As CSPs continue to evolve their networks towards 5G Standalone and cloud-native architectures, managing network functions from multiple vendors remains a significant challenge.
Whether dealing with traditional VNFs or newer CNFs, CSPs often face many of the same underlying problems:
- Different vendor architectures and deployment models: Each supplier has its own approach to packaging, configuration, interfaces, lifecycle management and upgrades. Interoperability and integration complexity: Getting network functions from different vendors to work together consistently across the core network is rarely straightforward.
- Documentation and tooling: Different data models, APIs, configuration formats and operational procedures create additional integration effort.
- Dependency management: Understanding infrastructure requirements, software dependencies, network connectivity and interactions with other functions can be time-consuming.
- Testing and validation: Each new function, release or configuration change can require extensive integration, performance, security and acceptance testing.
- Operational fragmentation: Monitoring, fault management, scaling, upgrades and troubleshooting often involve multiple vendor-specific tools and processes.
- Skills and cost: Specialist engineering effort is needed to onboard, integrate and maintain each vendor’s solution, making repeatable automation difficult.
Cloud-native network functions bring greater flexibility, scalability and the potential for more automated operations, but they don’t automatically remove these challenges.
And physical appliances should not be overlooked. They still offer benefits in certain network scenarios, including dedicated hardware, predictable performance and purpose-built capabilities. However, they bring their own set of challenges:
- Physical deployment and site readiness: Hardware delivery, rack space, power, cooling, cabling and site access all need to be coordinated.
- Commissioning and configuration: Each appliance needs to be installed, connected, configured and integrated into the wider network.
- Hardware and software lifecycle management: Firmware updates, patches, replacements, capacity upgrades and hardware refresh cycles add operational complexity.
- Multi-vendor interoperability: Appliances from different suppliers still need to work together and integrate with existing OSS, BSS and network management systems.
- Distributed operations: Managing and maintaining hardware across multiple sites can be resource-intensive and difficult to standardise.
- Limited flexibility: Scaling or changing capacity can require additional hardware, physical intervention and longer lead times.
So, whether the network function is delivered as a VNF, CNF or physical appliance, the underlying challenge remains: how can CSPs onboard, integrate and manage multivendor network functions in a consistent, repeatable and largely automated way?
Taskgentic.ai believes, Agentic AI, particularly multi-agent systems, could make a real difference here.
Imagine a vendor-independent intelligence and automation platform sitting between network function suppliers and the CSP’s existing cloud, infrastructure and orchestration environments. Taskgentic.ai view this as having four closely connected layers, supported by a common trust, security and governance framework.
1. Multi-agent intelligence layer
At the heart of the solution would be a team of Taskgentic specialised agents, each responsible for a different part of the onboarding process.
A package analysis agent examines vendor CNF/VNF packages, appliance specifications and documentation, identifying deployment requirements, interfaces and configuration options.
Validation agents could check security, standards compliance and compatibility with the target environment.
Architecture and dependency agents could determine what the new function needs from the existing network, while integration agents identify the interfaces, data models and adaptations needed to connect it to the wider 5G Core.
Deployment and testing agents could prepare deployment configurations, integration plans and acceptance tests.
An orchestrating intelligence agent would bring their findings together, resolve conflicting recommendations and ensure the proposed approach is technically consistent.
2. Agentic orchestration layer
This is where the different pieces come together. Rather than simply executing a predefined workflow, an Tasgentic agentic orchestration layer manages the overall onboarding journey.
It would understand the intended outcome, decide which agents and capabilities need to be involved, determine the sequence of activities and coordinate dependencies across vendors, cloud infrastructure and network domains.
It dynamically adapts the process when something unexpected happens. For example, if a compatibility check fails, it requests the relevant agents to investigate the issue, identify possible solutions and revise the integration plan. If a required interface or dependency is missing, it flags the gap and determine what needs to be resolved before proceeding.
It would also manage approvals, hand-offs and progress tracking, maintaining a clear view of the onboarding process from initial vendor submission through to operational acceptance.
3. Execution and automation layer
The third layer turns the agreed decisions and plans into actual network changes.
This layer connects to existing technologies and platforms such as Kubernetes, GitOps, CI/CD pipelines, cloud infrastructure managers, NFV management systems and network orchestration platforms. It handles tasks such as resource allocation, configuration generation, deployment, service integration, testing, scaling and lifecycle management.
For physical appliances, it also coordinates installation workflows, commissioning, configuration and integration with network management systems.
The agents would not need to replace these established platforms. They would instruct and orchestrate them through standardised APIs and vendor-specific adapters, with policy checks, security controls and approval gates governing what can be executed.
4. Trust, security and governance layer
For Taskgentic.ai, this is one of the most important parts of the architecture. CSPs will not put agentic systems in control of network onboarding and operations unless they can trust how decisions are made and how actions are executed.
This layer needs to provide identity and access management for agents, role-based permissions, secure handling of vendor packages and credentials, and protection against malicious or manipulated input. It ensures that agents can only access the systems and perform the actions they are authorised to.
Just as importantly, every recommendation and action needs to be traceable. Therefore, clear audit trails showing what information was used, why a decision was made, which agent proposed it, what approvals were obtained and what changes were executed.
Policies define which activities can be automated, which require human approval and which must never be performed autonomously. Independent validation, security checks, testing and rollback mechanisms are built into the process rather than added afterwards.
In other words, trust cannot be an assumption. It has to be engineered into the platform.
Together, these layers would create a continuous onboarding process: analyse, validate, plan, orchestrate, execute and verify. Lessons from each onboarding exercise is captured and reused, reducing the effort required to integrate the next vendor’s network function.
Of course, there are still important challenges for Taskgentic to solve. How do we ensure agents make technically correct decisions? How do we deal with proprietary vendor interfaces and inconsistent documentation? How do we prevent incorrect configurations from reaching production? How do we manage interactions between multiple agents? And how do CSPs retain control, accountability and trust?
This requires more than adding AI agents to existing automation. It means combining telecom-specific intelligence with standardised network models, deterministic validation, policy-based execution, human oversight and independent verification. AI can help interpret, reason and orchestrate, but production changes must remain controlled, secure and auditable.
Taskgentic believes, the biggest opportunity is not another AI-powered CNF deployment tool. It’s creating a vendor-independent, agentic integration and orchestration platform that significantly reduces the time, cost and engineering effort involved in onboarding and managing network functions across different vendors and deployment models.
And the potential extends beyond 5G Core to RAN, transport, edge and eventually 6G.
Could this be the change automation requires?
About the Author
Chief AI Officer at Taskgentic.ai. Leading the development of enterprise agent architectures that combine collective intelligence, modular workflows and rigorous security guardrails for complex operational environments.
About Taskgentic.ai
Taskgentic.ai is developing the next generation of AI-driven task intelligence, helping Communications Service Providers move from isolated AI agents to secure, reusable and governed ensembles and assemblies that deliver measurable business outcomes.
