Consulting firms moved quickly to make AI part of employees’ working lives. Now they are confronting the cost of using it at scale.
One of McKinsey’s strategies for managing rising costs isn’t to cap consultants’ use of AI but to let them know when they’re using too much.
The firm tracks consumption on a user-by-user basis and alerts its employees via email when their AI usage gets too high, said Debasish Patnaik, who leads QuantumBlack — McKinsey’s AI, data, and analytics group — in the UK. The alert system was introduced firmwide in the summer.
“Similar to using mobile data on a work phone, we tell them this is how you could do things to make it more cost-effective for the firm,” said Patnaik.
Companies that initially focused on encouraging AI experimentation are now confronting the cost of operating the technology at scale.
The age of freewheeling AI token use came to an end earlier this year as LLM providers moved from subscriptions to consumption-based pricing models. Consulting firms and other enterprise users are now charged based on the number of tokens — the small chunks of text an AI model reads and produces — they use, rather than a flat access rate, making efficiency more important as usage grows.
In September, OpenAI announced that its most prolific users of AI coding agents now consume more than $7,000 worth of tokens a day.
McKinsey’s approach to managing that spending pressure rests on transparency and education, Patnaik said.
By May 2026, the firm was processing about five trillion AI tokens a month, according to a company blog post. Consumption was highly concentrated: about 10% of users accounted for roughly 65% of the total, with consultants and software engineers among the heaviest users.
The AI usage alerts aren’t intended to discourage employees from using AI, but to show them what they are consuming and help them complete the same work more efficiently.
“We really believe in giving autonomy to the consultants,” said Patnaik.
Ensuring they have the right education about how to use tools has been key for managing usage, he added, noting that McKinsey has already seen shifts in how employees use the tools since giving users more visibility into their AI habits.
Beyond token alerts
McKinsey’s internal AI spending has not yet reached a problem stage, Patnaik said.
Most of the firm’s AI use is for personal productivity or to deliver engagement faster, both of which are cases where “the cost-benefit is still skewed towards the benefit side,” he said.
Usage levels could be more “egregious” in another six months, but Patnaik said the firm had already implemented additional controls to manage usage in addition to alerts.
An internal AI gateway optimizes requests before they reach model providers, while circuit breakers temporarily pause access around particularly high token usage while the firm assesses whether the usage is productive.
Caching enables answers to repeated questions to be reused and pools consumption costs across the business rather than locking unused capacity into individual licenses.
Other consulting firms are pursuing similar controls to manage AI spending.
The Big Four firm EY has set up an “AI Value Realization Office” to manage AI spending and has installed an “invisible” router behind some specialized AI tools that directs employees to the model best suited to a task. EY told Business Insider that the router, alongside other governance measures, had reduced token consumption by 60% since April.
In June, a senior software engineer at Deloitte US told Business Insider that the changes to GitHub’s pricing model are “already wreaking havoc” on expectations for work, with developers quickly burning through their new monthly quotas, which took effect that month.
Taking the strategies to clients
Patnaik’s role as leader of QuantumBlack in the UK primarily focuses on providing AI capabilities to McKinsey’s clients, rather than the firm’s internal use of AI. McKinsey established a formal practice this spring to advise clients on deploying AI cost-effectively.
Patnaik’s broad advice for clients mirrors the firm’s internal approach.
Over the last quarter, clients have increasingly realized there’s a “hidden cost that we haven’t completely thought through,” and executives are having to weigh AI’s potential competitive benefits against its return on investment, Patnaik said.
Patnaik said companies should assess token cost per outcome rather than per employee. Simply optimizing for fewer tokens could discourage valuable use of AI.
“What you don’t want is to take costs out here, but incur costs on the other side without knowing about it,” he said. For now, he added, there’s still value in encouraging adoption, rather than restricting it.
Read the full article here
