API
< Back to all clusters
[TECHNOLOGY] · United States · 2 sources

started · updated

AI operational costs rise as enterprises shift from pilots to daily operations

As artificial intelligence moves from experimental pilots into daily business operations, companies are facing significant rising costs associated with model inference and tool integration.

For AI agents, the total cost of a completed task is often driven by more than just model tokens. Expenses include tool and API calls for tasks such as web scraping, data enrichment, and email verification, as well as infrastructure for hosting, databases, and observability. Because agents autonomously decide how many tools to use, the cost per task functions as a distribution rather than a fixed number.

In the telecommunications sector, the scale of usage is driving a shift toward domain-specific models. For example, AT&T reportedly processes over 40 billion tokens daily. Operators are increasingly wary of the “context tax” imposed by general-purpose models, which require expensive Retrieval-Augmented Generation (RAG) to process industry-specific standards. This necessity increases both token costs and latency. Consequently, the industry is pivoting from prioritizing model intelligence to seeking the lowest cost-per-answer for specialized tasks.

Entities

AT&T · GSMA · Gartner · TM Forum