Accelerate your journey to Autonomous Networks with Agentic AI powered by Amdocs, AWS and NVIDIA
Simplifying complexity for customers and network operations
Telecom operators are in the middle of a perfect storm, one that combines stagnant revenue growth, escalating operational costs, and increasing customer demands. This challenging environment is further complicated by the rapid evolution of technology and the need to modernize legacy systems. Their lighthouse is the Autonomous Network (AN) vision, aiming to provide zero-wait, zero-touch, and zero-trouble customer experiences. This vision relies on intelligent infrastructure and agile operations supporting self-service, self-fulfillment, and self-assurance of telecom networks.
The expected outcomes are transformative: enhanced customer experience with higher availability of network services, quicker fulfillment of orders, and easier description of orders as "intents" rather than detailed specifications. This shift from complex, technical specifications to intent-based ordering represents a significant leap in user-friendliness and efficiency, particularly for B2B customers.
However, achieving this vision is complex. The TM Forum defines a six-step maturity model for the AN journey, with most telecom operators currently at level 2 (Partial AN), striving to reach level 4 (Highly AN) in the next 3-5 years. This journey involves overcoming numerous challenges, including integration of legacy systems with new technologies, development of AI and machine learning (ML) capabilities, ensuring network security and reliability in an increasingly automated environment, and upskilling the workforce to manage and maintain autonomous systems.
The urgency for this transformation has never been greater. Operators need to reinvent their operations and processes to tap into the age of AI.
As we stand at this critical juncture, the integration of Agentic AI, digital twins, and intent-based operations offers a promising path forward. By leveraging these technologies, Operators can address their immediate challenges while paving the way for future innovations, improved customer experience and operational efficiency.
The secret sauce: Data, AI, and Telco Know-How
While the telecom sector has embraced generative AI at pace, many deployments fall short of true operational integration. In a recent whitepaper, Amdocs, AWS, and NVIDIA define a telco-grade Agent as one that is “verticalized”—deeply embedded with a robust data ingestion pipeline, telco-specific skills, ontologies, reasoning capabilities, agent communication, and simulation using Digital Twins. These agents go beyond surface-level interactions. They deliver network-aware, context-sensitive responses and execute intelligent decision-making rooted in domain expertise.
In this blog we will showcase how the unique collaboration between Amdocs, AWS, and NVIDIA accelerates the AN journey for telecom Operators and how Agentic AI and Digital Twins will revolutionize how Operators sell, monetize, and operate their services.
The critical role of Digital Twins
A Digital Twin is a virtual representation of a real-world entity or system, aiming to mirror a physical entity, process, organization, or other abstraction accurately. Its goal is to help make more reliable decisions by simulating the impact of potential changes to a product or service without risking the entity itself.
For telecom operators, different types of Digital Twins can be created depending on the use case. For example, a virtual replica of their network can be used to simulate routing scenarios and support reliable decision making. It does so by deriving insights from comprehensive network data and enabling simulations, leading to the trusted execution of any action.
Data, analytics, AI/ML, and visualization are key foundational elements of a Digital Twin.
Data powers Digital Twins
For digital twins to be effective, a comprehensive data layer is essential. This layer ingests and stores data from multiple sources, including equipment, topology, configuration and assurance data to build an accurate representation of the network to support operations use cases. Establishing this data layer relies on capabilities from each of the partners.
Amdocs provides foundational data sources such as the Amdocs Network Inventory, which holds comprehensive service and network data, and the Amdocs Service Assurance Suite, which provides real-time insights into fault, performance, and service quality data.
AWS provides the scalable infrastructure required to efficiently ingest and store vast amounts of network data. Amazon Kinesis data stream is a scalable AWS streaming platform that can capture gigabytes per second of data with AWS Glue serving as the cornerstone for data integration services, enabling automated ETL processes and metadata discovery across diverse data sources. Amazon Neptune, a fully managed graph database service, forms the crucial knowledge representation layer, supporting both property graph and RDF models, making it ideal for storing complex relationships and knowledge structures that AI agents need to navigate. The time-series capabilities are handled through InfluxDB on AWS, providing high-performance temporal data management essential for agent decision-making and pattern recognition.
