Only 12% of telcos saw balance-sheet gains from AI, McKinsey says
The survey also showed widespread AI pilots, but limited scaling.
Most telcos have yet to translate artificial intelligence (AI) investments into significant productivity gains, with only around 12% of surveyed operators reporting meaningful balance-sheet impact, according to a McKinsey & Co. report.
The survey, conducted in December 2025 with 49 telecom executives across global regions, found that whilst operators are piloting AI widely, scaling use cases remains limited.
McKinsey & Co. notes that agentic AI’s ability to automate tasks, coordinate work across functions, and influence operational decisions could reshape workflows and value creation—but only if organisations restructure around AI rather than embedding it into legacy processes.
Executives expect the greatest near-term cost reductions from AI in customer support, network operations, and IT, with potential savings of roughly 10% over the next one to two years, and up to 30% by 2030.
Challenges remain, as immature operating models, data limitations, and ineffective change management were cited by more than three-quarters of respondents as barriers to scaling AI value. Around half of respondents identified employee adoption as the primary hurdle.
The report highlights early adopters deploying agentic AI across customer service, network planning, and internal support functions.
For example, one telco’s contact centre uses AI to analyse all inbound calls daily, identifying leads and performance insights, which reportedly increased inbound sales by 40% in 10 weeks.
McKinsey & Co. outlines elements for a coherent agentic operating model, including linking AI initiatives to budgets and financial goals, and redesigning workflows around task-level activities.
It also cited clarifying roles for human workers and AI agents, embedding change management, and establishing reusable AI platforms.
The firm warns that operators who implement these changes can materially improve return on invested capital and margins, whereas those treating AI as an incremental tool risk only marginal productivity gains.
