A new study from Boston Consulting Group (BCG) reveals that only a handful of companies are turning artificial intelligence into measurable business value. The study, "The Widening AI Value Gap," finds that just 5% of firms are realizing AI’s full potential, and most are still struggling to move beyond pilot programs.
Across industries, roughly 70% of AI’s impact now comes from core operations such as maintenance, supply chain, and pricing.
Airlines and telecommunications lead the way, with AI contributing to about 80% of value in core functions, up 15 percentage points and eight percentage points, respectively, from 2024. Software, telecom, and payments and fintech show the strongest maturity gains overall.
BCG calls these companies “future-built,” describing them as organizations that treat AI as a strategic capability rather than an experiment.
A major shift is also underway with agentic AI, which can function as autonomous, tool-using “digital workers” that can perform tasks and make limited decisions under human supervision. Agents represent about 17% of total AI value today, and BCG expects that to rise to 29% by 2028. Nearly half of surveyed companies are already piloting or deploying agentic systems.
For network operators, these trends are already visible in predictive maintenance, network operations center (NOC) triage, field-tech assistance, and customer-experience automation. AI tools are helping identify outages before they occur, triage alarms across complex infrastructure, and provide technicians with guided workflows and parts availability in real time.
The report also shows that technology alone won’t close the gap. 72% of organizations report unmanaged AI-security risks, often tied to poor data quality and unclear oversight. Companies that realize the most value tend to share certain habits: clear ownership between IT and business leaders, strong data governance, and training programs that help employees work alongside AI systems.
The telecom sector’s data-rich networks and high service expectations make it a prime candidate for end-to-end AI. But as BCG’s findings show, realizing that potential will depend less on cutting-edge models and more on disciplined execution.
