EY does not send every employee request directly to its most powerful AI model. The Big Four firm has deployed a stealth router that directs queries to the most appropriate model, cutting costs, according to Dan Diasio, who oversees EY's global consulting AI efforts.
Launched worldwide in April, the router works in the background of multiple EY-specific AI tools, serving as a go-between to match each task with the optimal model. "To the users, it makes no apparent change," Diasio said. "They open the frontier model, enter their prompt, and get a response. But behind the scenes, the router is deciding where their query should go."
He added, "Many times, it's faster than the frontier models, because to use the heavier models means there's more thinking that gets put in place, and it takes longer to be able to get back to the answer."
Token expenses have become a significant worry for businesses as AI vendors shift to usage-based billing. Between February and June, OpenAI, Anthropic, and GitHub launched pricing structures based on tokens - the units that measure AI input and output - rather than flat-rate plans. This change means deploying advanced models for simple requests can quickly drive up costs.
According to EY's US AI Pulse survey published Tuesday, 82% of senior decision-makers at firms adopting AI expressed concern over token expenses. The survey gathered input from 534 senior executives at American companies during April and May.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
"A lot of these big bills that companies are getting surprised by are by the top one or the top 2% of people inside the employee base that are just using the wrong tool for the job," said Diasio.
A router picks the appropriate model for each job, preventing token waste and reserving costly AI requests for tasks where they provide the most value. Diasio noted that many EY clients are developing similar solutions to manage expenses.
EY has not rolled out the router to every one of its AI tools. Instead, it's been implemented on department-specific platforms for tax and risk. The system does not cover the standard Microsoft Copilot chatbot that all EY employees can use.
The solution has produced a "meaningful impact" on how employees use AI tokens, Diasio said. Combined with training and governance measures, the router has lowered token consumption by 60% since April.
The early corporate AI boom emphasized getting workers to adopt the technology - leading to the "tokenmaxxing" trend that encouraged maximum usage. Rising costs have now pushed organizations to rein in that spending.
Companies such as Disney and JPMorgan have implemented dashboards to monitor AI usage, and AI-routing startups are gaining traction by helping developers direct tasks to appropriate models, track overspending, and resolve outages quickly.
EY assigns token budgets based on job roles and departments, Diasio said. Workers who go over their budget can ask for more tokens, which must be approved. According to the EY AI Pulse survey, 64% of senior executives reported that their companies monitor token usage and have defined budgets. Diasio expects that figure to climb rapidly: "If we ask this question in six months from now, it would probably be like 90% of people."
He also pointed out other ways companies can increase the value of their AI spending. Businesses often underestimate the need for strong internal knowledge layers and well-designed prompts grounded in their own data. Instead of spreading resources across every function, they should concentrate investment where AI can most deeply change how work gets done.
Diasio noted that while keeping an eye on employee usage stays crucial, the focus of corporate AI adoption is shifting toward value and outcomes. Measuring AI's true benefit requires looking beyond profit and loss, he argued. The bigger questions businesses should ask include: Is AI helping employees complete tasks faster? And how can you motivate workers to take on new initiatives and responsibilities that fall outside their current role?
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
