The National e-Governance Division has selected CoRover.ai to deliver a unified artificial-intelligence assistant for central digital services, beginning with the DigiLocker document wallet, after a tender that drew bids from more than 80 companies including Deloitte, EY, PwC, KPMG, Tata Consultancy Services and Jio Haptik.
Tender outcome
CoRover, a Bengaluru agentic AI startup, scored highest on technical evaluation and emerged as the sole L1 financial bidder, according to people familiar with the process cited in industry newsletters this month. Founder Ankush Sabharwal said implementation on DigiLocker could go live within four months, giving citizens a single conversational entry point for fetching marksheets, licences and other verified records.
MeitY secretary S Krishnan has linked the project to India’s wider AI public infrastructure roadmap, which includes datasets and compute access through the AIKOSH initiative. The award lands while electronics and IT minister Ashwini Vaishnaw prepares a regulation-focused consultation paper on AI safety, deepfakes and skilling expected within weeks.
Why DigiLocker goes first
DigiLocker already hosts hundreds of millions of issued documents and acts as the identity backbone for several state portals. Adding an assistant layer could reduce call-centre load if citizens can retrieve certificates through voice or chat in Indian languages, but it also raises data-minimisation questions because the bot will need authenticated sessions tied to Aadhaar-linked accounts.
CoRover has previously deployed multilingual virtual assistants for railways and banking clients; the government contract will test whether a mid-sized Indian vendor can scale to nationwide traffic without leaning on hyperscaler models alone. Competitors who lost the bid may still supply components through subcontracting if the master systems integrator model allows it.
Policy context
Vaishnaw has argued that industry should carry primary responsibility for AI risk, with government setting guardrails through a techno-legal framework rather than prescriptive code-level rules. The forthcoming consultation paper is expected to address psychological harms and synthetic media alongside cybersecurity expectations for deployers of agentic systems.
That tone matters for DigiLocker: any assistant that can read a user’s vault must log prompts, restrict tool access and survive red-team reviews similar to those CERT-In now recommends for enterprises experimenting with autonomous agents.
What success would look like
Officials will judge the pilot on completion rates for common requests—downloading a Class X certificate, sharing a driving licence QR code—without escalating to human help desks. If metrics hold, NeGD could extend the same orchestration layer to UMANG and state service portals, turning the award into a reference architecture for sovereign AI interfaces rather than a one-off chatbot.