Enterprise AI Token Spending Outpaces Measurable Returns, Accenture Study Finds

Technology leaders face mounting difficulty in connecting their artificial intelligence expenditures to concrete business results, according to research from Accenture. The consulting firm surveyed 750 senior executives across enterprises and conducted 15 in-depth conversations with Fortune 500 technology and finance leaders in July 2026 to understand how organizations are managing AI costs.

The financial scale of token spending has become substantial. Organizations deployed approximately $2.5 billion toward tokens throughout 2025, with projections indicating consumption will surge by 78 percent within the coming two years. Among the drivers of AI expenses, token consumption ranks third, trailing only infrastructure costs and software development and maintenance.

Despite productivity improvements and enhanced customer interactions from AI systems, enterprises struggle to measure the return on their token investments. According to the Accenture findings, "Only one dollar in five of enterprise token spend shows up as a quantified financial outcome. The remaining four dollars sit in a space companies believe is productive but cannot prove."

Expansion Continues Despite Cost Uncertainty

Organizations are proceeding with AI initiatives despite unresolved questions about costs and measurable impact. An EY report published in July showed that more than 4 in 5 companies expressed anxiety regarding token usage and implementation expenses, yet 37 percent still intend to broaden their AI deployment footprint.

Cost reductions in token pricing may not automatically lead to lower overall spending. When Accenture asked executives how they would respond to a 25 percent or greater decline in token prices, 42 percent indicated they would increase the scale of their existing AI operations.

Recognition of the token consumption challenge has prompted industry action. The Linux Foundation established the Tokenomics Foundation earlier this year to tackle the issue of rising AI expenses tied to token usage.

Strategies for Better Cost Visibility and Control

Accenture's report outlines several approaches that technology leaders can implement to establish clearer connections between token consumption and business value. The first step involves gaining granular visibility into how tokens are consumed across different workloads.

Deploying an AI gateway or observability layer that tracks each AI interaction—including the specific model deployed, associated token costs and generated output—enables organizations to attribute 53 percent of token usage to particular teams or individual users, the report found.

Once visibility improves, assigning token cost responsibility to engineering teams and other departments encourages managers to evaluate spending decisions more rigorously. This approach pushes teams to select appropriate AI models that balance capability with expense considerations.

Making cost information transparent to developers and staff members can shift model selection behavior. Most technical personnel gravitate toward the most advanced models because they lack visibility into pricing differences, according to the research.

Routing workloads toward less expensive models represents another critical lever for managing token expenses. The Accenture analysis indicates that approximately 10 percent of workloads require advanced frontier-level reasoning capabilities, suggesting most tasks could operate on lower-cost alternatives.

Technology vendors have begun responding to enterprise concerns by introducing cost management capabilities. Snowflake introduced a model routing function within its Cortex AI Gateway that automatically selects the optimal AI model based on both performance quality and cost. Google has rolled out flexible billing arrangements that enable enterprises to establish caps on their monthly AI spending.

Before launching new AI initiatives, Accenture recommends that technology leaders establish baseline costs for current processes in time or financial terms, identify which business metrics AI will improve, and determine "how that outcome will be measured in dollars."