This foundation is strengthened by AWS's comprehensive security framework, including VPC for network isolation, AWS IAM for access control, and AWS KMS for encryption management, ensuring that AI agents operate within secure boundaries. The integration of these core services creates a robust platform that enables organizations to build sophisticated AI agents capable of processing vast amounts of data, understanding complex relationships, and making time-aware decisions while maintaining enterprise-grade security and scalability.
NVIDIA enables GPU-acceleration with NVIDIA RAPIDS and Morpheus– part of the NVIDIA AI Enterprise software platform — for data analytics pipelines, speeding up data processing and near real time processing
The result is a comprehensive data layer that provides a virtual representation of an Operator’s network, tailored to support operations use cases in a highly efficient manner.
Analytics and AI/ML transform data into knowledge
Analytics and AI/ML transform this data layer into knowledge by deriving actionable insights that accelerate the journey towards autonomous networks. AI/ML models trained on the network data identify patterns and predict changes, leading to efficient anomaly detection and fault prediction.
Amdocs leverages its telco know-how and provides patented AI/ML algorithms from its Service Assurance Suite, supporting root cause analysis and anomaly detection.
AWS provides a comprehensive set of services for data-driven digital twin solutions, leveraging cutting-edge AI/ML. With services like Amazon SageMaker, Operators can build, train, and deploy ML models at scale.
NVIDIA CUDA-X™ Libraries, part of the NVIDIA AI Enterprise software, offer a set of AI , ML capabilities deployed on accelerated compute to speed up performance and optimize compute resources; examples include: cuML, cuGraph, Morpheus and more.
The outcome is an optimized analytics and AI/ML layer that provides valuable insights by analyzing historical and real-time data, transforming Operators into data and AI-driven organizations.
Reaching autonomy with agents
In the journey towards AN, Agentic AI plays a crucial role. Coupled with Digital Twins, which provide AI/ML-derived insights and support simulations, AI agents autonomously make decisions and act accordingly to achieve specific goals, such as delivering and maintaining business intents.
These use-case oriented agents efficiently analyze, simulate, recommend and implement network changes, leveraging Digital Twins to enhance operational efficiency and customer experience.
The agents are built using Amdocs amAIz Suite which brings deep telco know-how, providing a telco-specific generative AI platform that optimizes token usage and improves efficiency.
AWS’ infrastructure for AI agents centers around Amazon Bedrock, which provides foundation models and APIs for creating sophisticated AI agents that can understand, reason, and generate responses. These agents can be enhanced through knowledge integration using a RAG (Retrieval Augmented Generation) pattern, where Amazon Neptune serves as the graph database backbone for storing complex relationships and knowledge structures.
AI agents can access this knowledge infrastructure to provide context-aware responses and make informed decisions. The agent architecture typically involves AWS Lambda for serverless compute and Amazon SageMaker for custom ML model deployment when needed. The integration of different models through Bedrock allows these agents to perform sophisticated natural language understanding and generation tasks, while maintaining context through the knowledge graph.
Security and governance are enforced through AWS IAM, and agent interactions can be event-driven using Amazon EventBridge. This architecture enables organizations to build intelligent agents that can handle complex queries, maintain conversation context, and leverage enterprise knowledge while scaling efficiently and maintaining security compliance.
AWS enhances enterprise AI governance through an integrated set of services, anchored by Amazon Bedrock's advanced capabilities including multi-agent orchestration and automated reasoning systems. The framework enables organizations to implement risk-calibrated controls that maintain compliance without sacrificing operational agility. For telecom applications, this translates into practical AI deployment through various approaches: from straightforward Retrieval-Augmented Generation (RAG) with contextual boundaries, to hybrid systems combining RAG-based knowledge access with defined compliance checkpoints. Organizations can also systematically convert operational procedures into AI-compatible formats using structured workflows and formal process definitions. This modular approach allows telecom operators to scale AI adoption based on their specific requirements and risk tolerance levels.
