CoRover.ai Selected by MeitY to Develop Government AI Platform

India’s Ministry of Electronics and Information Technology (MeitY) has selected Bengaluru-based CoRover.ai to develop a reusable artificial intelligence platform for government services, with DigiLocker expected to be its first implementation. The initiative aims to move beyond basic chatbots towards agentic AI systems that can answer citizens’ questions, help them complete tasks and escalate problems that require human intervention.

The project is part of India’s broader push to integrate AI into digital public services. Reports published in October 2026 say the National e-Governance Division (NeGD), which operates under MeitY, awarded the contract to CoRover after a competitive selection process. The proposed platform could eventually support multiple government departments, although its expansion will depend on implementation, security testing and the performance of the initial DigiLocker deployment.

Key takeaways

• Government AI contract: CoRover.ai has been selected to develop a unified AI assistant and bot platform for government services.
• DigiLocker first: The digital document service is expected to be the first implementation of the framework.
• Agentic AI approach: The proposed system is designed around an “ask, do and escalate” model—answering queries, helping carry out tasks and handing unresolved cases to human support.
• Reusable infrastructure: The objective is to build a framework that other government services can adopt rather than creating a separate chatbot for every department.
• Deployment timeline: NewsBytes reported an initial DigiLocker rollout in approximately four months, followed by continued improvements over the subsequent year. This is a reported timeline, not confirmation that the service is already live.
• Key challenges: Data protection, accuracy, language support, transaction reliability and clear human escalation will be crucial to public trust.

What is MeitY’s government AI platform?

The project aims to establish a common AI assistant framework that government departments can use to interact with citizens digitally.

Many government websites already offer searchable information, online forms, help pages and chatbots. These tools can help users find instructions, but they may struggle when a request involves several steps or when a person cannot easily identify the correct department or procedure.

An agentic AI platform is intended to go further. Instead of merely generating an answer, it may be able to connect a request to an approved workflow, retrieve relevant information, help complete a transaction and determine when human assistance is needed.

Moneycontrol reported in September 2026 that the proposed framework would follow an “ask, do and escalate” approach. The design would allow the AI assistant to answer a question, help the user take an action and escalate a problem when it cannot resolve the issue appropriately.

The framework is intended to be reusable across government services. This could reduce the need for each department to independently procure, develop and maintain an entirely separate conversational AI system.

However, a common platform does not automatically mean that every department’s data will be combined or that one AI agent will have unrestricted access to government systems. Actual integrations, permissions and access to citizen records would need to be defined for each service.

Why DigiLocker will be the first use case

DigiLocker is an Indian digital public service that allows users to access and manage digital documents and records. It is a logical starting point for an AI assistant because users frequently need help understanding document requirements, finding records and navigating service-related procedures.

The planned AI upgrade could make it easier for users to interact with the service through conversational questions rather than relying entirely on menus, instructions and conventional search.

For example, a user might ask how to find a document, which records are needed for a particular process or how to resolve a problem accessing a service. An assistant could provide guidance and, where the relevant integrations permit it, help the user move through an approved workflow.

These are illustrative possibilities, not a confirmed list of launch features. The precise transactions the assistant will be authorised to perform, and the level of access it will have to personal records, remain important implementation details.

From answering questions to completing tasks

A conventional chatbot might explain how to retrieve a document from DigiLocker. An agentic assistant could potentially take the next step by directing the user to the relevant function, checking the status of an authorised request or initiating an approved process.

This distinction is central to the project. A system that only answers questions can still leave users to navigate multiple pages and repeat information. An assistant capable of performing approved actions could reduce that friction.

But performing tasks introduces greater responsibility. A wrong answer can mislead a user; an incorrect action could affect a document request, transaction or service application. For that reason, the platform will need clear boundaries around what it can do independently and what requires explicit user confirmation.

Human escalation remains important

The “escalate” part of the proposed framework is just as important as the ability to answer and act.

AI systems may encounter incomplete records, unusual requests, conflicting information or problems outside their authorised scope. A reliable public-service assistant should recognise these situations instead of improvising an answer or repeatedly sending a user through the same automated steps.

