Air India
genAI customer agent (automated customer service)
Air India's generative customer agent, AI.g, handles about 40,000 queries a day and has resolved more than 13 million conversations with a 97 percent automation rate.
Key points
- GenAI customer agent (AI.g / Maharaja) on the site and app.
- Built on Azure OpenAI, with escalation to a human agent.
- About 40,000 queries per day, 97% automated, 13 million conversations resolved.
- Evidence level B, confirmed active status.
Objective
Absorb the rise in contact volume after passenger traffic doubled without growing the call center, by automating recurring questions and reserving human agents for complex cases.
The deployment
AI.g (also called Maharaja) is Air India's generative virtual agent, built on Azure OpenAI Service. It answers passengers on the site and app about flight status, baggage, check-in, changes, refunds, and the loyalty program, covering more than 1,300 question types. Piloted in March 2023 and then expanded, it handles questions in natural language and hands off to a human agent for out-of-scope cases. Microsoft documents a continuous scale-up: about 40,000 queries per day and more than 13 million conversations resolved, with half of customers now choosing the AI as their first point of contact.
Results Proof B
Quantified Microsoft (vendor) case study, updated between 2024 and 2026 with consistent figures, complemented by Air India's official deployment press release.
How it works
Documented architectureThe stack in detail
- plateforme Microsoft Azure OpenAI Service Azure service exposing OpenAI's models, the foundation of the AI.g agent.
- llm Modèles GPT (OpenAI) Generative language models served through Azure OpenAI; the exact version is not published.
- infra Azure AI Azure data and application components around the agent (hosting, security, scaling).
- infra Connecteurs vers les systèmes de réservation et d'opérations In-house integration to booking, flight status, baggage, and the loyalty program, required for transactional answers.
- integrateur Microsoft Technical partner for the deployment and the scale-up.
How it runs, concretely
For ops teams-
1Receiving the question customer
The passenger asks the question in natural language on the site or app.
-
2Classification and retrieval AI
The agent identifies the topic among more than 1,300 categories and queries the internal systems for context (booking, flight status).
-
3Answer or action AI
It answers, guides a change or a refund, or handles the request automatically when it is within scope.
-
4Escalation AI / human
About 3 percent of cases are handed off to a human agent for complex situations.
-
5Supervision and improvement data team
The teams review failures and expand the covered scope wave after wave.
The passenger's question, matched against the booking and operations systems (flight status, baggage, loyalty). If the agent loses live access to these systems, it can no longer answer transactional cases and human escalation rises.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- structured customer service knowledge base
- API access to the booking and operations systems
- ticket history to frame the scope
Org prerequisites
- product and data team
- defined human escalation process
- answer governance and compliance
Possible stack
- Azure OpenAI or equivalent
- RAG layer over the knowledge base
- connectors to the PSS / CRM
The plan, step by step
- Step 1Map the 20 to 30 most frequent contact center intents and structure the associated knowledge base.Deliverable: Locked v1 scope plus a structured knowledge corpus
- Step 2Build the agent (LLM plus retrieval over the base) and define the human escalation rules.Deliverable: Agent in pre-production on informational questions
- Step 3Connect the transactional systems in read mode: booking, flight status, baggage, loyalty.Deliverable: Context-aware answers tested on real cases
- Step 4Run a pilot on one channel (website) and track resolution rate and failures.Deliverable: Pilot report with a production go-live threshold
- Step 5Move to production, expand the scope in waves, and review failures each week.Deliverable: Automation / escalation dashboard monitored routinely
First step: Map the 20 to 30 most frequent contact center intents and connect the agent to that base before expanding.
The same pattern, proven in other industries
ADT
genAI customer agent
AT&T
Governed enterprise genAI assistant: an LLM connected to internal knowledge, open to tens of thousands of employees and called at scale by automated systems
Bank of America
conversational financial assistant at scale
Sources
- S1 How Azure AI helped Air India reinvent customer service by answering 40,000 daily queries instantly Interested party archive pending
- S2 Air India elevates customer support while saving money with Azure AI, data, and apps Interested party archive pending
- S3 Air India successfully deploys airline industry's first generative AI virtual agent, powered by Microsoft Azure OpenAI service Primary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.