Frequently Asked Questions
What is ‘agentic AI’ in telecom, and how is it different from a chatbot or automation script?
Agentic AI describes systems that go beyond responding to a single prompt or executing a fixed script; they observe ongoing conditions, reason about what action makes sense, and execute that action with some autonomy, operating within policies and guardrails set by network engineers. In telecom, this might mean an AI agent noticing a cell site’s performance degrading, diagnosing the likely cause, and either fixing it directly or dispatching a technician with context already prepared, without a human manually triaging the issue first. This differs from older rule-based automation, which only executes pre-written if-this-then-that logic, and from generative AI chatbots, which mostly produce text rather than taking independent action on live infrastructure. Telefonica’s virtual assistant Aura, for example, has scaled to handle hundreds of millions of interactions annually across dozens of channels, increasingly augmented with generative capabilities for more natural replies.
Are telecom networks becoming ‘AI-native,’ and what does that actually mean in practice?
Yes, and the shift is structural rather than cosmetic. For years, telecom infrastructure moved toward cloud-native design, meaning network functions ran as software on standardized servers instead of dedicated hardware. The current shift goes further: that same cloud infrastructure is being redesigned specifically to run AI workloads efficiently, supporting GPU orchestration, real-time inference, and generative AI models alongside traditional network functions. Standards organizations including ETSI, the Linux Foundation, CNCF, and TM Forum are actively working to formalize technical frameworks for this transition, since AI-native infrastructure needs different performance guarantees than the general-purpose cloud infrastructure most networks were originally built on. Vendors have responded with dedicated products; Huawei, for instance, launched a hyper-converged infrastructure platform aimed at letting operators run general-purpose and AI workloads on unified infrastructure.
Will AI replace telecom jobs, or change what those jobs look like?
Industry research suggests the more immediate effect is workload reduction rather than outright replacement. Surveys of telecom operators found that a strong majority expect AI-driven reductions in routine workload to actually improve retention in technical and customer-facing roles, on the logic that overworked, understaffed teams are more likely to burn out and leave, while AI handling repetitive tasks frees skilled staff for higher-value work. This doesn’t mean headcount is immune to change; certain repetitive or purely transactional roles are likely to shrink over time. But the dominant framing from the industry in 2026 is augmentation: extending the reach of existing specialized staff, often in short supply industry-wide, rather than purely cutting roles.
What is ‘AI sovereignty,’ and why are telecom operators interested in it?
AI sovereignty refers to the push, largely driven by national governments, to ensure AI models, the data they’re trained and run on, and the infrastructure powering them stay within a country’s borders rather than depending entirely on foreign cloud providers. This matters to telecom operators because they already own much of the infrastructure such sovereignty requires: extensive fiber networks, edge computing sites, and increasingly, data centers. Rather than just being connectivity providers, operators are positioning themselves as the infrastructure backbone for sovereign AI deployments, partnering with or competing against traditional cloud hyperscalers for this business. Industry analysts describe this as one of the most significant new infrastructure opportunities operators have seen in years.
How is AI changing the core network itself, not just customer-facing applications?
Several vendors are now embedding AI directly into core network functions rather than treating it purely as an add-on analytics layer. Huawei’s AgenticCore solution, for example, is designed to embed AI across the entire core network stack, including mobile data handling, voice services, network operations and maintenance, and the underlying telco cloud infrastructure that runs all of it. The broader industry framing is that networks are evolving from simply transporting data to actively hosting, managing, and orchestrating AI-driven processes themselves, meaning the network increasingly becomes a platform for running AI workloads, including third-party ones, rather than just a pipe carrying traffic between endpoints. This shift is closely tied to edge computing, since running AI inference physically close to users is part of what makes real-time, AI-driven network behavior possible.
What are the biggest challenges operators face in deploying AI at scale?
Operators describe three recurring hurdles. First, the cost-efficiency of AI computing power; running AI workloads, especially generative AI and large models, at carrier scale is expensive, and operators are still working out which use cases generate enough value to justify that spend. Second, unifying AI across multiple generations of network technology, since most carriers operate a mix of legacy 4G, current 5G, and emerging 5G-Advanced infrastructure simultaneously, and AI systems need to work across all of them rather than in isolated silos. Third, tailoring AI models to specific telecom scenarios rather than relying on generic, off-the-shelf AI tools, since network operations have domain-specific requirements, like real-time decision-making at extremely high reliability, that general-purpose AI products aren’t necessarily built to handle out of the box.
Is AI helping or hurting network security?
It cuts both ways. On the defensive side, AI dramatically speeds up threat detection by spotting unusual patterns across enormous volumes of network traffic far faster than manual monitoring, and it increasingly powers automated incident response. On the risk side, AI agents given operational access to network systems represent a new kind of attack surface; if an agent’s decision-making can be manipulated or its access misused, the consequences could be more severe than a typical software vulnerability, since the agent is specifically designed to take autonomous action. Security researchers are increasingly studying AI-related risks alongside other emerging threats like quantum computing, treating AI agents with network access as a risk category needing dedicated scrutiny rather than assuming existing security tools automatically cover it.
