- Fine-tuning offers telcos a cheaper alternative to building GenAI models from scratch
- Yet many telcos are still building costly models despite the availability of ready-made solutions
- Challenges and costs aside, telcos like Windstream expect long-term benefits from their GenAI investments
Since ChatGPT essentially sparked the big bang of generative AI (GenAI) in 2022, telecom operators have rushed to implement GenAI models across their operations. But the costs of training and running these models can be significant, and the return on investment (ROI) of GenAI remains a topic of debate across sectors.
Fine-tuning GenAI models has emerged as a more affordable option for telcos. Fine-tuning in the context of GenAI models refers to the process of taking a pre-trained model and further training it on a specific, smaller dataset to optimize its performance for a particular task or domain.
Ishwar Parulkar, chief technologist for Telecom and Edge Cloud at AWS, outlined the cost breakdown for different stages of GenAI model development, noting that pre-training – building a model from scratch – can cost millions. This task is typically handled by specialized providers like Anthropic and Cohere.
On the other hand, fine-tuning existing models with proprietary data is a less expensive alternative, costing only thousands.
Windstream is one example of a telco opting for fine-tuning. According to Stephen Farkouh, Windstream’s chief information officer, the company relies on Azure OpenAI and GPT-4O, avoiding the high costs of building models from the ground up. Instead, Windstream subscribes to Azure’s AI services, with costs dependent on usage and the “complexity and volume of data we’re tweaking,” Farkouh said. This approach saves the company “a significant amount of money and resources.”
New research from SAS shows that 70% of telcos are already using GenAI and 89% plan to invest in it within the next financial year.
Despite this widespread adoption, some telecom operators are making avoidable mistakes.
A McKinsey & Company report pointed out that many telcos are still building GenAI solutions from scratch, even though there are numerous off-the-shelf options. Only one-third of telco leaders surveyed said they buy products off the shelf, which, according to the report, suggests “that many telcos continue to embrace a do-it-yourself model. This move is likely to slow innovation and distract talent from more differentiating use cases, as it has in the past with other technologies.”
Is GenAI worth the price tag?
GenAI brings inevitable costs, no matter how it’s adopted. Even fine-tuning is an iterative process, requiring repeated evaluations through LLMOps/FMOps procedures, which contributes “significantly to the overall cost structure,” Parulkar noted.
AI adoption at scale also comes with other unique challenges for telcos, especially when dealing with complex networks, vast data volumes, sensitive data and real-time processing needs.
Managing massive, heterogeneous data from sources like network logs, customer interactions and IoT devices is not only complicated but also costly due to the high storage and computing power required. Real-time decision-making, which depends on low-latency data handling and model inference, often calls for large instance types or GPUs. Additionally, telcos must navigate stringent security measures to protect sensitive customer data, further adding to the complexity and cost of these projects.
Finding the right specialized expertise within its enterprise isn’t cheap, either.
“Telecom and AI expertise is competitive and can be costly,” Farkouh said.
The payoff for telcos
Despite these hurdles, companies like Windstream remain confident that their GenAI investments will yield long-term benefits. Farkouh said the company is “banking on several long-term benefits” from its AI investments, among them customer experience and operational efficiency. “Automating routine tasks and predicting maintenance needs will help us save costs and improve service reliability,” he added.
And there are already promising results. According to McKinsey & Company, one European telco increased its marketing campaign conversion rates by 40% while reducing costs by using GenAI to personalize content. Another telco in Latin America boosted call center agent productivity by 25%. What’s more, both companies deployed their GenAI models in a matter of weeks – the first in just two weeks, the second in five.
“For an industry with a mixed track record for capitalizing on new technologies and legacy systems that slow innovation, these early results and deployment times illustrate the potentially transformative power of GenAI,” the McKinsey analysts wrote.