When new technology ignites as much hype and expectation as generative artificial intelligence (gen AI), there’s bound to be a fair amount of uncertainty in the early days over what use cases the technology can be applied to, what problems it will solve, and how it will make money.
Communications service providers (CSPs) are still investing in cloud, automation, and traditional AI to reach autonomous networking, so it’s fair to ask: How does gen AI fit into my current automation strategy?
The good news is that the telecom industry is on time to answer this question. The immediate move is to combine the capabilities of traditional AI and gen AI by integrating current AI/ML-driven insights into large language models (LLMs) designed and fine-tuned specifically for the telecom industry.
Data and intelligence are not new concepts to CSPs or the telco industry at large. For over 50 years, service providers have worked with basic data to write programs that could carry out specific actions.
Over the last 15 years, this data has evolved into telemetry data used to model insights into actions. Today, the rise of gen AI and LLMs signals the dawn of the “Knowledge Era,” where insights are extracted from a unified knowledge source; this knowledge is then used to automate and problem-solve across all operational domains.
It is in this Knowledge Era that CSPs will begin to move the needle closer toward the ultimate vision: reaching Level 5 Autonomous Networks, as defined by TM Forum’s Autonomous Networks Maturity Model. At this level, CSPs can run intent-driven autonomous operations that deliver zero-wait, zero-touch and zero-trouble services that meet the changing needs of their customers.
CSPs can move beyond the gen AI hype and take that first step by uniting the capabilities of traditional AI and gen AI. This move requires a pragmatic approach, and CSPs must look for partners that have deep domain knowledge to help bridge the AI gap by taking some critical steps. Such as:
- Curating a vast collection of federated knowledge sources: Tap into a wealth of unparalleled domain knowledge spanning network planning, design, operation, and customer care to enhance productivity gains and future product enhancements.
- Developing a sophisticated LLMOps system to optimize knowledge sources: Leverage state-of-the-art LLMs designed to maximize the benefits CSPs will get from knowledge sources. Evaluate the use of leading LLMs from the open-source community that are powerful, task-relevant, and cost-effective to deliver customized results for CSPs and enterprises.
- Implementing domain-driven hallucination management: Employ prompt engineering and fine-tuning techniques to telco-specific LLMs for consistency and accuracy. This will mitigate the risks of low-quality outputs by implementing domain-driven hallucination management. Likewise, by prioritizing upskilling existing talent on data vectorization tools, fine-tuning techniques, and prompt engineering skills and by providing training and resources to strengthen their understanding of these tools and techniques, they will be able identify and manage hallucinations across the network effectively.
Nonetheless, CSPs will have concerns and questions. For starters, many operators are in the middle of deploying 5G and other cloud-based transformation projects and want to know, should I wait for genA I to evolve and mature?
The answer is no; you do not have to wait. Innovative partners are delivering measurable impacts today with CSPs willing to make the gen AI leap.
Imagine a leading Tier 1 CSP that wants to achieve significant cost and efficiency gains in network and service operations. This CSP would send out multiple requests for information (RFIs) and then choose a partner that could guide CSP personnel along the traditional AI/gen AI blueprint outlined above. It would then go on to achieve actual, measurable KPIs, including an 80% reduction in knowledge acquisition time, a 72% increase in data analysis efficiency, and annual savings of over $6.5 million USD in network operations.
This success would be leveraged to create a forward-looking blueprint for large-scale automation using gen AI, further improving high-impact areas like automatic generation of network design configurations and optimization recommendations. It all starts with combining today’s traditional AI/ML insights into telco trained gen AI LLMs.
The symbiotic relationship between gen AI and traditional AI is key to realizing the full potential of AI-driven autonomous networks. Strong CSP partners will know why and how today’s AI is driving tomorrow’s gen AI potential.
By adopting the steps above, CSPs and enterprises can expedite the adoption of gen AI in networks and move closer to the fully autonomous network vision.
