Dive Brief:
-
CIO Challenges: CIOs are struggling to predict token consumption and link spending to business outcomes, amidst rising AI costs, according to a report from Accenture. The report, based on a July 2026 survey of 750 senior executives and 15 interviews with Fortune 500 technology leaders, highlights evolving concerns.
-
Token Spending: Enterprises spent approximately $2.5 billion on tokens in 2025, with consumption projected to grow 78% in the next 24 months. Token costs rank as the third-largest driver of AI expenses related to infrastructure and software development.
-
Returns on AI Investments: While AI is generating productivity gains, establishing a clear link between token spending and financial outcomes is challenging. Only 20% of token spend demonstrates quantified results, leaving the majority uncertain in terms of productivity.
Dive Insight:
Despite concerns regarding costs and connections to business outcomes, enterprises are advancing AI implementations:
-
Growth Plans: More than 80% of businesses express concerns about AI token usage costs, yet 37% plan to augment their AI deployments, as noted in an EY report from July.
-
Cost Efficiency: As token prices become more efficient, many businesses indicated intentions to expand AI workloads, regardless of price drops. According to the Accenture report, if token prices decreased by 25%, 42% of executives would increase existing AI workloads.
-
Tokenomics Foundation: Launched earlier this year, the Tokenomics Foundation aims to address rising AI costs. CIOs are advised to enhance transparency in token usage by deploying an AI gateway to track consumption accurately. This includes monitoring models used, costs, and outcomes.
-
Workload Management: Assigning token cost responsibilities to teams can lead to better cost justification and model selection. Most developers are likely to choose costly models due to a lack of visibility regarding expenses.
-
Vendor Innovations: Companies are implementing features to manage costs effectively. Snowflake has introduced a model routing feature for optimal AI model selection based on quality and price, while Google offers flexible billing to help cap monthly AI spending.
Before adopting new AI workloads, CIOs should outline existing costs, potential financial improvements, and metrics for measuring outcomes, as suggested by Accenture’s report.
