Agentic AI: Transforming telcos into techcos

As telcos evolve to stay competitive, innovate faster and diversify revenue streams beyond traditional voice and data services, there is an urgent need to rethink how these services are delivered to their customers.

Over the past year, generative artificial intelligence (AI) has been at the forefront of digital transformation agendas across industries. Today, we are entering the era of agentic AI, where AI does not just answer questions or optimise tasks but becomes a capable, collaborative agent. These autonomous systems can set goals, take initiative, interact across tools and systems, and make decisions in real time.

As AI agents transform diverse areas across industries, communication service providers (CSPs) are embracing new ways of working by developing different AI agents across their consumer services, business-to-business operations and captive units. These areas primarily revolve around customer experience enhancement, generating insights for operational efficiency and predictive analytics to unlock new upselling or cross-selling opportunities.

Agentic AI for autonomous networks

CSPs have made several strategic investments in analytics-driven network operations to manage the growing complexity of mobile wireless infrastructure, particularly in radio access networks (RANs), where multi-vendor environments, diverse technologies and surging data volumes present ongoing operational challenges. Now, with the advent of agentic AI, telcos are moving beyond basic automation toward intelligent, zero-touch operations. This next phase enables systems to autonomously detect anomalies, analyse root causes, assess potential impacts and execute corrective actions without human intervention.

Despite this progress, a key gap remains – the disconnection between network performance metrics and the actual customer experience. In many organisations, data is still analysed in operational silos, where network teams focus on infrastructure key performance indicators (KPIs), often unaware of user-level issues unless a service ticket is raised. This reactive approach limits proactive problem-solving and directly affects customer satisfaction. To fully realise the value of automation, telcos must deploy AI agents that can break down these silos by linking operational insights with user experience data to enable true end-to-end intelligence and service assurance. To achieve this, telcos must implement a well-defined framework with the following capabilities:

• Utilisation of distributed telco datasets: CSPs often capture the performance of their network and customer experience in different datasets, including configuration management, performance management, fault management and other systems that track customer sentiment and satisfaction ratings. These datasets are generally distributed across multi-vendor systems and utilised by different teams in silos. To give a unified and proactive customer experience, we need an intelligence layer to understand and normalise the data from these systems and define a common baseline of individual KPIs.
• Proactive monitoring and predictive analytics: Once baselines are defined, another intelligence layer should be integrated to analyse continuous data streams from systems and predict anomalies or deviations. These anomalies must also be mapped to risk scores to enable impact analysis and prioritisation. This intelligence layer should be trained on telecom-specific metrics, KPIs and CSP-specific thresholds to identify deviations from baselines.
• Execution layer for actions: Deviations should be relayed to the execution layer of AI agents in real time. This layer would then assess the results of risk impact analysis and initiate appropriate autonomous actions across various components of the network. To enable effective resolution, this layer requires a well-defined and standardised integration framework to interact with the target systems or tools. This can be achieved through a dedicated multi-agent system layer, where individual agents are assigned specific tasks as per predefined instructions and goals.

CSPs lay the foundation for nationwide critical infrastructure. Therefore, it is crucial to achieve the required level of accuracy and validate the results of each layer of the above framework. This demands strong collaboration between industry experts, network engineers and AI agents.

Building agents at scale

Leveraging cloud platforms or hyperscalers such as Google Cloud to build this complex framework of enterprise AI agents is not just convenient, it is essential for telcos. These hyperscaler platforms offer the required scalability, integration capabilities and, most importantly, AI-native infrastructure needed to build, deploy and scale agentic AI systems effectively. CSPs can benefit from strong partnerships with hyperscalers to build agentic AI capabilities at scale.

Apart from the core AI-ready infrastructure, hyperscalers provide additional benefits, including:

• Access to advanced AI tools: Enterprises often need access to pre-built AI services with options of first-party and open-source large language models, multiple agentic frameworks, vector databases, embeddings and reinforcement learning with human feedback tools necessary for building reasoning-capable agents.
• Easy integration with enterprise ecosystems: AI platforms offered by the hyperscalers come with numerous supported, ready-to-use connectors for enterprise systems, significantly accelerating development. AI agents can seamlessly interact with required databases, application programming interfaces, customer relationship management systems and enterprise resource planning solutions, enhancing their effectiveness within real business workflows.
• Enterprise-grade security and compliance: Hyperscalers offer robust security frameworks, such as identity and access management, encryption and audit logging, and most importantly, they support compliance with regulations such as the Digital Personal Data Protection Act, 2023 and those formed by the Telecom Regulatory Authority of India.
• Cost optimisation and pay-as-you-go: CSPs can avoid upfront infrastructure costs and scale computing based on usage, while also gaining access to cost optimisation tools and tiered pricing models for AI workloads.
• Observability for strong governance: Hyperscalers provide built-in tools for observability, drift detection, feedback loops and performance monitoring of AI agents and associated infrastructure, which would help telcos maintain control and reliability of autonomous AI agents in production environments.

Addressing compliance and sovereignty needs

Some CSPs face stringent regulatory requirements or data sovereignty rules that mandate maintaining data residency and data processing strictly on-premises. However, to still leverage the advantages of cloud technology, hyperscalers like Google Cloud provide a suite of edge and even air-gapped cloud solutions that comply with these regulatory requirements. By offering managed platforms and AI services similar to those of public cloud, these edge solutions enable telcos to accelerate their adoption of cloud-native transformation across both IT and network operations.

In summary, it is crucial for telcos to adopt agentic AI frameworks across the network and IT domains, accelerating their transformation journey by leveraging the investments hyperscalers have made in democratising AI capabilities.