AT&T Says Model Routing Cuts Enterprise AI Costs By 90 Percent

The Common Question

How can we reduce our AI spending without hurting productivity?

The Lesson in This Case

AT&T's chief AI officer put a concrete number on the model-routing opportunity: companies can cut AI costs by as much as 90% by matching tasks to appropriately capable models rather than defaulting every request to frontier systems. That single architectural decision – not broader usage restrictions – is what separates enterprises that scale AI economically from those that cap spending and accept slower, lower-quality output as the trade-off. Senior leaders defending AI budgets right now should stop measuring adoption by token volume and start requiring use-case-level accountability: defined outcomes, model-routing logic, and a direct line between AI activity and whether delivery times, error rates, or close cycles actually moved.

Originally reported by ERP Today. Read the full story here.