NVIDIA AI Enterprise deployed on AWS further enhance the Agents allowing them to be skilled with Telco capabilities and deployed more efficiently and securely:
NVIDIA NeMo is used for customizing generative AI models while NVIDIA NIM™ microservices ensure fast, optimized and secure model deployment, resulting the generation of Large Telco Models (LTM) NIMs. LTMs NIMs are the base of Agentic AI in Telco Network allowing the agents to be portable with optimized performance.
Figure 1. Amdocs, AWS and NVIDIA Blueprint Architecture
Putting it into practice
To demonstrate the transformational impact of Agentic AI and digital twins, Amdocs, AWS and NVIDIA implemented two high impact use cases for customer experience and operational efficiency.
Agentic AI for Self-Service
In the self-service use case, a Sales Agent interacts with the customer to simplify the ordering process by capturing the needs in simple terms and the business intent. The Sales Agent then identifies the optimal solution by collaborating with a Network Operations Agent that has visibility into the network resources.
The ordering process starts with the Sales Agent leveraging the Amdocs Customer Engagement Platform and the Amdocs Catalog to identify potential solutions for the customer’s requirements. In the background, it engages with a Network Operations Agent that autonomously performs technical validation using a Digital Twin which aggregates network data from multiple sources including Amdocs Network Inventory. The Network Operations Agent evaluates all viable options, including those based on planned network rollouts and partner networks.
Once the Network Operations Agent identifies the feasible options, it communicates them back to the Sales Agent. This interaction showcases the power of multi-agent deployments, where agents collaborate to autonomously complete complex tasks, such as the ordering process.
The entire end-to-end process, including the interaction between the agents, is transparent to the customer, resulting in a seamless and optimized ordering experience that enhances customer satisfaction.
Agentic AI for Autonomous Network Operations
In the second use case, we demonstrate the transition from human-driven to system-driven network operations, where a Network Operations Agent continuously monitors the network health and reacts in real-time to AI/ML-based predictions for service-impacting faults. These predictions are generated by models trained on historical raw data including fault, performance, and telemetry.
Upon identifying a pattern that may lead to a service disruption, the Network Operations Agent autonomously initiates a diagnostic process. It analyzes the network health data sourced from the Amdocs Service Assurance Suite and performs a service impact analysis using network and service topologies from Amdocs Network Inventory.
Next, the Network Operations Agent evaluates potential resolutions, prioritizing them based on effectiveness and impact on the broader network. Low impact resolutions are implemented autonomously by the Network Operations Agent, while those with broader impact require human approval.
Leveraging the Digital Twin, the Network Operations Agent simulates recommended resolutions before applying them, increasing trust in the proposed solutions and minimizing risk.
This use case showcases the combined value of AI/ML, Agentic AI, and Digital twins to proactively predict and prevent service-impacting issues by simulating what-if scenarios before implementation.
Achieving benefits along the way
In conclusion, the integration of Agentic AI and Digital Twins by Amdocs, AWS, and NVIDIA demonstrates to Operators a transformative path towards autonomous networks, enhancing customer experience, and operational efficiency.
A practical approach must be taken, gradually addressing incremental use cases that deliver value along the way. Operators should build the data layer incrementally, incorporating only the essential data to support their prioritized business and operations goals(e.g., zero-touch service fulfillment or preventive maintenance).
In alignment with the data layer, analytics, AI/ML models, and agents should be specifically developed and trained to support identified use cases. As trust is built on their insights and actions, their level of autonomy can be increased.
As this journey evolves, the Agentic AI and Digital Twins will become integral parts of Operators’ operations, leading to benefits that will impact every aspect of operations. For instance:
- Reduced Meant-time-to-Repair (MTTR)
- Increased automated incident resolution
- Enhanced service quality
By embracing this approach, Operators can navigate the complexities of modernization and achieve significant improvements throughout their journey.
