GenAI implementation still in its preliminary stages; challenges include data quality and regulation

Amdocs commissioned Analysys Mason to carry out the research* titled, “GenAI Challenges and Opportunities.” The results from the survey showed that there is a clear understanding among 90% of CSPs of the profound impact GenAI will have on business goals. However, deployment of GenAI is still in its preliminary stages, with just 22% of CSPs having already implemented GenAI solutions, 32% running proof of concepts (PoCs) and 29% looking to explore it in the next 12 months.

Findings from the research include:

CSPs expect significant changes with GenAI: Most respondents expect GenAI to have a high impact across a variety of domains, such as software development (84%), data management (82%), monetization (82%) and network operations (81%).

Specific benefits are expected: Generating new enterprise revenue (51%) and reducing time to market (40%) were leading use cases, followed by improved employee productivity (39%), new consumer revenue opportunities (35%) and improved customer experience (34%).

Automation is an area of critical importance: 75% of survey respondents expect that integrating GenAI with operational systems will expand the range of automated actions and tasks that operational systems can perform.

However, GenAI implementation is a struggle: 84% of CSP Data, AI and IT organizations say they face great challenges in acquiring the capabilities they need to deploy GenAI successfully. 81% consider GenAI integration with business processes as the largest challenge they face.

Most CSPs consider data quality, LLM (large language model) training and regulatory compliance as being the most challenging hurdles to overcome: Almost half (45%) of CSPs said alignment with compliance, privacy and security regulations were the biggest challenge, followed by training LLMs with telco specific data (43%), and access to high quality data (40%).

The path forward: Almost three-quarters (74%) of respondents are exploring what it takes to build their own GenAI platform. Further, 42% are using professional services providers to implement specific telecom use cases. Additionally, 84% of CSPs are expecting to deploy GenAI as an embedded feature via a vendor’s B/OSS applications.

“Some CSPs are even implementing or considering building their own large language models (LLMs). However, building these platforms internally can be slow and costly, especially when skills and resources are scarce, at this point in the market. These CSPs will have to undertake the training and fine tuning of these models for the telecom environment on their own and this can present challenges if they do not have the required skillset in-house,” said Adaora Okeleke, Principal Analyst, Analysys Mason. “Adopting a specialized telco-focused LLM platform from a single vendor that consolidates capabilities from multiple LLM providers can accelerate adoption. Once Generative AI becomes an integral part of core business processes, it has the potential to bring additional value and can transform CSPs’ performance, speed, and operating costs.”

“GenAI relies on a vast array of web data, but it lacks access to the data associated with the telecom technology ecosystem, creating a crucial gap and lack of context. To address this challenge, expertise is required to unlock access to pertinent telco data sources, comprehend the intricate telecom taxonomy, and draw necessary, correct conclusions,” said Gil Rosen, Chief Marketing Officer at Amdocs. “Through Amdocs’ amAIz TelcoGPT platform, CSPs can seamlessly incorporate GenAI to enhance network efficiency, elevate customer service, and drive overall business improvement. Additionally, they can effectively manage costs, security, scalability, and efficiency while optimizing use of tokens.”