A well-designed escalation process would explain why the issue needs human attention, preserve the relevant context and direct the user to an appropriate support channel. That could make the service more useful for people whose problems do not fit standard automated workflows.

The effectiveness of this process will depend on how the system identifies uncertainty, what information it passes to human support and how users can track unresolved cases.

Who is CoRover.ai?

CoRover.ai is a Bengaluru-based conversational AI company that develops voice, video and text-based assistants for organisations. Its platform is marketed under the BharatGPT name and is designed to support conversational interfaces across multiple languages and digital channels.

The company says its technology can be deployed through websites, applications, messaging platforms and voice-based services. It also promotes AI agents that can automate workflows and transfer conversations to human staff when necessary.

CoRover’s existing work provides relevant experience for a government-facing project. The Government of India’s India Science, Technology & Innovation portal, for example, describes AskDISHA 2.0 as a conversational AI platform used for railway-related queries and transactions, while DigiSaathi provides information about digital payment products and services.

These examples illustrate the kinds of conversational services that can help users navigate complex systems. They do not, however, establish that the new government platform will use exactly the same technology or that all the capabilities of existing deployments will be available in DigiLocker.

The new contract would give CoRover an opportunity to extend its technology into a broader government-services framework. Its performance will be assessed not only on conversational quality but also on integration with public systems, reliability, security and the ability to handle sensitive citizen interactions responsibly.

How the selection process unfolded

The selection process involved a technical evaluation and a competitive bidding procedure. Moneycontrol reported on September 23 that MeitY had narrowed the field to CoRover and Kyndryl Solutions after the technical evaluation.

Subsequent reports in October said CoRover emerged as the highest-ranked bidder under a quality-and-cost-based selection process and was awarded the project through NeGD.

Quality-and-cost-based selection is designed to evaluate bids using both technical considerations and financial criteria. It differs from choosing a supplier solely on the basis of the lowest price.

The sequence matters because being shortlisted is not the same as winning a contract. The September reporting described the narrowed field; the later October reporting described CoRover’s selection and award.

Public reporting has not established a complete set of contract details, including a confirmed total contract value, all delivery milestones or the final list of participating government departments. Those details should not be inferred from the selection announcement.

What the platform could mean for India’s digital public services

India has built large digital services around identity, document access and electronic transactions. AI assistants could provide another layer on top of this infrastructure by helping people understand which service they need and how to use it.

Easier access to complex procedures

Citizens may struggle to find the right instructions when information is spread across different websites or written in administrative language. A conversational interface could let users describe a problem in ordinary language and receive more direct guidance.

This may be particularly useful for people who are less comfortable navigating complex websites or who need help understanding the steps required for a service.

However, accessibility will depend on the assistant’s ability to understand the user’s question, provide accurate instructions and support the languages and communication methods people actually use.

Potential support for Indian languages

CoRover promotes multilingual conversational AI, making language support a relevant capability for the government project. A common assistant could potentially help citizens interact with digital services in languages they are more comfortable using.

Yet broad language claims should be tested against real government tasks. Understanding everyday conversation is not the same as accurately interpreting document names, legal requirements, abbreviations and procedural instructions.

Performance may also differ between languages and between text and voice interactions. The government will need to assess whether the service delivers consistent results across the groups it is intended to serve.

A shared framework for departments

If the platform can be reused successfully, government departments may be able to adopt a common foundation rather than repeatedly building similar conversational systems.

A shared framework could make it easier to standardise basic functions such as user interaction, escalation, monitoring and service integration. It could also reduce duplicated development work.

The trade-off is that a common platform must be flexible enough to accommodate different departments’ rules and systems. A document service, a tax-related service and a public grievance portal may require different permissions, workflows and safeguards.

A shared technical foundation should therefore not mean identical policies or unrestricted access across services. Each integration needs clearly defined responsibilities and controls.

Security and privacy will determine whether citizens trust the platform

DigiLocker handles access to personal documents, making security a central requirement rather than an optional feature.

