In a conversation with Sam Keys Toyer, global head of portfolio for network managed services at Ericsson, a clearer picture emerged of how one of the sector’s largest vendors is attempting to answer it: not through new spectrum or hardware cycles, but through software, automation and, increasingly, AI-led operations.

Toyer’s remit sits firmly in the operational heart of the network. Ericsson’s managed services business today spans mobile, fixed, cable and IoT environments, often in complex, multi-vendor settings.

The proposition is straightforward in principle: take over the running of networks on behalf of communications service providers (CSPs) and deliver outcomes such as higher availability, improved performance and lower cost. In practice, however, that remit has become significantly more complex as networks have evolved.

This complexity is precisely where Ericsson sees its opportunity. Over the past several years, the company has rebuilt its operational model around what it calls the Ericsson Operations Engine, a framework that integrates people, processes, tools and security into a single delivery architecture. At its core is a shift towards data-driven operations, underpinned by extensive automation.

According to Toyer, Ericsson now runs tens of thousands of automated processes every second across dozens of global contracts. Around 80% of operational tasks have some degree of automation embedded. But while automation is not new to telecoms, the emphasis is shifting from task execution to decision-making.

“We started with automation and data foundations,” Toyer explains. “The next step was applying machine learning to predict faults before they happen. Now, we’re moving into systems that can make complex decisions.”

This progression reflects a broader industry ambition: the move towards autonomous networks. Working alongside frameworks such as those developed by TM Forum, Ericsson is targeting higher levels of operational autonomy, where networks can analyse, decide and act with minimal human intervention.

Central to this evolution is the concept of intent-based operations. Rather than instructing a network how to perform a task, operators define the desired outcome. The system then determines how best to achieve it.

One example Toyer highlights is energy optimisation. Operators can set an “intent” to minimise power consumption without degrading user experience. The system continuously monitors network conditions, predicts when adjustments are needed and executes them automatically—without human involvement.

A similar approach is being applied to service assurance. For instance, operators can define quality targets for video services on platforms such as YouTube or WhatsApp during high-traffic events. The network then dynamically reallocates resources, tests potential solutions, sometimes using digital twins, and implements the optimal configuration.

Crucially, these systems are not operating blindly. Observability layers provide full transparency, audit trails track every action, and “human-in-the-loop” controls can be introduced where required. This reflects a pragmatic acknowledgement that trust in AI remains a work in progress.

“We’re still in the early stages,” Toyer notes. “But when you can show how these systems work, and provide the right level of explainability, confidence builds.”

That confidence is essential, particularly as AI begins to encroach on areas traditionally reserved for human judgement. Network operations are filled with trade-offs, adjusting one parameter to fix an issue can create another elsewhere. Replicating this kind of contextual decision-making has long been a barrier to deeper automation.

Ericsson’s approach is to combine machine learning with newer reasoning techniques and agent-based models, enabling systems to weigh competing outcomes more effectively. Yet Toyer is clear that this is about augmentation, not replacement.

“Humans are not leaving the loop,” he says. “This is about using machines to increase productivity and handle complexity, while people focus on higher-value activities.”

That balance is likely to define the next phase of telecoms operations. As networks become more programmable and service-driven, the ability to manage complexity at scale will be critical to unlocking new revenue streams, particularly in enterprise and private 5G use cases.

For Ericsson, the bet is that operational transformation is the missing link between 5G capability and commercial success. The technology itself can already support differentiated services, but its complexity has limited widespread monetisation.

“If we can simplify how networks are operated,” Toyer argues, “we open the door to the full promise of 5G service diversity, differentiated connectivity, and ultimately, new revenue.”

In an industry still searching for growth, that promise remains compelling. The question now is whether AI-driven operations can deliver on it.

RELATED STORIES

Ericsson, SK Telecom sign MoU to advance 5G and 6G deployment

Ericsson: Turbocharging 5G, market collaboration and accelerating UK connectivity