Consulting firms AI spending is under fresh scrutiny, as a wave of cost reality bites across the industry and forces both advisory firms and their clients to move from open-ended investment to measured, accountable deployment.
The mood shift has been building across corporate America. Amazon recently removed an employee-made leaderboard that tracked AI token usage after it encouraged excessive spending. Walmart, which built a vibe-coding tool for its workforce, has now set limits on token consumption. Uber COO Andrew MacDonald said it is hard to justify the money his company spends on AI.
Cisco Chief Product Officer Jeetu Patel put the frustration plainly at a recent event, saying the price of tokens is ‘far higher than the actual value these tokens are generating at scale.’
Consulting firms caught between selling AI and managing its cost
For consulting firms, the pressure is acute. The rise of AI posed an early existential question for the industry: if chatbots can perform the analytical legwork of junior consultants, what is the business model? Most firms moved quickly to position themselves as integrators, helping clients adopt AI while deploying it internally. Now the bill is arriving for both sides of that bargain.
KPMG has built a dashboard to track how often employees in its US advisory division use AI tools, going live late last year, with the goal of pushing adoption from basic to sophisticated use. The system compares individual usage against a personal target and against teammates, according to Business Insider. Some employees told the outlet the tool is easy to manipulate and may not accurately reflect genuine day-to-day use.
The dashboard arrived as KPMG laid off 400 employees in that same US advisory division, which employs approximately 10,000 people. The firm has also launched an AI Spark Innovation Awards programme to reward creative AI use and partnered with the University of Texas at Austin to study effective adoption. A KPMG spokesperson told the Times of India: ‘the people who get the most value out of AI did not have the most technical knowledge or simply used AI for basic tasks.’
McKinsey is pushing further still. CEO Bob Sternfels said in January that the firm operates roughly 25,000 AI agents alongside its 40,000 human employees, with the ambition that one or more agents will eventually support every person. A later account, attributed to Sternfels on an HBR IdeaCast podcast episode by Yahoo Finance, put the agent count at 20,000; both figures are attributed to Sternfels at what appear to be different points in time. The January figure of 25,000 is used here.
AI spending set to double, but return expectations sharpen
The broader data on consulting firms AI spending confirms the scale of the commitment. Boston Consulting Group found that companies expect to more than double their AI spending in 2026, from roughly 0.8% of revenue to about 1.7%. For large enterprises, that shift represents billions of dollars flowing into strategies that remain, in many cases, experimental.
BCG’s AI Radar 2026 report, based on a survey of 2,360 executives across 16 markets and nine industries including 640 CEOs, found that 94% of companies plan to keep investing in AI even without immediate returns, while 90% believe AI agents will produce measurable results in 2026. PR Newswire reported that 72% of CEOs now describe themselves as the primary decision maker on AI, and half said their jobs depend on getting it right.
More than 30% of organisations’ AI budgets for 2026 have been committed to agentic AI, according to the BCG AI Radar 2026 report. The most active CEOs, described by BCG as ‘trailblazers’, spend more than eight hours a week on their own AI upskilling and invest twice as much as peers in capability-building across their organisations, allocating 60% of their budgets to workforce skills.
Russell Fradin, CEO and co-founder of Larridin, a platform that helps companies including major consulting firms measure AI returns, said the direction of travel is not in doubt. ‘We haven’t seen anyone talking about spending less in AI next year,’ he told Business Insider. ‘They’re just talking about instrumenting to understand where it goes.’
Companies, Fradin said, are arriving at the consensus that they ‘can’t 10x spend every year forever.’ The question now is whether the measurement infrastructure catching up with the spending will, at last, tell them what they are actually getting for it.