As generative AI and AI agents are rapidly being adopted across the telecommunications industry, AI is increasingly viewed as an important tool for improving customer service, network operations, and overall business efficiency. However, Bain & Company’s latest analysis points out that declining AI model costs do not necessarily translate into lower overall AI spending for telecom operators. Without simultaneously redesigning existing workflows, large-scale AI adoption could instead drive up telecom operators’ operating expenses (OPEX).

According to Bain, the cost of using AI models has declined significantly over the past year, with the cost of some models falling by nearly tenfold. However, as telecom operators deploy AI agents across customer service, network operations, software development, and corporate management, the scale of AI usage and token consumption is also expected to increase. This could offset some of the anticipated cost savings. Over the next three to five years, as AI agents become widely integrated into telecom operations, AI-related expenses could account for approximately 20% to 30% of telecom operators’ OPEX, potentially becoming a new structural cost component.

Bain argues that the key challenge facing telecom operators is not a lack of AI technology, but the tendency to simply layer AI onto existing processes. Without simultaneously transforming existing operational processes and systems, improvements in the efficiency of individual tasks may not translate into lower overall operating costs. In addition, many telecom operators remain at the proof-of-concept (PoC) stage, where individual AI applications can demonstrate technical benefits but have yet to generate productivity gains significant enough to be reflected in corporate financial performance.

Bain therefore recommends that telecom operators move beyond simply expanding AI adoption and instead redesign their core operating models around AI. This includes establishing mechanisms to manage AI computing and usage costs, conducting end-to-end redesign of high-value workflows, and eliminating AI applications that fail to deliver tangible benefits. The key is not to “deploy more AI,” but to ensure that the productivity gains generated by AI exceed the additional costs associated with computing and operations.

This trend indicates that competition in AI within the telecommunications industry is gradually shifting from “AI adoption” to “AI-driven operating model transformation.” Going forward, telecom operators’ ability to effectively integrate AI agents, network automation, and existing systems—and translate AI investments into revenue growth and measurable productivity gains—will be critical to realizing the business value of the AI-native telco.