Myra, the conversational AI travel agent of MakeMyTrip, has scaled from 25,000 conversations daily in Q2FY26 to over 50,000 during Q4FY26, reaching 80,000 conversations per day at the time of the call, co-founder and Group Chief Executive Officer (CEO) Rajesh Magow told analysts during the company’s Q4FY26 earnings call on May 19. “AI allows us to bridge language, trust, and discovery barriers in ways that were previously impossible,” he said.
Myra: from planning tool to transaction agent. The upgraded version covers flights, hotels, buses, trains, cabs, and full itinerary planning through a multilingual voice feature, and directly assisted over 200,000 bookings in Q4FY26. MediaNama reported in Q2FY26 that Myra had scaled to 25,000 conversations daily, with over 35% of travellers engaging with the agent up to 90 days before their trip.
Key developments this quarter:
On the proprietary data layer, Magow said: “We have been on our journey to embed GenAI all through the consumer journey, leveraging our own proprietary data.” On measurement, he said: “We have clear metrics defined on measurement, specifically on quality of conversation, something called a good conversation versus not so good conversation. There’s a clear quality metric attached to it.”
Regional language users in tier 2 and smaller cities are now completing payments through Myra’s voice interface, with Myra collecting behavioural, voice, and transactional data in the process. The Digital Personal Data Protection (DPDP) Act, notified in November 2025, requires informed consent from the data principal, the individual whose personal data is being processed, before that data is collected. The call did not address how consent is obtained or governed for these users.
Magow’s case against agentic disruption: four moats. Asked directly whether AI agents could displace online travel agencies (OTAs), Magow said:
Magow’s four-moat framework is more specific than Eternal co-founder Deepinder Goyal’s argument in his Q4FY26 shareholders’ letter, which relied primarily on consumer habits and physical infrastructure. The question that neither has publicly answered is what happens when an AI agent acting on a user’s behalf gets a travel booking wrong. Travel adds international jurisdiction and airline liability to a regulatory gap that Indian law has not addressed.
On horizontal AI players, meaning general-purpose AI agents like ChatGPT or Perplexity that handle travel queries alongside everything else rather than specialising in travel, Magow said: “They have already stated that they want to probably focus a lot more on the planning and the discovery step of the overall journey and not necessarily go deep because it’s not easy and probably not their DNA to go really deep in the funnel.”
On MakeMyTrip’s own positioning, Magow said: “We on MakeMyTrip will leave no stone unturned to ensure that leveraging this technology, whatever it takes, we continue to be the first place of choice for all the new users for travel when they come online.” He also said the company would “partner in that scenario if we need to be.”
Smart Search and voice reviews. Smart Search, a semantic free-text search capability that MakeMyTrip deployed this quarter, lets users describe what they want in natural language. “For example, family stay near Baga Beach with Jain food or a rooftop pool hotel in Jaipur with spa access,” Magow said. “Through this feature, customers now receive contextually precise, explainable results. This feature delivers much higher conversion versus traditional filter-based journeys, clearly demonstrating that understanding intent outperforms matching keywords.”
A user searching for a Jain-friendly stay near Baga Beach is revealing dietary preferences, travel companions, and destination intent — intent data, specific signals about what a user wants that natural language captures more precisely than keyword search. This feeds into Myra and the knowledge graph of MakeMyTrip, a structured database of travel information that powers its AI recommendations. No public disclosure exists on how long this data is retained or what it is used for beyond the immediate search.
On voice reviews, Magow said: “Voice reviews are generating a lot more content per submission compared to typed reviews. Customers describe their stays naturally in detail in their own language. This richer signal feeds directly into our knowledge graph, improving the quality of AI-generated summaries, safety scores, and contextual recommendations for future travelers. Voice is becoming the default input for Indian customers, increasingly, and we are building our content infrastructure around that reality.”
A negative voice review that an AI misinterprets could affect a property’s safety score and ranking, with no human review or appeals process disclosed. The governance framework for how review data is used, stored, or weighted in algorithmic outputs has not been made public.
MakeMyTrip Organisational AI Key Developments:
The call does not address what happens to the 45% of customer service queries the AI cannot resolve or whether AI-resolved queries fall within the Consumer Protection Act, 2019’s grievance redressal framework, which requires platforms to acknowledge and resolve consumer complaints within defined timelines.
On cost timing, Magow said: “There is going to be AI tooling cost, and then there is going to be efficiency kicking in. At some point in time, efficiency is going to show bigger impact than the additional cost that is coming from the AI tools.”
RedBus AI: RAY and voice bots. RAY is now driving 2x engagement among regional language users in the pre-booking journey on RedBus. “Around 6% of total queries come via voice as input. RAY is emerging as an assist layer that improves decision confidence before booking and deepens engagement in core booking funnels,” Magow said. RedBus has recorded 33% efficiency gains in customer support by replacing legacy interactive voice response (IVR) systems with voice bots, Magow said. Regional language users engaging with RAY via voice are generating the same data collection and consent questions as Myra, with no separate disclosure on the call.
Also read:
For You
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。