Bain warns that AI investments could bloat telco opex

• AI model prices are dropping but this doesn’t guarantee cost-savings for telcos
• Bain & Co report warns that scaling AI could negate efficiency gains
• Report comes as operators embark on their AI-native journey

While the price of AI models has dropped nearly tenfold during the past year, the costs associated with agentic operating models could negate the cost savings that telcos hope to see from becoming AI native, according to a new report.

In AI in Telecom: The Opex Reckoning, Bain and Company warns that scaling up AI will simply add expenses to a legacy operating expenditure (opex) that is already heavy for telcos, with danger signs emerging in early agentic operating models.

Bain & Co said that telcos are likely to experience one of the most significant shifts in their operating cost structure for decades over the next three to five years as AI agents become embedded across customer care, network operations, software engineering and enterprise functions. Funds channelled into AI tokens are also set to account for a growing share of costs.

The global management consulting firm foresees an agentic operating model in which traditional operating expenses make up approximately 70% to 80% of the total, while the remaining 20% to 30% consists of AI agent and token costs. In that agentic operating model, humans will continue to make critical decisions, but much of the execution will be performed – or augmented – by specialised AI agents working alongside them.

The challenge for telco CTOs, CIOs, CFOs and other executives is “how can they scale up AI without simply adding token costs to an already heavy legacy operating cost base?” and make sure they see proportional gains in productivity, customer experience or growth. If they fall into the trap of a “cost-creep scenario”, as shown below, they could end up with higher opex and no tangible advantages from their AI strategies.

The concerns related to the cost-creep scenario come from three misconceptions, according to the report. 

Firstly, the falling price of AI models (which have dropped by around tenfold each year) will automatically lead to lower AI costs. 

Secondly, the idea that AI can just be bolted onto legacy processes. Telecom operators generally have large legacy footprints, but to get the most out of AI, networks require redesigned workflows. 

The final misconception is one that tracks with discussions held at TelecomTV’s recent AI-Native Telco Forum – AI demonstrations, or proofs of concept (PoCs), are mistaken for transformation.

“Telcos often focus on low-risk demonstrations that only marginally improve efficiency,” explains Bain & Co. “This sets the ambition too low and misses the P&L upside.”

The report goes on to challenge telecom leaders to make immediate changes that will also help with longer-term transformations. These include: Measuring the cost per task on individual workflows; creating a dedicated AI compute budget; picking three high-value workflows for end-to-end redesign; eliminating wasteful AI usage; and establishing governance before proliferation.

Hannes Schneider, Alex Bhak, Danielle Stekelenburg, Alex Martynov and Fabio Caiazzo, who co-authored the report, concluded: “The goal isn’t simply a 70:30 cost ratio. It’s an operating model that’s faster, simpler and higher performing because it’s AI native.”

- James Pearce, Editor, TelecomTV