ENTERPRISE use of artificial intelligence is creating a new cost challenge as rising token consumption makes AI spending harder for companies to predict and measure, according to an Accenture Research report.
The report, The CIO’s Guide to AI Tokenomics, said enterprises surveyed spent about $2.5 billion on AI tokens last year. Without intervention, the amount could rise to about $3.6 billion within 24 months as token consumption increases.
Accenture surveyed 750 senior executives from companies with annual revenues of more than $1 billion across 17 countries in July 2026 and conducted 15 interviews with technology and finance executives.
The report said companies expect token consumption to increase 78 percent over the next 24 months, while token prices are expected to decline 19 percent. Without sufficient optimization, the cost increase could reach about 44 percent.
“It’s a real problem ... growing by a minute, by a day ... keeping track of [token volume and spend] is definitely a challenge, especially if it’s distributed across multiple organizations, multiple teams,” a technology executive interviewed for the report said.
Accenture said companies currently optimize about 23 percent of token consumption. It estimates that roughly 31 percent optimization would be needed to keep spending flat as usage grows.
The report identified five disciplines for managing AI costs: workload-level visibility, financial accountability, model routing, proof of value and AI skills.
Only about one in five dollars spent on AI tokens can currently be linked to a quantified financial outcome. Organizations using chargeback and workload-level measurement could increase that to about 32 cents per dollar.
“The economic question is not ... how many [tokens] were consumed, but what did that token actually do?” a director of AI strategy at a Fortune 500 financial services institution said.
Accenture also found that one in three organizations had exhausted their annual token budget before the end of the year. Token consumption is spread across employee and developer tools, internal process automation, enterprise software, customer applications, analytics and agentic workflows.
“Even if the models become cheaper, teams just use them more,” an executive interviewed for the report said, describing the effect of falling token prices on AI consumption.
The report said model selection is another major cost factor. Accenture’s analysis of 9,368 occupational tasks found that fewer than 10 percent genuinely require frontier AI capabilities, while the remainder can use less expensive models.
Frontier models can cost 10 to 20 times more per token than capable mid-tier alternatives, while 54 percent of AI requests are being routed to a model tier higher than the task requires.
Accenture recommended making AI workloads observable before they scale, assigning financial responsibility for consumption, routing tasks to appropriate models, requiring measurable business cases before production deployment and developing AI skills across the organization.
“Your Toyota Camry will do just fine. You don’t need to upgrade to a Ferrari anytime soon,” an executive quoted in the report said, illustrating its recommendation to match model capability with the requirements of each task.
The report concluded that AI cost management requires CIOs to work with finance, business and human resources leaders to establish visibility, accountability and measurable value before AI consumption expands further.
