(Bloomberg) -- Big companies are goading workers to start using AI. Or they're gatekeeping access to it. The stance is evolving as corporate employers waffle between justifying their investments in artificial intelligence and keeping a lid on usage as computing costs mount.
Walmart Inc. has capped staffers’ use of an in-house AI agent that helps with workplace tasks, according to people familiar with the decision. And Uber Technologies Inc. is limiting each employee’s monthly spending on certain AI coding tools to $1,500 per tool. The ride-hailing company had already blown through its annual budget for Claude Code, a popular tool from Anthropic PBC.
Across industries, companies are gobbling up so many tokens — the basic unit of measurement for AI computing — that some have had to dial back their earlier anything-goes stance. It’s a sharp contrast with earlier messages encouraging or directly pressuring corporate employees to boldly experiment with a range of in-house and third-party AI tools. Firms including Accenture Plc and Coinbase Global Inc. told staffers that failing to embrace the technology could jeopardize their career development, or even their jobs. At Starbucks Corp., a quarter of the bonuses paid to tech employees will be pegged to department-wide goals that include AI adoption.
Sundar Pichai, CEO of Google parent Alphabet Inc., recently noted that monthly usage of the company’s AI products has increased sevenfold in the past year to 3.2 quadrillion tokens. “We've heard that many companies are already blowing through their annual token budgets, and it's only May,” he said at the company’s Google I/O conference on May 19. Google’s success has made its computing resources so valuable that its own AI researchers sometimes have to get in line.
“Companies have been correctly pushing employees to embrace AI, and rising token costs are a feature, not a bug,” said Matt Kropp, chief technology officer of Boston Consulting Group’s BCG X divsion, which helps clients implement AI. “That said, few companies know yet how to budget for AI, and employees are still learning how to use these tools effectively, so there is definitely waste happening.”
Amazon Inc. has sought to minimize so-called “tokenmaxxing” by shutting down a leaderboard that tracked AI token use because it encouraged some staff to perform tasks that didn't necessarily solve problems but helped employees climb the ranks, according to Business Insider. “Please don't use AI just for the sake of using AI,” Dave Treadwell, an Amazon senior vice president, reportedly told staff. Salesforce Inc., owner of the workplace productivity platform Slack, said last week in a webinar for customers that it’s tracking how token use creates real business value.
While AI can quickly boost the productivity of individuals and small teams — at McKinsey & Co., staffers are saving 30% of the time previously spent gathering and analyzing data — it’s a much bigger leap to deliver firm-wide savings. Most large companies are struggling to justify their ballooning investment with actual efficiency gains, according to a new Bain & Co. global survey of large companies. The result is a muddled minefield that employees might struggle to navigate.
“Companies have been driving their people to see who could use the most AI without a clear plan about why,” said Siobhan Savage, co-founder and CEO of Reejig, which makes software for managing workflows. “It was just ‘Get usage up.’ Some of those firms are backpedaling now and creating a bit of whiplash. If I were an employee, I’d be questioning leadership.”
At many companies, what it means to be AI-ready can vary and is often unclear. Often, it comes down to simply plowing through tokens.
“Everyone is still figuring this out,” said Helen Russell, chief people officer at software developer HubSpot Inc. “The playbook we’ve adopted, and one I would tell anyone to use, is more of a carrot versus a stick approach.”
But as spending grows, and the fear of missing out on AI’s promise spreads, many leaders have ditched carrots for sticks. Micha Kaufmann, CEO of freelance marketplace Fiverr, told employees in a memo that “AI is coming for your jobs” and warned that those who failed to embrace AI risked becoming obsolete. Amazon.com CEO Andy Jassy has said employees who “become conversant in AI” will be “well positioned,” implying those that don’t, won’t.
Some 94% of senior executives plan to continue investing in AI even if it does not pay off in 2026, according to a BCG survey published in January. But companies might need to keep closer watch on usage, especially as budgets shift toward agentic AI designed to perform work tasks. Agentic AI is “uniquely expensive,” according to new research by a team of academics working with researchers from Google and Microsoft Corp., consuming one-thousand times more tokens than more basic tasks since agents continue to soak up computing power in the background.
That might prompt more companies to take a Goldilocks approach to adoption — not too little but not too much. But pushing back on AI use just as employees are starting to get the hang of it “could be counterproductive,” BCG’s Kropp said, and may limit the benefits companies ultimately realize from AI.
Workplace consultant Phil Kirschner, who teaches corporate employees how to use AI in their jobs, said that any shift in workplace policies has potential to create whiplash for employees. But as anxieties mount over AI’s impact on jobs, changes in rules about AI adoption “come with an existential threat on top.”
--With assistance from Natalie Lung.
More stories like this are available on bloomberg.com
©2026 Bloomberg L.P.
