Key Takeaways of MWC 2026

After several days of conversations at MWC 2026 with CTOs, network leaders, partners, and fellow vendors across the ecosystem, these are some of my personal takeaways from the event.

This is not meant to be a definitive industry analysis, but rather a synthesis of recurring themes that consistently appeared in discussions around network automation, AI, and Autonomous Networks.

New technologies must land in a specific network domain and a clear E2E use case

At MWC 2026 there is a noticeable shift in how operators evaluate new technologies. Concepts like AI, Digital Twin, or Agents are no longer compelling on their own.

Operators increasingly expect every innovation to be anchored in a concrete network domain and demonstrated through a clearly defined end-to-end use case.

The typical questions are straightforward:

– In which network domain does this apply? (RAN, IP Core, Access, Transport, Telco Cloud)

– What operational problem does it solve?

– What does the E2E flow look like from detection to action?

There is also visible skepticism around the idea of a “full network digital twin.” What starts to resonate are domain-specific implementations tied to real operational scenarios.

Another important point is the timeline. Operators are not expecting these capabilities to transform operations in a few months. Most initiatives are being framed as multi-year operational transformation programs rather than short-term deployments.

In practice, technologies gain credibility only when they are grounded in a network domain and a concrete operational workflow.

The conversation has moved from “automation” to closed-loop operations

Across the industry, the focus is shifting toward closed-loop operational models.

Operators are increasingly interested in architectures that combine:

• observability and telemetry
• event correlation
• policy engines
• automated remediation
• human-in-the-loop governance

Reference architecture such as CLADRA are starting to appear more frequently in discussions, emphasizing structured approaches to closed-loop automation and operational decision-making.

Another clear trend is the growing need for cross-solution collaboration. Closed-loop operations often require combining expertise and capabilities from multiple layers of the stack, bringing together vendors and specialists responsible for different domains such as assurance, observability, automation, orchestration, and AI.

In practice, meaningful closed loops are rarely delivered by a single solution. They require cross-solution integration that brings the expert of each layer into the operational workflow.

The key idea is not automation in isolation, but automation that can detect, decide, and act within a governed operational framework across multiple systems and domains.

Multi-vendor and multi-domain complexity remains the core challenge

Despite years of transformation initiatives, the operational reality of telecom networks remains highly fragmented.

Most networks are inherently:

• multi-vendor
• multi-domain
• multi-generation

Operators must orchestrate environments that include legacy infrastructure, cloud platforms, multiple OSS/BSS stacks, and equipment from different vendors. In addition, the industry is starting to move toward multi-agent operational models, which add another layer of complexity.

Between these domains, platforms, and emerging agent-based systems, there are often gaps in orchestration, data consistency, and operational ownership. These gaps accumulate over time and translate into significant technical debt, making it harder to scale automation across the network.

This structural complexity continues to be the main barrier to operational simplification and large-scale automation.

ANL4 remains the strategic north, but the operational reality is still far from ANL3

Autonomous Networks Level 4 continues to serve as the long-term direction for the industry. However, the operational maturity across network domains remains uneven.

In many operators:

• some domains are approaching ANL3
• many still operate around ANL2
• and legacy environments may even resemble ANL1 practices

The immediate challenge is therefore not jumping directly to ANL4, but raising the baseline maturity across domains. In practical terms, operators are focusing on:

• consistent automation across domains
• reliable topology and inventory
• closed-loop automation for concrete operational cases
• reduction of manual operational work

GenAI is having stronger impact in build time than in run time

Another consistent observation is that Generative AI is currently delivering more value in build time than in run time.

GenAI is accelerating activities such as:

• Automation development
• Development of operational playbooks and procedures
• Faster integration between systems

In contrast, operators remain cautious about introducing AI directly into runtime decision-making in production environments.

At runtime, AI is currently being used more often for analysis and assistance, for example:

• configuration analysis
• troubleshooting support
• operational diagnostics

As a result, a practical pattern is emerging: GenAI for build time acceleration + governed automation for run time execution. AI helps teams build operational capabilities faster, while the execution layer remains deterministic, auditable, and controlled.