IT Home reported on November 11, 2026, citing Business Insider, that after several years of encouraging employees to adopt AI, companies including Goldman Sachs, Cisco and JPMorgan Chase are shifting to more targeted deployment. Measures include usage budgets, tracking of calls, and limits on which employees or tools can be used, as firms raise AI spending while trying to reduce related security risk.
Lari Hämäläinen, a senior partner in McKinsey & Company’s technology, media and telecommunications practice, said more companies see employee AI use “rising sharply” and therefore need clear rules and limits.
Goldman Sachs chairman and CEO David Solomon said this month that the bank is rapidly widening AI-tool coverage so its 45,000 employees can use more products, but that not every employee needs daily access to the most advanced models. In a Bloomberg Television interview, he said some work does require frontier models, “especially in cybersecurity,” while “many jobs do not need the most cutting-edge models” and ordinary AI tools are enough to improve efficiency.
Other firms that first used loose policies to drive adoption are now capping spending. Cisco chief product officer Jeetu Patel said last month on the CXOTalk podcast that the company’s early approach was like learning to ride a bicycle: development teams were allowed unlimited tokens for AI coding tools for months so people could learn the technology before efficiency was emphasized. As use grew, Patel said he plans to set token budgets for managers at every level in a team of 32,000 employees.
Business Insider previously reported that in September this year JPMorgan Chase limited some engineers’ use of Anthropic’s Claude Code to $2,000 a month. IT Home noted that, at the exchange rate it cited, that was about 13,376 yuan. A bank spokesperson said the firm is tying AI cost to productivity, work quality, risk reduction and actual business impact.
The tighter controls come less than a few months after some companies publicly highlighted how many tokens employees were consuming. Visa said monthly token use rose from 1 trillion in February this year to nearly 2 trillion in March, and that it rewarded the heaviest users. In March this year, Nvidia CEO Jensen Huang said the company planned to spend up to $2 billion a year on tokens for engineers. IT Home’s cited conversion was about 13.376 billion yuan. He joked that he would “feel very worried” if an engineer paid $500,000 a year—about 3.344 million yuan at IT Home’s cited rate—used less than $250,000 worth of tokens, about 1.672 million yuan.
Some companies also introduced dashboards so managers could see AI use, and in some cases to encourage more of it. A Disney dashboard launched in April this year showed the most active employees averaging hundreds of requests a day and tens of millions of tokens. In May this year, KPMG US said a dashboard for its consulting business was meant to push about 10,000 employees to use AI more often and more deeply. As AI became part of daily work, KPMG stopped monitoring use through that dashboard. A spokesperson said the tool had finished its original adoption role.
With AI bills still rising, the focus is moving from scale to return on each token. A Menlo Ventures study found enterprise AI deployment spending had reached $177 billion by September this year—about 1.18 trillion yuan at IT Home’s cited rate—compared with $33 billion for all of 2025, about 220.696 billion yuan. Higher spending has not clearly produced matching returns. A Gartner survey of about 2,000 technology executives from May to July this year found that only 13% said their organizations had gained significant value from AI.
Alex Sobol, co-founder of Millennium Alliance, an invitation-only forum for corporate executives, said bills are swelling in part because employees have little incentive to limit use. Without caps, he said, employees “will try every way to use AI.” Broad access can also create security problems if it is misused. A KPMG survey in 2025 of more than 1,000 U.S. employees found that 44% had used AI at work without authorization or inappropriately, and 46% had uploaded sensitive company information or intellectual property to public AI platforms.
HR software startup Rippling took a similar step. In August this year it set monthly caps on AI tools after projecting that token spending could reach 40% of the R&D team’s overall labor-cost budget. Rippling said limits plus optimized model routing cut the projected token-spend share to 10%–15%, while token volume stayed roughly at the earlier peak. Chief product officer Matt MacInnis said: “Our purpose is not to cut AI costs. As long as this spending produces real production value, we can fully accept costs continuing to grow.”
Original article: IT Home