For the past few years, the conversation around artificial intelligence (AI) has been dominated by generative AI, the systems that write, summarise, translate and converse with fluency. However, a new wave, that is, agentic AI is now reshaping how enterprises think about automation. Unlike traditional AI models that respond to a single prompt and stop, agentic AI systems can plan, reason, make decisions, use tools and execute multi-step tasks autonomously, often with minimal human supervision. While a generative AI chatbot might draft an email if asked, an agentic AI system can figure out that an email needs to be sent, decide who should receive it, check the calendar for the right time, draft the content, send it and then follow up if there is no response, without any human intervention.
Telecom operators are among the industries where this transition could have the greatest impact. The sector combines enormous volumes of customer interactions, complex networks, vast quantities of data and constant pressure to reduce costs, while improving service. For India, with hundreds of millions of mobile subscribers and a highly diverse customer base, agentic AI could become one of the most consequential operational technologies of the decade.
Why now?
Three developments have brought agentic AI closer to practical enterprise deployment. First, large language models have become reliable enough at reasoning and instruction, following to be trusted with multi-step tasks. Second, standardised protocols have emerged for connecting AI models to external tools and data sources, allowing an AI system to safely trigger an action in another piece of software. Third, enterprises have accumulated years of data such as customer records, network logs, transaction histories that agentic systems can now use to make grounded decisions. Together, these developments are moving agentic AI from impressive demonstrations towards real-world workflows.
Telecom: Where agentic AI becomes operational
India’s operators are no longer experimenting only with chatbots that answer customers’ questions or models that summarise information. They are building systems that can interpret an objective, work across multiple databases and applications, make decisions and execute the next step.
The telecom sector can be one of the biggest beneficiaries of agentic AI because of enormous volumes of structured and unstructured data. Every day, operators manage millions of customer interactions, network events, payments, service requests, field interventions, fraud signals and enterprise transactions. Much of this activity involves connecting multiple systems and making decisions based on changing circumstances. This makes the sector a natural fit for agentic AI.
India offers a particularly interesting environment for agentic AI because of the sheer scale and diversity of its digital economy. Operators serve subscribers across urban centres, smaller towns and rural areas. Customer interactions can happen in English, Hindi and numerous regional languages. At the same time, telecom companies operate enormous physical infrastructures and increasingly sophisticated digital networks. India’s digital infrastructure also provides a strong foundation for this next phase of automation. The country’s decade-long expansion of digital public infrastructure, including Aadhaar and Unified Payment Interface has created large-scale, interoperable digital systems on which new AI-driven services can operate.
The potential applications span the telecom value chain. In customer service, an agent could verify identity, understand a customer’s problem, assess available plans, recommend an appropriate service, activate it, resolve a billing issue and follow up if the problem remains unresolved. In network operations, agents could correlate alarms from multiple systems, identify likely root causes, recommend remediation and execute low-risk corrective actions, while escalating high-impact changes to human engineers. Field-service agents could diagnose faults, check spare-parts and technician availability, schedule repairs and update tickets automatically. In fraud and revenue protection, agents could investigate suspicious behaviour across SIMs, devices, payments and network activity. Enterprise sales agents could research accounts, identify opportunities, prepare proposals, coordinate internal teams and maintain records, allowing human sales teams to spend more time on relationships and strategic decisions. The important change is that these systems can potentially coordinate the entire workflow rather than performing just one task within it.
From chatbots to digital operators
Recent initiatives by India’s telecom companies illustrate this transition. Reliance Jio’s Jio Call Agent and MyJio.AI Care, announced in June 2026, represent an evolution beyond conventional customer-service chatbots. Rather than simply telling a customer which roaming pack to choose or explaining how to complete a transaction, an agent can interpret the request, access relevant account information, determine what needs to happen and, with user consent, execute tasks such as activating services or making bookings. That is the defining characteristic of an agentic system.
Bharti Airtel is pursuing a similar direction, building a foundational agentic platform that includes an agentic SDK, evaluation frameworks and a no-code builder. AI agents are already being incorporated into customer-facing operations, including inbound voice, next-best-action decisioning and proactive assurance. Airtel has also reported more than 4 million customer interactions handled daily through AI-powered voice agents across eight languages. Its broader approach includes agents for buying, billing, payments and customer care.
