After encouraging employees to adopt AI tools, some companies are setting budgets and access rules as usage grows. According to an IT Home report citing Business Insider, the shift reflects concerns about cost, the value of AI spending and security risks.
Goldman Sachs chairman and CEO David Solomon said in November 2026 that the bank was expanding AI access for its 45,000 employees, but not every role needed daily access to the most advanced models. Cybersecurity was one area that could require a frontier model, he said, while less advanced tools could help with many other tasks.
Cisco chief product officer Jeetu Patel said the company initially gave developers months of unrestricted access to AI coding-tool tokens so they could learn to use the technology. As usage rose, he planned to assign token budgets to managers across his 32,000-person team. Business Insider also reported that JPMorgan Chase set a $2,000 monthly Claude Code limit for some engineers in September 2026. A bank spokesperson said it was weighing AI costs against productivity, work quality, risk reduction and business impact.
The emphasis on limits follows a period when some companies highlighted rising usage. Visa said its monthly token consumption grew from 1 trillion in February 2026 to nearly 2 trillion in March. Disney introduced a usage dashboard in April; KPMG US said in May that it hoped a similar dashboard would encourage roughly 10,000 consulting employees to use AI more. KPMG later stopped monitoring usage through the dashboard, saying it had served its adoption goal.
Costs and returns remain difficult to balance. A Menlo Ventures study put enterprise AI deployment spending at $177 billion through September 2026, compared with $33 billion for all of 2025. In a Gartner survey of about 2,000 technology executives conducted from May to July 2026, 13% said their organizations had obtained significant value from AI.
Security is another consideration. In a 2025 KPMG survey of more than 1,000 US workers, 44% reported using AI at work without authorization or inappropriately, while 46% reported uploading sensitive company information or intellectual property to public AI platforms.
Rippling said it introduced monthly AI-tool limits in August 2026 after projecting that token spending could reach 40% of its engineering and development payroll budget. With limits and changes to how requests were routed to models, it said the projected share fell to 10%–15%, while token use remained near its previous peak. Chief product officer Matt MacInnis said the aim was not simply to cut costs: the company would accept higher spending if it produced real value.