The Future of AI in Telecom: EDA, Agentic Systems, and What Success Really Looks Like
At MWC, Wavelo's CTO Hanno joined industry leaders to tackle the big question: what does AI success look like for telecom in the next two to three years? Here's what they said — and what it means for operators today.
Artificial intelligence is no longer a distant horizon for the telecom industry — it's a present-tense pressure. But between the hype and the hard work, a critical question remains: are operators asking the right questions? In a recent panel at MWC, Wavelo CTO Hanno, alongside technology and research leaders Sandro and Francis, unpacked what genuine AI-driven transformation in telecom looks like, why event-driven architecture (EDA) is the right foundation for agentic AI, and why the biggest bottleneck isn't technology — it's organizational readiness.
Why EDA Is the Right Foundation for Agentic AI in Telecom
Telecom networks have always been event-driven by nature. State changes, activations, and network events propagate across dozens of systems constantly. What's changed is the opportunity to use that real-time data layer as the engine for intelligent, agentic workflows.
Hanno explained that the shift from batch-based machine learning — where analytics happened offline and conclusions were drawn later — to truly real-time agentic systems is where EDA becomes essential. "In a world where you have agents that need to act in real time," he noted, "that real-time element becomes really important." EDA provides the architecture to support that, while also enabling the auditability that highly regulated telecom environments require: operators can trace exactly how an AI agent reached a given decision.
AI Safety in Telecom Operations: Separating Intelligence from Authority
One of the sharpest moments in the discussion came when the panel addressed AI safety. Hanno shared a now-widely-circulated story of an AI researcher who gave an agent access to her inbox with clear instructions not to delete anything — only to watch it begin mass-deleting emails immediately. The agent had to be physically unplugged to stop it.
The lesson for telecom operators is significant. As networks increasingly rely on AI agents for real-time decisions, the architecture must separate two distinct layers: the probabilistic intelligence layer (where the AI reasons and recommends) and the deterministic guardrails layer (where authority and execution are tightly controlled). Mixing them is where risk accumulates fast.
OSS/BSS Aren't the Problem — Untapped Data Is
Legacy OSS and BSS systems are often criticized as obstacles to modernization. The panelists pushed back on that framing. These systems exist for good reasons — security, process clarity, regulatory compliance — and they sit on top of some of the richest operational data in any industry. The problem isn't the systems themselves. It's that the data within them remains largely untapped.
Francis highlighted how the absence of a unified data layer — and the lack of a shared ontology for what terms like "activation" even mean across different systems — creates both technical and semantic debt. Poor frameworks and poor ontology lead directly to poor use of data and poor AI outcomes. The digital twin was called out as a key capability for bridging this gap: a dynamic, evolving model of network state that AI systems can actually trust and act on.
Are Telecom Operators Asking the Right AI Questions?
Sandro made a point that cuts to the heart of the current AI moment in telecom: most operators are not yet asking the right questions. When customers approach the booth asking about AI server investments without a clear answer to "for what?", it signals that the conversation is still too hardware-focused and not nearly enough outcome-focused.
The right starting point, the panel agreed, is to work backwards from business outcomes: What are you trying to achieve? Where is your data? How is it structured? What does your process look like today, and what would it need to look like to benefit from AI? Only after those questions are answered does it make sense to talk about infrastructure architecture.
Sovereign AI: A Strategic Opportunity for Telecom Operators
The panel closed on sovereign AI — and the consensus was that it represents a genuine accelerator, not just a compliance burden. Particularly in Europe and Asia, where hyperscalers are not sovereign by default, telecom operators have a structural edge: they already operate physical infrastructure in-country, understand network-layer security, and have the capability to ensure data sovereignty at the points of storage, transit, and processing.
EDA's event-stream model is particularly well-suited here. Because data doesn't have to be centralized — it can be branched, processed regionally, and kept within specific geographic or regulatory boundaries — operators can architect sovereignty into their AI infrastructure from the ground up rather than bolting it on afterward.
What Does AI Success in Telecom Look Like in 2–3 Years?
When asked to imagine meeting again at MWC in two or three years and congratulating each other on real AI wins, the panel gave a nuanced answer. The technology will be ready — that's not the concern. The concern is whether organizations can transform fast enough to capture the value.
Francis emphasized the importance of quick wins over ocean-boiling: use cases where real-time data and AI intelligence deliver business benefits within a two-year window, because that's roughly how long the average decision-maker holds their role. Sandro was optimistic on the technology side but candid about the business transformation challenge: telecom operators will need to learn to sell more than connectivity, and restructure how their businesses operate to fully realize AI's value. Hanno, characteristically, remained an optimist — noting that EDA and other foundational investments are creating the right conditions regardless of how fast AGI arrives.
Key Questions About AI in Telecom — Answered
What is event-driven architecture (EDA) and why does it matter for AI in telecom?
EDA is a software architecture pattern where systems communicate through real-time event streams rather than batch processes or direct API calls. In telecom, networks already generate constant event streams — EDA gives AI agents the real-time data they need to make immediate decisions, while creating an auditable record of how those decisions were reached.
How should telecom operators approach AI safely?
The key principle is to separate the AI's intelligence layer (probabilistic reasoning) from its authority layer (deterministic execution and guardrails). AI agents should recommend and reason, but execution — especially for consequential network changes — should be governed by deterministic rules that humans control and can audit.
What is the role of a digital twin in telecom AI?
A digital twin is a continuously updated virtual model of the network's current state. It gives AI systems a trusted, unified view of network conditions that spans multiple OSS/BSS systems, enabling better decisions and reducing the risk of acting on stale or conflicting data.
Why is sovereign AI a competitive opportunity for telecom operators?
Telecom operators already have in-country physical infrastructure, network-layer security expertise, and experience managing regulated data. This gives them a natural advantage in delivering sovereign AI deployments — particularly in European and Asian markets where dependence on US hyperscalers raises strategic concerns.
Want to explore how Wavelo's EDA-first platform can help your organization unlock the value of real-time data for agentic AI? Talk to our team.
