Air India launches world’s first AI agent for rapid customer service

Air India has achieved a pioneering milestone by becoming the world’s first airline to implement a Generative AI virtual agent named ‘Maharaja.’ This virtual agent is powered by Microsoft’s Azure OpenAI service, and the project was initially introduced in the testing phase in March 2023. Air India reports that Maharaja has successfully addressed over half a million customer queries, currently managing more than 6,000 queries per day in four languages: Hindi, English, French, and German.

Maharaja’s primary function is to handle a diverse array of customer queries spanning 1,300 areas, including flight status, baggage allowances, packing restrictions, check-in procedures, frequent flyer awards, airport lounge access, flight changes, and refunds.

Out of the daily influx of 6,000-plus questions, Maharaja adeptly responds to over 80 per cent of them within seconds. For the remaining 15 per cent of queries that require additional assistance, Maharaja autonomously identifies this need and transfers the query to Air India’s contact center agents.

Satya Ramaswamy, Chief Digital and Technology Officer at Air India, noted the airline’s commitment to delivering exceptional service and adapting to changing guest preferences. With the emergence of Large Language Model-driven Generative AI capabilities, there has been a noticeable shift in guest preferences towards utilising chat interactions for quick and direct information and support.

The implementation of this AI-based Chat bot is expected to significantly reduce the need for human intervention in handling basic queries, automating repetitive tasks such as booking, cancellation, and confirmation. Air India aims to enhance human agents’ efficiency by allowing them to focus on more complex and value-adding interactions.

To ensure ethical and responsible usage, Air India has employed a strategy that combines various traditional machine learning techniques with Generative AI. The airline asserts that safeguards are in place to prevent biased or harmful language from infiltrating its conversational AI system.