As AI applications become more widespread, the demands on telecommunications (telco) networks are evolving.
Singtel Singapore CTO Tay Yeow Lian expects data to move significantly upstream, computing to be distributed across devices, edge, and cloud, and applications to increasingly require consistent performance, responsiveness, and resilience.
“This creates two priorities for the industry: building networks for AI, and using AI to make networks more intelligent,” he posits.
As a result, the evolving landscape will require more dynamic AI networks and a deeper understanding of application requirements. He also noted the need for sufficient connectivity, computing, and network resource allocation.
“AI for networks will embed intelligence more deeply into operations — anticipating demand, identifying anomalies earlier, diagnosing problems faster and progressively automating the appropriate response,” he said.
AI-ready network
According to Tay, an AI-ready network is designed to support increasingly intelligent applications while using intelligence to improve how the network operates continuously.
This involves real-time data, AI, automation, and orchestration, enabling the network to understand changing conditions and respond dynamically.
“Modern telecom networks are already highly advanced and software-driven. The next stage is to connect intelligence more deeply across different network domains and give AI the trusted context required to make increasingly sophisticated decisions,” Tay shared.
He emphasised the importance of context, as telecom networks span mobile, fixed, transport, cloud, and other systems, often involving multiple technologies and vendors.
“AI needs an accurate understanding of network topology, performance and operational knowledge, including how an action in one domain could affect services elsewhere,” Tay said, adding that this enables better decision-making based on the end-to-end customer experience instead of an individual network component.
He added: “Governance is equally important. AI needs clearly defined permissions, traceable decisions and actions, and mechanisms for human intervention when required.”
AI networks in operation
AI offers vast opportunities across the entire network lifecycle.
“One of the most valuable areas is resilience and service assurance,” Tay said, adding that AI can continuously analyse network behaviour, identify anomalies earlier, correlate signals across multiple systems, and help engineers determine root causes before problems escalate.
He shared that Singtel leverages AI-powered broadband anomaly detection, which has reduced detection lead time by 89% and given teams more time to intervene.
Telecom companies can also maximise the technology in fault diagnosis and field operations to determine whether an issue can be resolved remotely and, when physical intervention is needed, identify the appropriate expertise and response.
“This reduces investigation time and helps field resources respond more effectively,” Tay shared.
Another use case is change-risk management, where AI can analyse historical changes, dependencies, and current network conditions to identify potential service impacts before changes are implemented.
“The longer-term opportunity is to move progressively from detecting and responding to network events towards anticipating and preventing them, while keeping engineers responsible for governance, complex decisions and outcomes,” Tay underscored.
This shift towards more autonomous operations is also reflected in Forrester’s analysis of AI-powered telecommunications providers. In a May 2026 blog, Forrester principal analyst Tom Mouhsian highlighted predictive fault detection, network capacity planning and autonomous, self-healing networks among the emerging applications of AI in network engineering.
Addressing common challenges
“As telcos apply AI more deeply across network operations, three challenges stand out: fragmented data and knowledge, multi-vendor complexity, and building trust in greater autonomy.” Tay Yeow Lian
In preparing for the AI era, telcos need to address the most common barriers.
“As telcos apply AI more deeply across network operations, three challenges stand out: fragmented data and knowledge, multi-vendor complexity, and building trust in greater autonomy,” Tay shared.
Tay said telcos first need to connect context that often lives across systems and teams.
“Networks have evolved over decades, so data, network relationships and operational knowledge can become fragmented. We are improving data quality, mapping how network components relate and codifying engineering knowledge to give AI reliable context for operational decisions,” he said.
Next is managing intelligence across multiple technology domains and vendors.
“Our approach keeps specialised intelligence close to the underlying network technology while allowing us to orchestrate across domains and maintain an end-to-end view of service and customer impact,” he said.
Tay also emphasised the importance of introducing autonomy progressively.
“AI should first observe and recommend, then execute based on human-defined actions and approvals. As we build evidence and strengthen monitoring, governance and rollback mechanisms, we can move selected processes towards closed-loop operation, with engineers defining the parameters and remaining accountable,” he explained.
Preparing for the agentic AI era
“The architecture and governance decisions they make today will determine how effectively they can scale AI over the next five years.” Tay Yeow Lian
If asked for advice as telcos move towards the agentic AI era, Tay reminded organisations:
“The architecture and governance decisions they make today will determine how effectively they can scale AI over the next five years.”
He reminded telcos to start with the operational outcomes they want to improve, define how they will measure success, and determine the appropriate level of autonomy.
“This keeps AI investment tied to business and customer value,” he said.
Equally important is building reusable foundations early, including common data, knowledge, APIs, and governance. This allows operators to scale use cases without rebuilding underlying capabilities each time.
He reminded organisations that the goal is to make each successive deployment faster and more efficient than the last.
“As technology will continue to evolve, it is better to have an open, modular architecture that gives leaders the flexibility to work across different technologies and vendors without creating new AI silos,” he said.
For Tay, the most important thing is to invest in people alongside technology.
“As AI takes on more operational tasks, people will play a greater role in setting intent, exercising judgement, managing exceptions and providing accountability,” Tay said.
Preparing telcos for the agentic AI era
The era of agentic AI is not coming; it is already here.
As technology rapidly advances, telcos should evolve alongside it.
With the right strategies, investments, and priorities, the telecommunications industry will be better prepared for future developments.
