Reliance is embedding an AI assistant into Jio, India's largest mobile carrier with 500 million subscribers, so it can join your phone calls, transcribe them, and book a cab without an app.
Reliance is embedding an AI assistant into its telecom network so it can join a live phone call, book a cab, and order food, without the user installing anything. On Friday, chairman Mukesh Ambani unveiled Jio Call Agent, a voice helper bound to Jio, India's largest mobile carrier and a Reliance subsidiary with more than 500 million subscribers.
The mechanism matters more than any single feature. Jio Call Agent sits inside the carrier's network rather than running on a phone or being delivered through a third-party app like a US-style call screener. That gives Jio something the App Store AI race cannot easily replicate: a 500-million-plus subscriber base that already routes its calls, messages, and billing through a single operator, and a built-in voice channel that does not require users to install anything new.
That is the structural bet Reliance is making. In a global AI market dominated by US and Chinese model labs, the company is not racing to build a frontier large language model. It is racing to own the distribution layer that AI will run through for ordinary Indians, the same way Jio once used cheap data to upend India's telecom market a decade ago. Ambani has framed the move as industrial policy. "India should not be a mere consumer of AI created elsewhere," he told shareholders. "It must become a creator, adopter, and a global leader in AI."
The ambition is large. Reliance has said it will invest $110 billion in AI infrastructure this year — a figure whose commitment level is unclear from available disclosures, and which may aggregate spending across multiple Reliance entities rather than representing a single-year Jio capex commitment. The company also launched Reliance Intelligence last year, an AI unit aimed at consumers, businesses, and governments, with a stated commitment to support 22 Indian languages. Reliance has lined up partnerships with Google, Meta, and Nvidia for model access and compute. Alongside Jio Call Agent, Friday's announcements included an AI-powered version of the MyJio app that lets users activate an eSIM or pick a roaming plan by typing (or saying) what they want, and a home display called TeleFrame that surfaces weather alerts, schedules, and reminders through AI agents, putting Reliance in direct competition with Amazon's Echo Show and Google's Nest Hub lineup.
The skeptic's questions are already queued up. Jio Call Agent is "expected to launch later this year," per Reliance's own framing, which leaves a wide gap between announcement and shipped product. An AI agent that listens to live phone calls, transcribes them, and acts on them on the user's behalf is also a serious privacy surface, one the user cannot opt out of by simply uninstalling an app, because it sits in the carrier's network. That structural position is the same feature that makes the product work.
There is also the question of whether $110 billion is firm capital expenditure or aspirational commitment, and whether Reliance's "national champion" framing is a description of an organic market position or a quiet pitch for state-aligned industrial policy. India's AI ambitions do not currently include a domestic frontier-model competitor to OpenAI, Anthropic, or the Chinese labs, and Reliance is the closest thing the country has to a candidate. Whether that means Reliance deserves the slot, or has simply captured it by being too big to bypass, is a different question.
The 22-language commitment is worth taking seriously rather than waving through. Most consumer AI products in India today work in English plus a handful of high-resource languages. A voice interface that genuinely understands the full spread of Indian languages, runs on a network that already carries those calls, and does not require a smartphone app is a real differentiator, if Reliance can actually ship it at Jio's scale.
What to watch next: whether Jio Call Agent launches in 2026 with the feature set Ambani described, whether Indian regulators accept a carrier-embedded AI listening on calls without a separate opt-in, and whether the $110 billion capex shows up in Reliance's earnings disclosures as committed spend rather than a multi-year target. The interesting question is not whether Reliance can build a model as good as GPT or Gemini. It is whether owning the call is a better moat.