Why Network Function Lifecycle Management Is Still the Hardest Problem for Telcos

Telcos have been evolving through distinct phases of network function deployment: hardware appliances, ETSI-based Virtualized Network Functions (VNFs), Cloud-Native Network Functions (CNFs), and GitOps automation. Each phase introduced architectural improvements while perpetuating a fundamental challenge: separate tools for lifecycle management. Onboarding, Day-2 operations, and multi-vendor integrations remain inconsistent across VNF and CNF environments, creating operational friction and constraining automation velocity.

A recent approach has emerged that can address these challenges across diverse deployments. Model Context Protocol (MCP) provides a standardized interface for AI agents to interact with different systems and applications. For network lifecycle management, this means AI assistants can potentially work across various network functions through a common protocol, reducing custom integration overhead. Mission-critical network operations will continue to require human oversight.

This blog examines the lifecycle management evolution, identifies persistent gaps in current approaches, and explores how MCP-enabled AI agents might address fragmentation in mixed VNF/CNF environments while acknowledging both the historical challenges and operational realities of telco networks. 

From Appliances to GitOps

Early network virtualization focused on decoupling software from hardware. ETSI NFV provided a common architectural framework that introduced flexibility in how network functions could be deployed and managed. This enabled faster service rollouts with VNFs compared to physical appliances. However, while ETSI standardized the LCM interfaces, practical implementation remained vendor-specific with preference for specific VNFM rather than leveraging generic VNFM. Onboarding, upgrades, and scaling relied on bespoke workflows and scripts.

Cloud-native network functions became the preferred delivery method for 5G deployments. The Technology disrupted the network virtualization standardization efforts as well as the development cycles of the network functions themselves. Not all network functions were able to follow a micro-service architecture to support the IT/ enterprise way of lifecycling the infrastructure, creating more operations challenges on the longer run and emphasized the need to invest in more automation capabilities. At some point, the industry accepted the need to follow IT/enterprise best practices and is now being disrupted by GitOps. 

The evolution of GitOps, which uses declarative models and treats configuration as code, is improving deployment consistency. This is facilitated by tools like ArgoCD, a Kubernetes-native GitOps controller that synchronizes configurations between Git repositories and their intended clusters.

GitOps delivers benefits within its defined scope: managing Kubernetes resources through declarative manifests stored in version control. This approach provides traceability and enables rollback to known states. However, VNFs, infrastructure components, and platform services often require parallel automation tooling or integration layers. Service providers operate both VNFs and CNFs on separate cloud instances, with lifecycle management remaining separated between these function types.

The Core Challenge: Lifecycle Fragmentation

In spite of standard APIs, every Network Equipment Provider (NEP) is implementing the Lifecycle Management (LCM) in different ways, which requires vendor-specific validation cycles. Scaling operations use different APIs and tooling depending on workload type.

Traditional API standardization addresses some aspects but introduces their own challenges: lengthy standardization cycles, vendor implementation variability, and innovation constraints. 

This fragmentation directly impacts three operational dimensions:

• Automation velocity – Custom integrations slow onboarding cycles. Teams spend time managing tooling complexity and vendor-specific scripts rather than improving network capabilities.
• Operational risk – Inconsistent lifecycle patterns increase error probability during critical operations. 
• Workforce expertise – Teams must master VNF orchestration, Kubernetes operations, GitOps workflows, and multiple vendor management interfaces simultaneously. The constant evolution of technology necessitates continuous “re-skilling,” making this expertise gap a structural talent issue that persists alongside mixed environments.

MCP and AI-Driven Lifecycle Management

Model Context Protocol establishes a standardized interface between large language models and software systems. While MCP represents new terminology, intent-based and AI-assisted lifecycle management are not new concepts. The industry has explored these approaches for a few years now, with specifications developed in 3GPP and multiple vendor implementations attempted.

Earlier efforts faced substantial challenges. Translating high-level intent into concrete operations across diverse equipment proved far more complex than expected. Telco networks spanning 2G/3G/4G/5G generations cannot achieve clean-slate modernization. LCM with application vendors remains heavily human-driven, requiring tribal knowledge and domain expertise remain heavily human-driven, requiring tribal knowledge and domain expertise that AI systems have yet to replicate reliably.

Modern large language models show promise in contextual understanding and multi-step reasoning that earlier rule-based systems lacked. MCP provides a standard protocol layer while leaving vendor APIs intact. These developments are worth exploring and we have begun research into how AI might augment specific lifecycle workflows like troubleshooting analysis or configuration validation.

However, mission-critical networks cannot accept experimental approaches in production. 

MCP-Assisted Operations with Human in the Loop

We’re investigating MCP-based approaches as potential augmentation tools and not replacements. The vision of expressing intent in natural language while AI orchestrates complex workflows remains aspirational. Reaching that state requires solving persistent challenges: achieving meaningful vendor ecosystem participation, establishing semantic standardization across implementations, building trust for mission-critical infrastructure decisions, and transforming operational skillsets.

For now, operators should focus on maturing their CNF adoption and GitOps practices. Current CNFM and GitOps methodologies provide proven, reliable lifecycle management today. These approaches deliver tangible operational improvements today while providing the standardized foundation that future AI assistance might build upon. MCP represents an interesting research direction, but substantial work remains before it becomes operationally viable for telco lifecycle management.

What This Means for Service Providers

Model Context Protocol and AI-assisted automation represent exploration of a new approach to lifecycle fragmentation. By providing a standardized interface for AI agents to interact with operations across heterogeneous environments, MCP represents a promising step toward a unified LCM, though its effectiveness at scale remains unproven. It’s an emerging pattern with potential, constrained by real-world complexity.

Practical starting points include:

• Low-risk operations – Start with read-only operations (inventory queries, configuration backups, diagnostics) before extending to production-affecting changes.
• Workflow assistance – Use AI agents for procedure documentation and troubleshooting recommendations where errors have limited impact.
• Mandatory human oversight: Implement HITL approval for all operations affecting service availability.

Open Questions

Will vendor ecosystem participation reach critical mass? Can semantic standardization achieve consistency across implementations? Will operational teams trust AI recommendations for mission-critical infrastructure? These answers will emerge from implementation experience, not architectural diagrams.

Whether these capabilities overcome historical complexity is not yet clear. For service providers with mixed VNF/CNF environments and legacy constraints, MCP-based AI assistance offers a direction worth investigating with realistic expectations. Progress should be measured in reduced integration effort and improved operator productivity for specific workflows, not in wholesale transformation or autonomous operations.

Additional Resources

Explore more about our approach and solutions:

  – TelecomTV on VMware: Read the latest insights on our telco cloud strategy.

  – VMware Telco Cloud Platform Page: Discover the full features and benefits of the platform for your network.