Bain Warns Telecoms of 'Cost-Creep' as AI Scaling Risks Bloating Opex

Telecom operators are facing a critical inflection point in their AI adoption, risking a "cost-creep" scenario where new token expenses simply stack on top of heavy legacy operating costs. According to a report from Bain & Company, scaling AI without a fundamental shift in operating models may fail to deliver proportional productivity gains, leaving companies with higher bills and stagnant efficiency.

Bain identifies three primary "cost traps" currently hindering the industry. First, falling model prices are being offset by rising usage, leading to ballooning bills. Second, operators are frequently bolting AI onto legacy processes rather than redesigning them from the ground up. Finally, many firms are focusing on low-risk demonstrations rather than implementing changes that fundamentally transform the profit and loss statement.

The Shift to Agentic Models

To avoid these traps, Bain proposes a transition toward an "agentic operating model." In this framework, AI agents handle the bulk of execution and routine workflows, while human employees are reserved for critical decision-making. This shift could redefine the industry's financial structure, with an emerging model potentially seeing a ratio of 70-80% traditional operating expenses and 20-30% dedicated to AI agents and token costs.

Several operators have already demonstrated the potential of this approach. AT&T reduced costs by up to 90% and tripled throughput by redesigning its AI orchestration to utilize "super agents" that delegate specific tasks to smaller, specialized models. Similarly, Vivo has implemented a closed-loop network operations process where AI manages the entire detect-to-resolve workflow, escalating to human intervention only when necessary. In the retail sector, Telia piloted AI-enabled self-service kiosks to reach rural communities where traditional physical stores were not economically viable.

Why Workflow Redesign Matters

The challenge for telecom companies is rooted in their complexity. Unlike digital-native firms, telcos operate with rigid legacy business support systems, large outsourced workforces, and strict regulatory constraints. This environment makes simple AI integration difficult and often ineffective.

If operators fail to redesign end-to-end workflows and govern AI as a strategic resource, they risk increasing their operating expenses without improving their competitive position. Bain argues that the industry must move away from measuring "cost per token" and instead focus on "cost per business outcome," such as the total cost to resolve a single customer issue.

The Path Forward

As software stacks evolve toward common systems of record with intelligent agent layers on top, the traditional line between capital expenditure and operating expenditure is blurring. The industry is now split between those attempting to optimize outdated processes and those rebuilding their operations for an AI-first era.

As Bain & Company puts it, "The real divide will be between telcos that deliberately reshape their operating model and those trapped trying to optimize yesterday's processes with tomorrow's technology."