Air India: AI.g customer service agent on Azure OpenAI
Air India · Travel and hospitality · India
The challenge
Millions of customer queries were overwhelming Air India's support channels, which raised costs, slowed responses and frustrated customers and staff. In 2024 the airline said it had doubled its passenger count since early 2022, so support had to scale without a matching rise in contact center work.12
What was implemented
Air India's own developers built AI.g, which launched in May 2023 on what was then Azure OpenAI Service. The 2024 Microsoft story also names Azure AI Search, Azure AI Speech, Azure AI Vision, Azure AI Content Safety, Azure Cosmos DB, Azure SQL Database, Azure Database for PostgreSQL and Azure Databricks among the services used. The 2026 story names Azure OpenAI in Foundry Models.
AI.g answers more than 1,300 types of question, from booking changes to refund requests, and passes the rest to human agents. The airline is trialling agents that carry out complete workflows across several systems, starting with refunds. Air India says the agent has saved it millions, without giving a figure or method.12
Reported outcomes
about 40,000 (2026), up from an average of 30,000 (2024)
Customer queries handled per day123
Measuredmore than 13 million
Conversations resolved since launch (Microsoft narrative; a quote in the same story refers to 4 million-plus queries, and the periods behind the two figures are not given)13
Measured97%
Success rate, meaning the share of queries handled without a human agent12
Measuredabout 9,000 queries a day, unchanged
Contact center volume while passenger numbers doubled since early 2022 (Air India statement, 2024)2
Measured
Each outcome links to its source; the label there says whether the company, Microsoft or a third party reported it.
What others can learn, and the limits
In our view, Air India is one of the more mature customer service agent cases: built in-house on a managed model service and tracked by volume and containment over several years. Readers should compare its figures with care, because all sources are Microsoft pages and the metric definitions change between stories (queries, conversations, success rate). Airlines and travel firms with fewer query types may reach similar containment with a narrower bot and well-maintained knowledge content before adding multi-system agents.
Sources
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.