose a human agent, and that Air India is testing agents that complete workflows across systems, such as refunds.
A small hotel group or tour operator can usually rely on the chat features of its booking engine or CRM. Building in-house, as Air India did, makes sense with very high query volumes, an internal development team and APIs into reservation and refund systems.125
Conceptual architecture
Diagram as a list
Copilot & business apps
- Web, app and messaging channelsWhere travelers ask questionsConnects to Customer service agent (Azure OpenAI in Foundry Models)
- Handover to human agentsComplex cases and customer choice
Agents & automation
- Customer service agent (Azure OpenAI in Foundry Models)Answers and completes routine requests
AI models & platform
- Policy and fare knowledge retrieval (Azure AI Search)Grounds answers in current rules
Data & analytics
- Reservation, refund and loyalty systems through APIsWhere changes are actually made
- Conversation and customer data (Azure Cosmos DB, Azure SQL Database)Session state and historyConnects to Customer service agent (Azure OpenAI in Foundry Models)
Cloud & hybrid infrastructure
- Modernized booking and operations apps (Azure Kubernetes Service, Azure SQL Managed Instance)Hosts the systems the agent callsConnects to Reservation, refund and loyalty systems through APIs
Technologies and their roles
Microsoft Foundry Models2
Language models for the assistant
Air India's AI.g runs on Azure OpenAI in Foundry Models.
Azure AI Search1
Knowledge retrieval
Listed among the services in Air India's 2024 customer story.
Azure Cosmos DB1
Conversation data
Listed among the services in Air India's 2024 customer story.
Azure Kubernetes Service (AKS)5
Application platform
Carnival Corporation is shown on Microsoft's travel page using Azure Kubernetes Service.
Azure SQL5
Booking and operations data
American Airlines is shown on Microsoft's travel page using Azure SQL Managed Instance.
What you need first2
- APIs into reservation, refund and loyalty systems with clear permissions
- Handover rules that let customers reach a person at any time
- A maintained source of fare rules and policies for retrieval
- Separation of payment card handling from the assistant
Risks and how to reduce them
- PCI DSS scope3
- Microsoft's PCI DSS attestation does not cover services customers build; keep card data out of the assistant or include it in your own assessment.
- Biometrics under the EU AI Act4
- Remote biometric identification is listed as high-risk; assess any facial recognition at airports or check-in against Annex III obligations.
- Changing metric definitions1
- Air India's figures move between queries, conversations and success rate across stories; define containment and success precisely before comparing.
Documented examples
Air India
In productionAir India: AI.g customer service agent on Azure OpenAI
about 40,000 (2026), up from an average of 30,000 (2024) (measured)
Travel and hospitality, India
Related
Sources
- Customer story: Air India (Azure OpenAI Service and Azure data services) (opens in a new tab)
- Customer story: Air India (Azure OpenAI in Foundry Models) (opens in a new tab)
- Payment Card Industry Data Security Standard (PCI DSS) - Microsoft Compliance (opens in a new tab)
- AI Act: regulatory framework for AI (opens in a new tab)
- Travel and Transportation | Microsoft AI (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.