Air India has reportedly become the world's first airline to introduce a generative AI virtual agent. The agent named Maharaja has successfully deployed, powered by Microsoft's Azure OpenAI service. The project was presented in its testing phase in March 2023. Air India claims that Maharaja has responded to over a million customer queries by the time and currently handles more than 6000 daily queries in four different languages: Hindi, English, German and French.  

How does the virtual agent Maharaja work? 

The virtual agent Maharaja handles a wide variety of customer queries across 1300 domains, such as flight status, baggage allowances, check-in, packing restrictions, airport lounge access, frequent flyer awards, refunds and flight changes. Among the 6000 questions Maharaja received daily, it successfully answered about 80% in seconds. For the remaining 15% of queries, Maharaja required additional assistance, and therefore, it transferred the queries to the contact centre agents of Air India.  

Satya Ramaswamy, Chief Digital and Technology Officer at Air India stated, "We are dedicated to providing the best possible service and experience to our valued guests across channels. We also want to be very responsive to their changing preferences. After the emergence of Large Language Model-driven Generative AI capabilities, we are increasingly seeing a definitive shift in guest preferences to use chat interactions to get the information and support they need quickly and directly as compared to browsing several web pages." 

Minimising human intervention 

The new AI-powered chatbot introduced by Air India will minimise the need for human intervention in simple tasks and basic queries. It also assists in automating repetitive tasks like booking, cancellations and confirmations. As per the company, this service will free up bandwidth for human agents to concentrate more on complex and value-added tasks. Furthermore, Air India has deployed a strategy that combines various traditional machine learning techniques with generative AI. Moreover, the company also claims to have implemented safeguards to stop biased or foul language from infiltrating its conversational AI system as well.