A panel we hosted in Bengaluru last February – featuring operators from fashion, hardware, education, and entertainment – made one thing unmistakably clear: India isn’t catching up. It’s running a different race entirely.
Here is what is actually changing, sector by sector.
Conversational shopping – understanding occasion, context, and preference in natural language – is the most promising direction. But the real test is calibration: an AI-powered discovery experience has to feel native, not bolted on. And, voice has not cracked commerce, yet. Visual browsing wins on cognitive load; scrolling and tapping is genuinely lower effort than speaking and waiting. Even among vernacular users on platforms built specifically for them, the thumb beat the voice (Indian customers struggle with articulation of tasks). Image search is growing. Voice remains largely a customer support input for now.
Platform moats in fashion are shifting as a result. Catalogue depth is increasingly replicable. The emerging moat is the feedback loop between user interaction and model improvement – the deeper a platform’s understanding of how a specific user’s taste evolves over time, the harder it becomes to displace.
Hardware: The AI Is moving off the Cloud and onto the Device
Consumer hardware is undergoing a more fundamental transition than most consumer tech categories, because the site of AI computation is moving. Real-time ambient noise cancellation on a call, sensor fusion on a smartwatch, recovery analysis on a ring – these decisions need to happen in milliseconds, on-device. This is not merely a technical shift; it is a product design philosophy with direct consequences for the consumer experience.
Additionally, what is distinctive about building AI hardware for India is the localisation challenge. Indian consumers are genuinely heterogeneous in what they expect from an agent – not just across languages, but across regions and cultures. Some users want one crisp answer. Others expect a thorough walkthrough. General-purpose voice models are entirely blind to this variation. The operators getting it right are building regional interaction data loops – feeding conversations back into the system to fine-tune responses. That corpus, accumulated conversation by conversation, is a moat that cannot be purchased.
A deeper insight for builders is that, carefully fine-tuned models have an emergent ability to pick up on regional lingo. Drop a few Bhojpuri words or write in a Marwari register – the model catches it. New age platforms are now feeding these cues back into the system, allowing it to respond to each user in the register “they speak”.
This draws a hard line: if your platform behaves identically for a Tier 1 English-speaking user and a first-generation smartphone user in Patna – you haven’t built for Bharat. You’ve built something that tolerates it.
Entertainment: AI is rebuilding the Content Supply Chain
Indian digital entertainment is in the middle of a supply-side disruption. The traditional OTT model – commission, film, post-produce, release – runs on timelines measured in months.
AI is compressing this from the inside out. Mobile-first content players are now going from concept to finished episode in under a month, with real-time viewer drop-off data feeding back into every creative decision – scene by scene, hook by hook.
The consumer behaviour insight driving this is stark. Indian mobile entertainment consumers, particularly on paywall platforms, are less forgiving of slow pacing content than any previous generation. The tolerance for a slow opening has shrunk to approximately the time it takes to swipe to a Reels feed. This elevates hook strategy to a first-order content problem – a strong story with a weak opening will lose to a moderate story with a compelling one.
Left to right: Anoop N Menon (Principal @Chiratae Ventures), Lakshminarayan Swaminathan (Head of PM @Myntra), Shyam Vedantam (CPO @boAt), Gautam Pratap Singh (SVP – Business @Seekho), Pratik Anand (Co-Founder @Flick TV)
In fashion, it is the discovery moment for a user with an occasion and no time to browse.
In hardware, it is the on-device inference that makes a wearable feel genuinely useful rather than a data dashboard.
In education, it is the personalised learning loop that meets a non-English speaker where they actually are.
In entertainment, it is the content iteration speed that lets a lean team compete with a large one on the quality of the hook.
“That specificity – the precise understanding of how a particular user thinks, communicates, and decides in a particular context – is the most defensible thing you can build in Consumer AI right now.”
Bengaluru gave us three things we’re still thinking about:
“Voice is not the interface unlock we assumed yet, the data corpus is the real moat, and Bharat will punish lazy localisation in ways that aren’t visible until it’s too late.”
This was Edition 1. We took the conversation to Mumbai next – different operators, different sectors, same honesty. And we’re taking it to three more cities after that, across health, gaming, travel, companionship and more.