CSPs have traditionally relied heavily on the knowledge and expertise of engineers to solve network issues – as a result, multiple rounds of human interactions may be required to tackle a problem. This manual approach can however no longer cope with CSPs’ increasingly complex operations and maintenance (O&M) requirements. This Catalyst aims to harness data patterns and best practice to build large language models (LLMs) that will enable humans and computers to collaborate effectively on network O&M and IT service lifestyle (ITIL) processes.
The project team will focus on developing LLMs to address several specific use cases, such as summarizing work order information, predicting network faults with the support of digital twins, and recommending next best actions for O&M tasks. Other priority applications will be identifying the root causes of network faults and issues and generating operational reports based on intent. In each case, the objective is to enable engineers to simply ‘ask’ an AI agent, underpinned by an LLM, to complete necessary tasks.
The overarching goal of the Catalyst is to help CSPs greatly reduce manual repetitive tasks, thereby improving the employee experience and achieving efficiencies in fault handling and incident diagnosis. As they become less reliant on expert/high-skill support functions, CSPs should make significant time and cost savings.
The project team will track the number of complex fault scenarios that can be quickly demarcated and located. Ultimately, the progress of the Catalyst will be assessed by tracking the amount of effective knowledge deposited in the models, which will be evident in various LLM parameters. Crucially, the success of the project depends on the solution’s ability to understand human intent and the accuracy of the answers it provides.