AI token costs push executives to rethink rollout plans, EY says
An EY survey of 500 U.S. senior decision-makers finds AI token costs are forcing companies to revisit budgets and deployment scope.
By Colin Brandt · Enterprise Reporter
· 3 min read
AI token costs are now changing enterprise rollout plans, according to an EY report based on a survey of 500 U.S. decision-makers at senior vice president level and above. EY said more than four in five businesses investing in AI have internal concerns about token usage and implementation costs, although more companies are widening their AI plans than cutting them back.
The numbers point to a budget fight rather than a retreat from AI. EY found 37% of businesses are expanding the scope of planned AI deployments despite cost concerns, while 15% have moved toward narrower implementations. The report did not disclose total AI budgets, token volumes, vendor mix or how much spending has changed by company size.
Dan Diasio, EY global AI consulting leader, said in the report that companies are starting to prioritize which AI projects deserve funding, rather than treating adoption itself as the goal. He also said organizations need to link AI initiatives to financial value or risk spending effort without making progress.
Why are AI token costs changing company plans?
Tokens are the units many AI model providers use to meter usage, so costs can rise as employees, applications and automated workflows send more prompts and process more outputs. For enterprise teams, that means a promising pilot can become materially more expensive once it is embedded across customer support, software development, analytics or back-office operations.
EY’s survey suggests the cost issue is pushing companies back into the build-versus-buy decision. Three-quarters of senior leaders with active AI investments said packaged software products do not match their IT requirements, according to EY. Nearly nine in 10 businesses have either fully deployed or are testing programs to build AI internally.
That shift is relevant for vendors selling AI into large accounts. If customers believe off-the-shelf tools do not fit their systems, vendors may face longer implementation cycles, tougher procurement reviews and pressure to prove cost controls. EY’s findings also leave open a key question: whether in-house AI reduces total cost or moves it into engineering, infrastructure and governance budgets.
Vendors are responding to cost pressure
AI providers and cloud companies are already trying to meet the cost objection. OpenAI said earlier this week it reduced pricing for its Luna and Terra models. Oracle and AWS announced product changes last month that were aimed at helping customers manage AI costs more effectively, according to CIO Dive.
The demand for spending controls is also creating room for management tooling and operating practices around AI usage. Flexera data released in June found that more than two-thirds of companies lack accurate visibility into AI software use, while roughly three in five reported a year-over-year increase in AI overspending.
FinOps, originally used to control cloud spending, is being extended to cover AI costs. The Tokenomics Foundation, an offshoot of the Linux Foundation announced last month, is intended to help companies manage AI spending across the enterprise.
For operators, the signal is straightforward: AI budgets are moving from experimentation into financial scrutiny. The companies expanding their deployments are still doing so under tighter measurement, and the winners among vendors will need to show not only model quality, but predictable unit economics at production scale.
This story draws on original reporting from CIO Dive.
