MakeMyTrip’s AI playbook balances emotion, efficiency and security

MakeMyTrip and travel booking have been practically synonymous for years. Now there’s a prominent addition to the conversation, and it’s AI.

Whilst the AI buzz has reached fever pitch recently, here’s the thing most people don’t realise: MakeMyTrip has been quietly working with this technology for far longer than the headlines might suggest.

“AI has been a long journey for MakeMyTrip, stretching back seven years or more,” Group Chief Technology Officer, Sanjay Mohan recalled, settling into his chair as he explained the company’s AI direction at TechSparks 2025.

The company’s AI journey began with the mundane, and often invisible, work of collecting, cleaning and structuring data at a scale that could support meaningful intelligence. The focus was patient and methodical, and it laid the groundwork for everything that followed.

Mohan described the early years as a period of preparation rather than application. He recalls the simple but fundamental truth that high-quality data is the first requirement for any credible use of AI.

“AI cannot be done without data, and therefore, the lead time for getting any AI feature out is rather long. You have to collect the data, control the quality and monitor it. It took us around two years to build a platform that ingests data and pushes out good-quality information from the other end,” he remarked.

Visible innovation

Customer-facing AI use cases and product innovation grew naturally from this long-term investment. MMT’s teams began experimenting with ways to make the experience more relevant, intuitive and immediate for the traveller.

The shift was noticeable. Features such as personalised hotel rankings, zero cancellation options and fare locks became embedded within the platform and began to influence behaviour at scale.

At the same time, the company moved early to integrate generative AI directly into its product rather than relying on external plugins. This allowed customers to search, transact and resolve queries inside the familiar environment of the app. 

“We were the first ecommerce company in India to launch a fully integrated generative AI capability in the product. Other companies were doing plugins, but that was not in their product. You could get information, but you could not transact or handle post-sales queries,” Mohan highlighted the importance of this approach.

The arrival of the company’s conversational assistant, Myra, added a new layer to this strategy. It showed that AI could simplify planning while reducing the effort a user must invest in form-filling or repeated browsing.

Voice interactions in particular revealed how comfortable many Indian travellers are when freed from the friction of typing. Myra also began assisting with real-time questions such as visa requirements and layover details, and early signs suggested higher conversion rates among users who engaged with these tools.

Reading emotion

Meanwhile, understanding customer emotion, intent, and conversational UX has emerged as one of the most complex yet essential parts of MMT’s work.

Voice engagement opened a window into new behavioural signals that text could never fully capture. Emotion, tone and urgency became part of the conversational data that the company needed to interpret.

The challenge, however, is far from straightforward. India’s linguistic diversity makes sentiment analysis particularly demanding, and expressions vary widely across regions.

“It is very important for us to capture the frustration level because post-sales conversations are sometimes unhappy. If you ask me something and I give you an irrelevant answer, I am frustrating you again. The nuances of regional languages are hard to capture, and voice technology for Indian languages is still not very good,” Mohan explained how the team is attempting to navigate this reality.

Despite these challenges, the company sees significant potential in emotions as a signal. Expressions that convey delight or disappointment can influence recommendations and help the system recognise when to escalate to a human. That blend of empathy and efficiency will determine how conversational interfaces evolve in Indian travel.

Smarter operations

Amidst these AI-related shifts, operational efficiency, productivity and the human AI balance have become central to the company’s internal transformation.

MMT has approached automation not as a replacement for people but as a way to free them to focus on tasks where judgment matters. The introduction of AI-assisted call transcription is a clear example. Instead of agents scrambling to take notes while listening, the system now transcribes entire calls, identifies intent and prepares summaries. This has shortened call duration, lowered error rates and improved customer satisfaction.

“Now the person is paying more attention to the consumer because somebody else is taking notes, intent identification also happens through the bot, and the human can verify it. The mistakes made by the call centre agent have gone down because there are two pairs of ears listening,” Mohan explains.

The same pattern is visible within engineering teams, where code generation and automated testing are changing the pace at which products can be built. Productivity gains are distributed rather than concentrated, and the company’s ambition is to raise overall output without dramatic changes.

Secure by design

As the use of AI accelerates, privacy, security, infrastructure and the future of the industry sit at the centre of the strategic debate.

For MMT, the principle is clear. Generative AI features must operate within the company’s protected environment, and customer data must never leave its controlled systems. This approach becomes even more important as AI systems become multimodal and begin to process voice, images and documents.

“Everything is within our controlled firewall, and all generative AI runs inside the app. Our partners practise responsible AI, and the frontier models operate within our own protected firewalls. Data does not go out of my system at all,” Mohan outlined.

Looking ahead, Mohan expects rapid change in user interfaces, devices and modes of interaction. He believes voice will become mainstream for travel queries within the next two years and that multimodal inputs such as photos, videos, or social media clips will reshape how people express travel intent. Mohan also suggested that the current dominance of text-heavy AI interfaces is temporary and that more natural and layered forms of communication will emerge.

Masterclass

During a masterclass that followed, held in a hall packed with tech enthusiasts, Mohan spoke about how generative AI is changing the way people use travel platforms. Travellers are moving away from short search queries toward longer, more expressive conversations, a shift seen most clearly among first-time users in smaller towns who prefer speaking to typing.

Interaction is also becoming multimodal, with people sharing images or short videos to describe the holidays they want and expecting richer outputs such as video itineraries. Yet many conversational tools have slipped back into basic text exchanges, a design challenge he believes the industry must confront.

These behavioural changes are pushing companies to rethink how their systems work. The neat, step-by-step flow of traditional ecommerce cannot handle queries that cut across destinations, budgets, policies and real-time needs, so the technology must adapt to follow the user’s lead.

He stressed that generic AI models are not enough for a sector built on specialised knowledge. Platforms must add their own travel intelligence so that suggestions are as accurate as they are creative.

The investment has been heavy, he said, but the results are clear, with generative AI features improving conversions by notable margins through sharper, more personalised summaries delivered securely within the company’s protected environment.