From detection to investigation
India’s telecom industry also provides a particularly strong use case for agentic systems because fraud is not a single-event problem. A suspicious transaction, SIM activity, call pattern or message may look harmless in isolation. The signal becomes meaningful when it is connected with other events.
Airtel’s recent AI initiatives show how operators are moving intelligence closer to the network itself. In February 2026, the company announced an AI-powered autonomous solution designed to detect and intervene against fraud linked to OTP leakage at the network layer. The trials had demonstrated significant effectiveness, building on its earlier AI-based spam-call and malicious-link protection. In March 2026, Airtel and Google announced an AI-powered RCS experience with enhanced spam and fraud protection. The initiative had blocked 71 billion spam calls and 2.9 billion spam SMSs.
The next step can be potentially more interesting with building agents that investigate rather than merely detect. An agent could connect suspicious call behaviour with subscriber history, device information, messaging activity and transaction signals; determine a risk level; apply an appropriate intervention; and escalate unusual cases to a fraud analyst. Instead of beginning with a collection of disconnected alerts, the investigator would receive a contextualised case with the relevant evidence already assembled. This could fundamentally change the economics of fraud management, particularly at India’s scale.
India’s multilingual opportunity
Because of its linguistic diversity, India could make telecom one of the most important proving grounds for agentic customer service. An Indian customer should not have to translate a complicated telecom problem into English before an AI system can solve it. Jio’s 2026 AI demonstrations explicitly position its services for Indian languages, while Airtel is already operating AI-powered voice agents across eight languages. The real opportunity, however, is not merely multilingual conversation. A customer could speak in Hindi, Tamil or another regional language, while the agent works underneath in English language, billing platforms and network-management tools. The customer sees a natural conversation, whereas the enterprise sees a structured workflow.
Beyond telecom
Telecom may be the strongest example in India, but the same design pattern is appearing in other industries. In oil and gas, Indian Oil is collaborating with IIT Bombay’s Centre of Excellence in Oil, Gas and Energy on AI-driven pipeline integrity. In May 2026, the collaboration specifically discussed the role of agentic AI in the oil and gas industry, against the backdrop of Indian Oil’s more than 20,000-km pipeline network. The system could detect any abnormal operating condition, optimise parameters and notify the relevant teams.
Transportation offers another interesting opportunity. In October 2025, Adani Airports announced a partnership with AIONOS to deploy a multilingual, omnichannel agentic AI solution across its airports. The system is designed as a 24×7 digital concierge that can provide flight, gate and baggage information and support passengers across voice, chat, web and mobile in English, Hindi and regional dialects.
In financial services, RazorPayX launched what it calls India’s first agentic connected banking platform in June 2026. Its Payout Agent can identify a payee, suggest whom to pay and when, prepare the payout and take the user to the OTP approval step. Razorpay had earlier launched Razorpay Agent Studio, a platform designed around agents that can manage payments and financial operations.
The road ahead
The biggest challenge is not intelligence. It is trust. A chatbot providing an incorrect answer can be frustrating. An autonomous system making an incorrect network change, approving a payment, suspending a legitimate customer or altering a critical industrial process is a fundamentally different risk. That is why the most practical enterprise deployments are unlikely to involve unrestricted autonomy. Instead, the emerging model is graduated autonomy. An agent may investigate independently. It may recommend an action. It may automatically execute low-risk changes, while requiring human approval for high-impact decisions.
India is well placed for the transition, given its telecom scale, digital infrastructure and technology ecosystem. The key question is no longer whether AI will become part of telecom, but how much of the network, customer journey and commercial operations can safely be entrusted to it. For telecom operators, the advantage will come not from the flashiest AI demo, but from the infrastructure behind it which includes connected data, modern OSS/BSS systems, reliable APIs, network intelligence, strong governance and secure permissions. The next competitive edge will belong to operators that can build a reliable bridge between AI intelligence and real-world execution.
Vaishnavi