A conversational AI assistant connected to a public service could process questions containing personal details or information about sensitive records. If the system can perform transactions, it may also need to interact with authenticated services and approved application interfaces.

The main question is how the government and its technology partner will separate general information from actions involving a person’s account or documents.

Access controls and data minimisation

The platform should receive only the information required to complete an authorised task. A general question about how DigiLocker works should not automatically grant an AI agent access to an individual’s records.

For operations involving personal data, access controls should verify that the user is authorised and that the requested action is permitted. The system should also avoid collecting or retaining unnecessary information.

These are important design principles for any AI system connected to sensitive services. The exact technical controls for this project have not been fully detailed in the public reporting cited here.

Accuracy and accountability

An AI assistant can produce plausible but incorrect answers. In a government setting, such mistakes may lead to missed deadlines, incorrect applications or confusion about official requirements.

The system should therefore rely on authoritative information wherever possible, indicate when it is uncertain and provide a way to reach human support. Critical transactions should include appropriate confirmation steps and records of what was requested and executed.

Accountability also needs to be clear. Users should know which department is responsible for a service, how to challenge an incorrect response and where to report a problem.

Testing before wider deployment

A successful pilot on DigiLocker would provide evidence about the platform’s performance in a real public-service environment. Testing should cover common questions, unusual requests, multiple languages, authentication failures and attempts to make the system act outside its permissions.

A rollout timeline alone is not enough to judge readiness. The relevant measures include answer accuracy, successful task completion, escalation quality, response times, security incidents and user satisfaction.

The platform’s expansion to other departments should depend on the results of such evaluations rather than simply on the availability of the underlying technology.

The competitive and strategic significance

The contract reflects a broader shift in how governments and businesses are approaching AI. Early deployments often focused on chatbots that answered questions from a limited set of documents. Newer agentic systems are designed to connect conversations with workflows and actions.

For technology companies, government deployments can demonstrate whether their platforms can operate at scale under demanding reliability and security requirements. For the government, the potential benefit is a more consistent way for citizens to navigate digital services.

The project also sits within India’s wider AI ambitions. The IndiaAI Mission, launched in 2024 with an approved outlay of ₹10,372 crore, supports several parts of the domestic AI ecosystem, including computing infrastructure, innovation and the development of AI capabilities.

That broader programme should not be confused with this particular contract. The ₹10,372 crore figure is the IndiaAI Mission’s overall approved outlay, not the disclosed value of CoRover’s government AI platform project.

The strategic opportunity is to make AI useful in everyday public services while retaining control over sensitive data and decisions. Whether the platform becomes a model for other departments will depend on operational performance, governance and public confidence.

The Bigger Picture

CoRover’s selection points towards a more integrated approach to government AI in India: a reusable framework that could connect citizen questions with authorised digital workflows. Starting with DigiLocker gives the project a concrete service to test, while the “ask, do and escalate” approach recognises that public-service interactions sometimes require more than a generated answer. If successful, the framework could reduce friction for citizens and help departments avoid building separate conversational systems for every service.

The opportunity also brings substantial responsibilities. An AI assistant connected to government services must protect personal information, follow service-specific permissions, provide accurate guidance and recognise when a human should take over. A wider rollout should be driven by measured results rather than claims about AI capability alone. The project’s long-term significance will depend on whether it makes public services genuinely easier to use while preserving reliability, security and accountability.

Looking Ahead

The next milestones are the DigiLocker deployment, the publication of more detailed product and security information, and evidence of how well the system handles real citizen requests. The reported initial timeline of around four months should be treated as a target rather than a confirmed live date until the government or company provides an update. It will also be important to see which tasks the assistant can complete directly, which require user confirmation and how unresolved cases are routed to human support.

If the DigiLocker implementation performs well, the reusable framework could be considered for other government services. Expansion will require separate integrations, appropriate access controls and testing against each department’s rules and data. For now, CoRover’s selection marks the start of a government AI deployment effort—not proof that a fully autonomous assistant is already available across India’s public services.