Talk to your network: Igniting the AI revolution in autonomous networks

TTYN shifts the telecom services lifecycle from current service realizations to unified, proactive services and operations, providing CSPs a competitive edge. This operational upgrade will turn networks into intelligent partners capable of understanding goals, predicting needs, and solving problems autonomously.

By enabling seamless communication through natural language inputs and intent-driven automation, TTYN eliminates manual tasks, streamlines service lifecycles, and unlocks a future of smarter, robust, and user-centric connectivity. This is the future of telecom—networks that evolve beyond infrastructure to become innovative collaborators in a digital-first world.

Talk to your network - Motivation

The telecommunications industry is undergoing an exciting transformation, unlocking a world of new opportunities for CSPs to provide new services, applications, and businesses. As CSPs embrace this evolution, they are poised to deliver innovative services worldwide, such as cutting-edge VR experiences and novel applications that create fresh business opportunities. 

On the other hand, such evolution will also pose significant challenges for CSPs in managing their complex, resource-intensive networks. These challenges include escalating network complexity, demanding customer experience expectations, and rising operational costs. Future telecom services and network management approaches will require integrated solutions, agile troubleshooting processes, and a proactive approach to service issues. Such approaches will improve efficiency, reduce operational expenses, and increase customer satisfaction.

Due to an ever increased number of applications expected in 6G, telecom systems must support continuous evolution of CSPs services and business models. Faster service creation and integration necessitate autonomous, dynamic, and flexible systems capable of handling new services and intent defined by the operator at run-time. Future telecom systems will enable continuous development, allowing coding, verification, and deployment of new software for implementing new services at run-time without disrupting existing services. Service performance will be closely monitored  and assured at every stage of the service lifecycle.

Table 1: Service lifecycle management requirement

Interactions and potential conflicts  will be proactively managed with resilience to ensure outcomes that align with service goals. 

To develop new services and handle the added complexity, AI becomes an essential ally, enabling the differentiated connectivity needed to provide on-demand, enhanced services that meet and exceed customer expectations. By leveraging AI-driven solutions, CSPs can proactively manage their networks, streamline operations, and enhance efficiency and customer satisfaction. With AI at the forefront, CSPs are well-equipped to lead the charge in this dynamic industry landscape, embracing a future full of promise and potential.

In summary, future telecom systems will require an AI-driven framework that supports the service operation, integration, creation, and assurance in a non-disruptive manner at run time. This is evident in all aspects of service lifecycle management as shown in Table 1.

Talk to your network – A paradigm shift 

The vision of the telecommunications landscape of the future is clear: design systems that empower humans to create and efficiently manage complex, interconnected telecom networks, services, and applications. These systems must meet operational demands and support a wide range of business purposes and use cases as depicted in Figure 1.

Figure 1. Talk to your network vision

This vision paves the way for new applications and services beyond our imagination. It calls for advanced automation solutions beyond AI and machine learning (AI/ML) techniques, which are already integrated into products.

Talk to your network (TTYN) is an umbrella term that brings together key AI technologies and enablers to establish autonomous networks capable of addressing the aforementioned critical areas. At the heart of TTYN are numerous autonomous network agents that operate seamlessly together as if they were a single unified system. These agents work together to analyze the network state of the telecom system, identify anomalies and their root causes, search for solutions, test different solution combinations, and make decisions about actions to be executed. In essence, these agents conduct a “sensing, searching, thinking, and acting” operation cycle in an orchestrated manner,  based on learned knowledge. A description of the four tasks of the operation cycle is as follows:

• Sense: Analyze and assess the network’s status, for instance, detecting throughput degradation in specific sites.
• Search: Retrieve critical information from different sources of inputs, such as product configurations in the network, or find relevant customer product information (CPI) and summarize the key details.
• Think: Diagnose issues, perform reasoning, and make decisions on remedial actions for any identified issues. 
• Act: Implement  the corrective action, like, adjusting configuration parameters.

