Unlocking Long-Term Operational Efficiency with Generative AI in the Telecom Industry

As the telecom industry continues to navigate the complexities of a rapidly evolving technological landscape, senior executives are under increasing pressure to drive down Operational Expenditure (OpEx) while maintaining competitiveness. In 2024, the adoption of Generative AI (Gen AI) marked a significant turning point, with widespread implementation across industries and a focus on practical applications. Now, as we look to the future, it’s clear that Gen AI will play a crucial role in shaping the telecom sector’s operational efficiency.

The telecom industry’s operational expenditures to sales ratio stands at a relatively high 73%, with a expenditures split across Network OpEx costs (30%), Non-Network costs (49%), and other costs (21%). Reducing OpEx is a pressing concern for most telecom providers, and leveraging Gen AI is emerging as a key strategy in achieving this goal.

The Transformative Impact of Gen AI on Telecom Operations

Last year’s integration of Gen AI in telecom has been characterized by the strict concern around data security and governance, the development of specialized AI models, and a growing emphasis on explainability and transparency. Retrieval Augmented Generation (RAG) applications were particularly effective in driving productivity gains and faster onboarding times. As we look to continued adoption and integration this year, the focus is shifting towards Gen AI Agents and Agentic Workflows, with our telecom customers seeking to automate Network Operations, enhance anomaly detection and response, and improve customer experiences.

Gen AI has the potential to transform telecom operations in several key areas:

Network Operations Automation

Gen AI-powered agents can automate complex network configurations, anomaly detection, and response, reducing the need for manual intervention and minimizing the risk of human error.

Use Case

Our client’s mobility network operations were heavily reliant on manual configuration changes across a wide variety of Cisco and Juniper devices. These manual processes increased the risk of errors, contributed to longer maintenance windows, and made coordination efforts across multiple teams difficult.

Evergreen’s team of experts coordinated with the client’s teams and designed, built, and deployed a tool that enabled their teams to automate configuration changes and updates to their Mobility Network. This directly led to reduced risk, better team collaboration, and shorter downtimes for maintenance resulting in a 54% savings in operational costs.

Predictive Analytics

By leveraging Gen AI-driven predictive analytics, telecom providers can better anticipate and prepare for future network demands, reducing the financial uncertainty associated with network maintenance and upgrades.

Use Case

One of our F100 clients was facing challenges with their legacy maintenance and fully life-cycle asset management & procurement processes. Evergreen performed a full review of their processes and identified potential for large operational gains by using modern predictive analytics automated in an agentic workflow.

We designed and built this system that highlighted when assets should likely be ordered based on project pipelines, notified when to expect an assets shortage during work in progress and suggested resolutions, and updated project completion based on changing logistic concerns. This tool revolutionized the client’s processes and resulted in a 42% operational increase by reducing shortages and work stoppages as well as providing better estimates of when projects would be finished, and revenue would be recognized.

Customer Experience Enhancement

Gen AI can enable personalized and immersive customer interactions, driving revenue growth, and improving customer satisfaction and retention.

Use Case

Our client was seeking a partner to own and operate 24/7 technical support for network services that deliver critical telecommunications for a state. Their goal was to find a strategic partner with a focus on cost savings and continuous improvement.

Evergreen took over their call center operations and after a few months established a baseline for the client’s KPIs. We built a continuous improvement roadmap with targeted Agentic AI based ticket handling. This work resulted in massive operational efficiency gains like reducing ASA from <290 seconds to <10 seconds (96.6%) and reducing AHT from <15min to <5min (66.7%) while improving graded ticket quality from ~70% to >95%.

Converting OpEx to CapEx: The Role of Gen AI

One of the key benefits of Gen AI is its potential to help telecom operators convert OpEx to CapEx, thereby reducing operational expenditure and improving the bottom line. Several methods can be employed to achieve this:

• Capitalizing Software Development Costs: By capitalizing Gen AI software development and implementation costs as CapEx, telecom providers can reduce their OpEx burden.
• Build vs. Buy: Building in-house private cloud or Gen AI infrastructure and equipment instead of buying it from third-party vendors can convert OpEx to CapEx.
• Gen AI Tools for Productivity: Developing in-house Gen AI tools can improve employee productivity and efficiency, reducing the need for additional headcount and associated OpEx costs.

Building Agentic Workflows: The Future of Telecom Operations

The development of Agentic Workflows is a critical aspect of Gen AI adoption in telecom. Several frameworks and tools are available to support the creation of these workflows, including LangGraph, MS AutoGen, CrewAI, and Model Context Protocol (MCP). These frameworks enable telecom providers to design and deploy multiple AI agents that can work together to handle complex tasks, interact with multiple stakeholders, and drive business outcomes.

Our experience in building Gen AI-powered solutions for telecom clients has yielded impressive results. Examples include:

• Robotic Process Automation (RPA) Agents: Automating responses to device anomalies and risks based on device metrics and log data.
• Network Configuration Agents: Assisting Network Operation Center users in setting complex network configurations in response to changing network events and statuses.
• RAG and Agent-based QA Development Agents: Restructuring unclear requirements into standardized formats, generating strong test case coverage for those requirements, and generating test scripts that are ready to be ran.
• Generative Business Intelligence (BI) Infobot Agents: Ask conversational questions of your data and get answers automatically without needing to know how to write complex queries yourself. These tools convert complex questions into multiple queries, run them, and transform the results into natural language answers.

Strengthening Telecom Through Gen AI

The strategic adoption of Gen AI throughout the telecom industry is poised to drive significant long-term OpEx reduction. By leveraging Gen AI-powered tools to build and run Gen AI Agents and Agentic Workflows, telecom providers can automate network operations, enhance customer experiences, and improve predictive analytics.

As both the industry and Gen AI continue to evolve, it’s clear that Gen AI will play a critical role in shaping the future of telecom operations. We believe that by embracing Gen AI and Agentic Workflows, telecom providers can unlock new levels of operational efficiency, drive revenue growth, and maintain competitiveness in an increasingly complex and dynamic market.

Justin Groseclose is Technical Architect, AI/ML for Insight Global’s professional services division, Evergreen. As a Data Scientist with more than 10 years of experience interpreting and analyzing data to drive improved business outcomes, he focuses on advanced numerical mathematical methods, statistics, data analytics, and machine learning. Groseclose is a graduate of Armstrong Atlantic State University with a Bachelor of Science degree in Applied Mathematical Sciences and has completed graduate studies at Freie Universität Berlin in Numerical Methods and Analysis, Mathematical Aspects of Machine Learning.