The generative artificial intelligence (GenAI) transformation is not only helping India’s mobile phone operators improve network and operations but also opening up new monetisation opportunities. The country’s two largest carriers, Reliance Jio and Bharti Airtel, have partnered with big tech companies to co-develop AI models and offer application-as-a-service.
Bharti Airtel’s partnership with Google is aimed at unique tech and telco AI solutions such as those in geospatial mapping with location intelligence, voice analytics trained across languages and marketing with high-precision ads targeting. Meanwhile, Jio Platforms, the telecom and digital services arm of Reliance Industries, is co-developing AI language models in partnership with Nvidia to build use cases in areas such as retail, healthcare, agriculture and education.
AI offers transformation in telecoms with abilities such as “learning, reasoning, perception, problem solving, data analysis and language comprehension”, Bharti Airtel said in response to ET’s queries.
“Airtel has partnered Google to bring together each of their respective strengths in connectivity and AI technology to develop industry-leading AI/ML (machine learning) solutions that Airtel will train on its large data sets,” it said, adding that the partnership is aimed at accelerating GenAI deployment in India.
Bharti Airtel is also using AI in areas such as anti-spam solutions, data mining for site installations, predictive analysis for network management and self-healing, and green 5G through efficient energy usage.
Meanwhile, telecom market leader Reliance Jio plans to offer affordable and personalised AI-as-a-service or AI agent applications on top of these models through Jio Brain – a full stack enterprise suite.
Jio Brain, which offers ML-as-a-service for enterprises, doesn’t need users to have costly infrastructure to build AI applications. “You can just tag on to JioBrain, and we’ll launch that in the coming quarters as we perfect the use cases for that,” Aakash Ambani, chairman, Reliance Jio, said at an event last week.
On the infrastructure side, Jio is building a 1GW AI data centre in Jamnagar both for captive use and offering GPU-as-a-service to commercial projects.
Ambani added that the company has a 1,000-plus team of data scientists, researchers and AI engineers who are overseeing AI integration across group functions.
“Today we monitor our network that now covers about 95% of the living population of India. We can proactively monitor, and before even a customer has a bad experience, we can predict it,” he said.
Commitment to GenAI by global telecoms is picking up pace. A 2023 study by NVIDIA found that nearly 90% of global telecom companies are utilising AI, with 48% in the piloting phase and 41% actively deploying AI solutions. Iliad, a French telecom group, announced plans to invest €3 billion to enhance AI infrastructure in Europe. Germany’s Deutsche Telekom aims to leverage AI to generate about €1.5 billion in new revenue streams and reduce costs by €700 million by 2027. On the usage side, Swedish telecom equipment maker Ericsson showed that telecom companies can enhance their 5G revenues by 5-12% by offering differentiated, high-quality connectivity for GenAI apps.
However, AI in networks presents challenges.
ML is already helping operators reduce costs with network automation, predictive maintenance and anomaly detection, but “scaling these solutions across complex, multi-vendor environments remains a hurdle”, said Nitesh Bansal, global CEO and managing director, R Systems, a digital engineering firm involved in core research and development with global tech-telco platforms.
The company is working on an AI-powered solution which could cut power consumption for Radio Access Networks by up to 35%, extending to older radio networks such as 4G and 3G.
“GenAI brings promise in areas like intelligent service provisioning and automated troubleshooting, but its practical deployment is still evolving. Operators are looking for AI use cases that deliver measurable ROI (return on investment), beyond just operational efficiencies,” Bansal said.
He said that with evolving consumption patterns, fine-tuning these models remains critical to make AI truly pervasive.