While these tasks can initially be guided by human intervention, the goal is intent-based autonomy, where systems operate independently without human involvement. The resulting orchestrated behavior allows the system to dynamically adapt and self-organize at run-time to fulfill high-level business and service intents. The emergent intelligence of such a system ensures flexibility and dynamic behavior. The TTYN autonomous workflow is fundamentally different from traditional chatbot interactions in conversational AI. Unlike chatbots, which primarily handle short-term, single-step interactions, TTYN workflows manage multi-step, long-running tasks that require complex and flexible planning and execution.

The TTYN framework integrates knowledge and agent-driven methods, transforming natural language into actionable intelligence. TTYN represents a paradigm shift in telecom products, services, and network management, as it leverages three AI technology pillars of autonomous networks, namely, —Generative AI, Agentic AI, and intent-based AI.

• Conversational AI, based on Generative AI and large language models (LLMs), enables natural language interactions, providing explainable insights and fostering transparent communication between networks and human operators, as well as between machines, to draw abstract conclusions.
• Agentic AI empowers autonomous entities to sense, think, and act independently, functioning as a cohesive system capable of dynamic adaptation.
• Intent-based AI orchestrates these entities, aligning their actions with overarching business goals for seamless coordination and execution.

At the core of TTYN is a translator module, which leverages Agentic AI and LLMs to break down high-level tasks or intents into smaller, actionable subtasks. These subtasks are designed for sequential or parallel execution, ensuring that the overarching task is completed effectively.

The aforementioned  key enabler technologies along with existing AI/ML solutions provide the framework foundations for TTYN use cases. The next section will detail how the technologies within the TTYN framework will contribute to the lifecycle management of future telecom systems and services.  

The execution of these plans depends on the degree of interaction between humans and AI agents, enabling seamless collaboration. The spectrum of autonomy within the TTYN system is adjusted based on the product's readiness and performance requirements of the task. This flexibility ensures that tasks are handled reliably and efficiently, maintaining an optimal balance between automation and human involvement. 

Benefits of TTYN

TTYN addresses the growing opportunities for new businesses, and the challenges faced by  CSPs  including increasing network complexity, heightened customer experience demands, and rising operational costs. A TTYN solution revolutionizes how CSPs control their infrastructure, turning complex networks into responsive, intelligent systems that operators can directly question and command while addressing critical pain points as outlined in Table 2.

Table 2: Benefits of TTYN

TTYN - Key technology pillars 

Key AI functions

To realize the above vision and meet the demands, TTYN introduces a holistic solution based on three key AI pillars: Conversational AI, Agentic AI, and intent-based AI. This framework is complemented by supporting functions for data and knowledge processing, AI/MLOps, and reliable features to ensure security, transparency, and scalability as shown in Figure 2.

Figure 2. Pillars of autonomous networks

Conversational AI and large language models (LLMs)

Conversational AI focuses on creating systems that primarily interact with humans through natural language and multi-modal inputs, enabling fluid and context-aware communication. An extension of Conversational AI will allow for agent-to-agent communication. These systems leverage advanced technologies, including natural language processing (NLP), Generative AI, LLMs, machine learning, prompt engineering, retrieval-based techniques, and dialogue management to determine context and intent scope from prompts. Additionally, powered by Gen AI, they can generate a rough plan of tasks, which acts as a blueprint for orchestrating intents and the tasks to achieve them. This plan forms the foundation for collaboration with intent-based AI and Agentic AI, bridging human interaction with advanced system capabilities. To this end, CSPs can communicate with the network using plain English, receiving clear, explainable insights into system operations, decisions, and actions. This reduces the cognitive load on human operators and fosters transparency while simplifying the management of complex telecom networks. 

Agentic AI

Agentic AI involves autonomous agents designed to execute tasks or achieve intents within dynamic environments. These agents, typically guided by reinforcement learning, planning algorithms, and adaptive decision-making frameworks, perform the actions needed to fulfill system-level objectives. They operate as part of an orchestrated system, collaborating to achieve user-defined intents or adapt to changes in real time. Different autonomous agents may perform different tasks including network state analysis (sensing), data analysis and retrieval (searching), root cause and conflict resolution analysis (thinking), and autonomous task execution (acting). With the assistance of Conversational AI, LLMs will enable dynamic agent-to-agent communication for real-time understanding of data by parsing unstructured and structured inputs and translating them into actionable insights. Furthermore, LLMs could serve as intermediaries, translating between different machine languages, APIs, or protocols to ensure seamless integration and communication between agents with task dependencies from disparate systems. Agents form a cohesive system that dynamically adapts to fulfill high-level business and service intents. 

Intent-based AI

Intent-based AI emphasizes recognizing and decomposing intents into actionable intents or sub-intents. It combines techniques like intent recognition, context analysis, and prediction to break down complex goals into manageable tasks. These sub-intents are further mapped to system functions or autonomous agents, ensuring precise alignment with service intents and user needs. In tandem with Conversational AI, it refines the rough plan into a set of structured intents, enabling seamless orchestration. To this end, intent-based AI provides the coordination layer, ensuring that autonomous agents act in alignment with overarching service intents and business goals. By translating high-level intents into actionable steps, it orchestrates seamless collaboration and ensures efficient service delivery and network management.

Together, the three key AI technologies contribute to an integrated system capable of conversation, orchestration, intent decomposition, and autonomous execution. Table 3 provides a more detailed description of how AI technologies contribute to service lifecycle management.

Table 3: Mapping AI technologies to service LCM

Supporting AI functions

To realize the full potential of the TTYN framework, the solution includes essential supporting functions:

• Data and knowledge life cycle management: Efficiently handle data ingestion, storage, processing, and lifecycle operations (DataOps), along with knowledge creation, relevance determination, and updates (Knowledge PS). 
• AI/MLOps: Manage the lifecycle of machine learning models and pipelines, including LLM-driven architectures, retrieval-augmented generation (RAG), and Agencies for dynamic adaptability.
• Trustworthy AI (TWAI): Ensure transparency, reliability, and ethical use of AI systems by incorporating safeguards against unintended outputs, data leaks, and hallucinations. Trustworthy AI includes robust guardrails, secure response modulation, and safe template outputs to maintain compliance with ethical and security standards. As part of TWAI, data privacy and security are paramount. Therefore, providers are taking significant measures to ensure that the data used to train LLM models adheres to internal privacy protocols and safeguards against data leakage.

Extensive reusability is required in addition to the above functions to streamline the evolution of TTYN use cases. TTYN solutions should build on common assets like Gen AI libraries, AI/non-AI agent templates, AIOps, and other existing AI/ML functions for consistent and scalable solutions.

Application categories and use cases
The TTYN framework can support a range of use cases, each designed to improve network operations, service creation, and system integration. These use cases progress from human-in-the-loop stages for input and feedback to human-on-the-loop oversight, eventually leading to Zero-touch operations with minimal human intervention. A short description of use case example categories is listed in Table 4.

Table 4: TTYN's possible use-case examples

In practice, multiple use cases from these categories can operate simultaneously within a network, covering a wide range of products and solutions from different vendors. By leveraging these capabilities, TTYN redefines network operations, creating intelligent, self-organizing, and trustworthy systems tailored to the demands of modern telecommunications. 

Conclusion

TTYN represents a paradigm shift in telecom network operations by leveraging advanced technologies like Generative AI and agent-driven orchestration.  It aims to transform networks into autonomous systems capable of self-organization, service delivery optimization, and rapid adaptation to dynamic market demands. TTYN achieves this by enabling seamless communication between CSPs and network nodes through natural language inputs, eliminating the need for manual command processing. This automation streamlines operations across various service lifecycle stages, from service creation and integration to assurance and operation. By enabling intent-driven automation, TTYN translates business objectives into actionable tasks executed efficiently by agents, reducing human intervention and fostering greater operational efficiency.  TTYN represents more than just a technological shift—it’s a reimagining of network services and operations as a proactive, purpose-driven system, designed to adapt to the evolving demands of future networks. 

Ericsson is at the forefront of this evolution, working closely with industry partners to define and develop the building blocks of autonomous networks. Together, we are paving the way for an era of transformative possibilities where smarter, trustworthy, and secure networks offer seamless connectivity to users in the digital age.