For telecom operators, running a modern network has always been a question of scale. But as data demand accelerates and infrastructure expands, the challenge is as much about coordination as capacity.
Across fibre, core and 5G networks, operational models have traditionally relied on large teams of engineers to monitor performance, troubleshoot issues and manage repairs. However, as networks grow ever more quickly in both size and complexity, that model is becoming increasingly difficult to sustain.
According to Rick Lievano, worldwide chief technology officer for telecommunications at Microsoft, the industry is now undergoing a fundamental shift. “Telecom operators are looking to move from traditionally managed, very operationally heavy networks to something far more autonomous,” he says.
That shift is being driven by advances in generative and agentic AI, which are enabling operators to move beyond static, rules-based automation. “In the past, automation was very much ‘if this happens, then do that’,” says Lievano. “Now we have the ability to use generative reasoning to make better decisions based on context, and to automate in ways that are much closer to how a human would operate.”
The transition is not a single leap, but a structured journey towards autonomy. The industry framework defined by TM Forum outlines a progression from manual operations – level 0 – through assisted and partially autonomous environments, to fully autonomous networks – level 5 – where systems can self-heal and self-optimise and humans provide final approval and accountability.
While full autonomy remains an ambition rather than a reality, Microsoft has been pushing the boundaries of what is possible in the management of its own Microsoft Azure network. With more than 600,000 kilometres of fibre optic and undersea cables and over 400 data centres in operation around the world, the network has been undergoing rapid expansion. To keep up, Microsoft’s network operations team has turned to AI agents to manage, troubleshoot and coordinate repairs.
These agents have been embedded directly into everyday workflows as what it describes as digital co-workers. Rather than building a single, monolithic AI system, the company has developed a line-up of specialised agents. “We realised early on that one big copilot wasn’t the right approach,” explains Lievano. “Instead, we’ve created agents that are subject-matter experts, modelled after a specific job role. An agent called Paul, for example, is a support engineer, who understands every piece of equipment we have deployed. Niobe, meanwhile, is a network operations centre manager who assigns cases, analyses reports and triages open cases.”
Another example is Miles, the agent responsible for managing fibre repair workflows. When an issue is detected, Miles identifies the location of the fault, determines the appropriate repair partner, and contacts them to dispatch a field crew. It continues to track progress, request updates and validate the repair once completed, ensuring that the issue is fully resolved.
Telecommunication providers are using AI agent capabilities to orchestrate workflows, analyse large volumes of data and apply organisational data to operations (Photo: iStock/halbergman)
“Today, Miles handles around 98 per cent of fibre outages,” says Denizcan Billor, engineering leader at Microsoft. “It’s effectively doing the work that a team of engineers would have done before.” The key advantage lies in speed and consistency. “Agents respond immediately,” he says. “They don’t wait, they don’t get distracted. They compress all the latency that humans naturally introduce into these processes.”
These agents operate within familiar tools such as Teams and email, allowing them to integrate into existing processes without requiring new interfaces or workflows. They can monitor network conditions, analyse incidents, coordinate responses and communicate with both internal teams and external partners, effectively taking on tasks that would previously have required continuous oversight from engineers.
Despite this, the role of human engineers remains central. “Previously, the majority of engineering effort within operators was focused on acquiring data, analysing it, reviewing output and orchestrating workflows, while decision-making and approval accounted for a relatively small proportion of the overall workload,” says Billor. “With AI capable of orchestrating complex work across teams, analysing large volumes of data and applying organisational knowledge, human attention is now on execution oversight. This includes identifying and addressing gaps in AI performance, providing ongoing training and retaining responsibility for approving and carrying out high-risk or irreversible actions.”
Having developed these capabilities within its own global network, Microsoft is now making them available to telecom operators through a framework designed to accelerate adoption. The approach provides a foundation that organisations can build on, combining pre-developed components with their own data and operational processes. “We can get customers a large part of the way there,” Lievano says. “They still need to connect it to their own systems, but the foundation is already in place.”
For telecom operators looking to adopt similar approaches, data remains a critical consideration. “Agents are only as good as the data they have access to,” Lievano notes. In many organisations, data is still siloed and difficult to access, limiting the effectiveness of AI-driven systems. Addressing this requires a more unified approach to data management, ensuring that information from across the network can be accessed, understood and acted upon in a consistent way.
Another key challenge is ensuring determinism in decision-making. While large language models are powerful, their outputs can vary depending on context, which is not acceptable in network operations. “You need consistent, repeatable outcomes,” Lievano explains.
To achieve this, Microsoft combines AI reasoning with curated data, predefined workflows and validated queries, reducing variability and increasing trust in the system. “AI can leverage deterministic systems to enforce safety constraints, while routing irreversible actions to human operators, ensuring that ultimate control remains with people,” says Billor. Lievano adds: “Developing and refining these practices has been part of our journey, and we can now provide that guidance to our partners.”
For operators, there is proven value in increasing the level of autonomy in their network. In its Scaling the AI-native telco report, McKinsey found that there was a 30 per cent reduction in network downtime from deploying AI for network management and maintenance, while Capgemeni reported in Networks with Intelligence 2024 that 71 per cent of telecom operators had reduced their energy consumption through autonomous network initiatives.
As demand for connectivity continues to grow, the pressure on telecom operators to embrace these improvements in efficiency and resilience will only intensify. For Lievano, the case for change is inevitable. “Autonomous networks are a high priority for virtually every single operator because they spend so much money managing their network,” he says. “With the pace of growth today, the traditional approach of simply adding more engineers isn’t going to be sustainable.”
The introduction of AI agents is not, therefore, simply a case of automation. They are redefining how work is organised, creating systems where humans and AI collaborate to deliver faster, more reliable and more intelligent networks that are prepared for the unprecedented demands on modern telecom operators.
“When we first started this project, the models were about the level of a good engineering intern,” says Lievano. “Now they are essentially senior engineers, with graduate-level knowledge. They’ve come a long way in just a few years, and they’re only going to get better.”
Partner perspective
We asked Microsoft partner Coretek how it is using cloud and AI solutions to help telecommunications providers enhance their operations
“Coretek partners with telecommunications providers to modernise network operations and reimagine customer engagement using Microsoft’s cloud and AI stack,” said Brian Barnes, chief technology officer at Coretek. “By deploying Microsoft Copilot Studio virtual agents and Azure AI services, we enable intelligent support automation, predictive network management and hyper-personalised service recommendations. A large telecom provider in the mid-west US implemented an agentic AI solution that deflected 45 per cent of inbound support contacts through intelligent virtual agents while cutting mean resolution time by 30 per cent. Employees transitioned from reactive troubleshooting to proactive value delivery – demonstrating how AI-driven modernisation reshapes both operational efficiency and competitive customer experience simultaneously.”
Brian Barnes
Chief Technology Officer, Coretek
Discover more about how organisations are using Microsoft’s AI technology in media and insights from experts at Microsoft Partner businesses including Intermedia, Velosio and AVEVA in the Summer 2026 issue of Technology Record. For more content like this, subscribe to the print edition or free digital edition of the quarterly Technology Record magazine.