EY Built an ‘AI Router’ to Keep Its Own AI Bills Under Control
The Big Four firm is steering tasks to cheaper models to control token costs, part of a wider corporate reckoning with the price of AI.
The Problem
July 30, 2026 – 1:24 pm
Image by: Sergio Rojo / Shutterstock.com
EY has built what it calls an “AI router,” a system that matches each task to the cheapest model capable of handling it, trimming AI bills without obviously cutting output. The tool, as reported by Business Insider, is EY’s solution to a growing problem among corporate IT departments: the cost of tokens consumed by AI.
Simple Arbitrage
Not every request needs the most powerful, most expensive model. The logic is straightforward arbitrage: easy work goes to cheap models, while hard problems are left for more expensive ones.
EY’s AI Investment and Concerns
EY invests over $1 billion annually in AI, operates a fleet of around 1,000 AI agents, and has seen its AI-related consulting revenue jump approximately 30%. This scale makes token costs anything but a rounding error, especially in a market where the most AI-focused firms spend thousands per employee each month.
Its research reveals widespread concern:
82% of senior US business leaders surveyed by EY expressed worry about token-usage costs, and 98% of those using token-based tools said these costs influenced their strategy.
However, many companies remain in the dark, with only 64% actively monitoring token usage and setting budgetary guardrails.
A Shift in Mood
EY’s global AI consulting leader, Dan Diasio, summarized the change:
“‘AI saves time’ is no longer sufficient when costs mount and remain unclear.”
This shift from unbridled adoption to cost-conscious value is driving efforts like EY’s AI router.
The Paradox and a Solution
Token economics are paradoxical: as models get cheaper, enterprise AI bills triple because agentic tools consume many more tokens than simple prompts ever did. An AI router addresses this by matching tasks to the least costly model that still meets requirements, preventing total costs from ballooning.
EY is not alone in this initiative. The industry’s two-year push for tokenmaxxing (encouraging excessive AI use) has given way to controls and budgets from companies like Atlassian and Amazon.
For a consultancy, however, the router is also a product. EY sells AI advice to other firms, making this tool a demonstrative evidence that it can help clients tame their costs as it has done for itself.
The survey supports this:
76% of leaders told EY off-the-shelf software no longer meets their needs, and 91% now consider building in-house AI tools critical.