NEC and AWS Show Agentic AI Running UPF End-to-End - From Design to Self-Healing
NEC, working with AWS, has demonstrated agentic AI that manages the full lifecycle of the User Plane Function (UPF) in 5G/6G core networks - from design and build to deployment, monitoring, anomaly detection, and self-healing. The result: work that used to take weeks with manual effort can be compressed into hours.
For operations leaders, this points to fewer tickets, faster rollouts, and less reliance on niche expertise for day-to-day tasks. It also sets a clear direction for how AI and cloud-native ops will scale across edge and on-prem footprints.
What was demonstrated
NEC validated autonomous orchestration of UPF using "Kiro," an AWS agentic AI with an IDE and CLI. The AI generated and executed deployment procedures, monitored live conditions, detected anomalies, and triggered self-healing - all without human intervention.
UPF is the data plane of the 5G Core (5GC), handling user traffic between devices and data networks. For context, see 3GPP 23.501.
Why it matters for operations
Manual builds and config drift slow launches, inflate costs, and increase outage risk. Embedding proven carrier-grade practices into AI closes that gap - turning tribal knowledge into repeatable, auditable procedures that scale.
This approach also broadens who can safely operate core functions. Teams without deep 5G specialization can manage deployments and recoveries through standardized, AI-driven workflows.
Key features
- Autonomous UPF deployment and operation: AI generates and runs procedures informed by NEC's carrier operations expertise, then monitors and self-heals in production conditions.
- Cloud-native flexibility on AWS: Apply consistent operational models across regions, edges, and on-prem systems using familiar cloud tooling and patterns.
Operational benefits
- Weeks to hours: Time from network design to initial service launch drops dramatically compared to NEC's traditional manual methods.
- Fewer errors, higher quality: Automation reduces manual touch points and the risk of misconfigurations.
- Wider participation: Personnel without deep 5G core expertise can deploy and run network functions safely.
- Faster revenue capture: CSPs can launch services sooner and respond to demand spikes more effectively.
How it works at a high level
The AI encodes deployment and operations knowledge - including sequencing, validation, and rollback logic - then executes via IDE/CLI, aligned with cloud-native patterns. Continuous monitoring flags anomalies and initiates recovery without waiting for human action.
Because it runs on AWS, the same operational model extends across core regions, edge sites, and on-prem environments for consistent control and visibility.
Voices from the field
Amir Rao, Global Director, GTM and Telco Solutions, AWS: "AI-driven automation is becoming essential to build a competitive advantage by serving end customers faster and reducing operating costs. NEC's autonomous UPF orchestration shows how AWS AI services can help CSPs deploy complex 5G infrastructure in hours rather than weeks, so teams can focus on service differentiation instead of manual tasks."
Takashi Sato, Corporate Senior Vice President and Managing Director, Network Solutions Business Division, NEC: "Validating UPF deployment and operations automation using AI and GitOps in an AWS environment is a key milestone. By embedding NEC's design and operational expertise into AI, we aim for autonomous operations and significant end-to-end efficiency gains, including on-site activities."
What operations leaders can do next
- Pilot with clear guardrails: Start with a non-critical UPF slice or lab environment; define SLOs, rollback criteria, and audit trails up front.
- Standardize the pipeline: Move artifacts and configs into GitOps/IaC to give the AI a single source of truth.
- Instrument for trust: Build observability that maps AI actions to outcomes; require explainability for any automated change.
- Upskill the team: Blend 5G core fundamentals with AI-driven ops and cloud-native patterns. See the AI Learning Path for Network Engineers.
Where to see it and additional notes
A live demonstration based on the verified environment will be at the AWS booth during MWC Barcelona 2026 (March 2-5, 2026).
A portion of the technologies used in this project were supported by a grant program (No. JPJ012368G50701) from Japan's National Institute of Information and Communications Technology (NICT).
About NEC
The NEC Group applies AI, security, and communications technologies to create social value and a more sustainable world. Founded in 1899, NEC has around 110,000 employees worldwide and deep experience in commercial carrier networks and mission-critical systems.
UPF note: User Plane Function processes user data traffic in the 5G Core Network (5GC).
Your membership also unlocks:
