logo_header
  • Topics
  • Research & Analysis
  • Features & Opinion
  • Webinars & Podcasts
  • Videos
  • Event videos

GenAI can help the industry leap to level 4 autonomy at speed

The GenAI for AN Catalyst uses genAI to make the jump between level 3 and level 4 autonomy, helping to realize true self-healing networks that require minimal human intervention

Ryan Andrew
02 Aug 2024
GenAI can help the industry leap to level 4 autonomy at speed

GenAI can help the industry leap to level 4 autonomy at speed

Commercial context

The telecommunications industry faces dual challenges of escalating network complexity and service diversity, which together greatly intensify operational difficulties for CSPs. Increased demands on network operations and maintenance (O&M) include those stemming from various key business functions such as network security, customer response, and operational efficiency. This year alone has seen several critical network faults across the globe that have had far-reaching and disruptive consequences - the root causes of which are commonly attributed to manual errors during network changes.

These faults have severe consequences, and their shared causes underscore the urgent need for enhanced capabilities in manual processes like change monitoring and troubleshooting. There are numerous scenarios where CSPs can improve manual processes, and to meet rising customer expectations, they must enhance O&M capabilities across key touchpoints such as service provisioning and complaints management. In so doing, CSPs also hope to reduce cost and enhance efficiency, save energy and improve resource utilization.

In the last couple of years, generative AI (genAI) has emerged as the most viable way to achieve these objectives at speed. Specifically, it can help CSPs fulfil their objective to shift from level 3 autonomous networks (AN), which are automated and can perform closed-loop processes, to level 4 AN. Level 4 incorporates more sophisticated AI and machine learning algorithms to handle a broader range of tasks and scenarios, including complex and unforeseen events, and requires minimal human intervention. This evolution will be critical in developing intelligent networks capable of self-configuration, self-healing, and self-optimization.

The solution

The GenAI for AN Catalyst, which won a 2024 Catalyst Award in the Best Moonshot Catalyst – Attendees’ Choice Award category, is helping the industry use genAI to make the stride between AN3 and AN4, via an architecture for integrating genAI into network O&M which comprises three layers.

The foundational model, layer 0, provides basic capabilities, including general knowledge Q&A, mathematical calculations, code generation, and security-sensitive question rejection. Building on this, the professional model, layer 1, is tailored for communication networks. It uses a vast corpus of over 100 billion high-quality data points, including unstructured data, to create a comprehensive communication-specific language model. This model uses a wide range of machine data—faults, signalling data, and key performance indicators (KPIs)—to enhance its proficiency in handling communication network scenarios.

Layer 2, the scenario-specific application layer, integrates the layer 1 model into O&M processes, adding a software verification model to enhance service accuracy. This layer includes applications for core network alarm checks, transport network troubleshooting, wireless network planning and optimization, home broadband (HBB) installation and maintenance scheduling, and intelligent charging management.

Core network complaint analysis demonstrates the solution's capabilities, using genAI for smart analysis of core network complaints. It employs data preprocessing and parameter extraction to improve complaint classification accuracy and automate trouble ticket processing. Transport network troubleshooting trains a large model for SPN mobile bearer networks, enhancing root cause analysis and troubleshooting through 24/7 human-machine interaction. HBB installation and maintenance assistance implements natural language human-machine interaction to assist in HBB service knowledge, fault location, and service queries, significantly improving efficiency for frontline engineers.

Wider application and value

This genAI-driven architecture to support the transition to autonomous networks has shown remarkable results in various pilot projects so far, and has demonstrated the wider applicability and value of this technology. For example, China Mobile has used it to support complaint analysis for 5G ToB service scenarios in Zhejiang province, and has reduced the average complaint handling duration from 19 hours to 8. According to the CSP, this has effectively added more than 30 experienced ‘digital employees’ to their roster. Similar deployments in Hebei and Guangdong provinces led to significant reductions in fault locating and resolution times, with Guangdong seeing an equivalent addition of over 130 digital employees per city.

Telkomsel has also seen strong results, using the architecture to improve the quality of their coverage by 46% where deployed, and projecting a revenue uplift of $40 million. MTN has also witnessed very promising results, implementing the framework for intelligent charging management to streamline marketing and IT communication processes, reducing the time to market for new services from several months to weeks, thus accelerating revenue realization.

As Yuan Yao, Project Manager at China Mobile explains, “these applications highlight genAI's potential to revolutionize network O&M by transforming interaction modes, enhancing system capabilities, improving operational efficiency, and reducing time to market. This Catalyst, and resulting framework, enable a leap in capability which is set to significantly elevate O&M efficiency and user experience across the telecommunications industry.”

GenAI for AN