Accenture has released the results of "The CIO’s Guide to AI Tokenomics," a survey regarding AI token usage and costs among companies in 17 countries.
Accenture Report: Potential to Reduce AI Token Cost Increases by Up to 44%
This article is a translation. Read the Japanese original
According to the survey, an average of 25% of the AI-related budgets of the surveyed companies are allocated to token consumption, and one in three companies exceeds its budget regarding annual token expenditures. Furthermore, management predicts that token consumption will increase by 78% over the next two years.
On the other hand, less than 20% of companies are able to demonstrate a connection between token expenditures and quantitative financial outcomes. Accenture's analysis estimates that tasks requiring the capabilities of frontier models (cutting-edge, high-performance AI models) account for approximately 10% of the total, yet it was found that 54% of actual AI requests are being assigned to high-performance models that exceed the required standard of the task.
To address this cost inflation, Accenture states that by thoroughly implementing optimizations such as "model routing"—a mechanism that distributes requests to the appropriate AI model depending on the difficulty of the task—it may be possible to reduce the expected increase in token costs by up to 44% over the next two years.
Sources
- 「AIトークンコスト増を44％抑制」できる可能性も 高性能モデルの“使い過ぎ”、どう減らす？ （ITmedia AI+、2026-09-26）
- PR TIMES