Goldman, Cisco and JPMorgan Put Budgets and Gates on Employee AI

In September, JPMorgan capped some of its engineers' monthly use of Anthropic's Claude Code at $2,000 — a concrete sign that the era of unlimited workplace AI tokens is ending at major employers.

Goldman Sachs, Cisco and JPMorgan are all putting new guardrails on how staff use AI, according to a Business Insider report. The goal isn't austerity. It's making sure each dollar of AI spending maps to actual work, while shrinking security exposure at companies pouring ever-larger sums into the technology.

Lari Hämäläinen, a senior partner in McKinsey & Company's technology, media and telecommunications practice, said AI use is "really picking up" and employers now need to set limits.

What are the new limits?

Goldman Sachs CEO David Solomon said this month that the bank keeps expanding the range of AI tools for its 45,000 employees, but not everyone needs the top-tier models.

"We need frontier models for certain things, particularly cyber," he told Bloomberg TV, adding that "there are a lot of things where we don't need the frontier" to make employees more productive.

Cisco took a staged approach. Chief Product Officer Jeetu Patel compared the company's early policy to learning to ride a bicycle: coding teams got months of unlimited tokens to build fluency before efficiency mattered, he said on the CXOTalk podcast last month. Now managers across his 32,000-person team will receive token budgets.

JPMorgan, alongside its $2,000 monthly cap, said it was tying costs to productivity, quality, risk reduction and business impact.

Why did the bills balloon?

Months ago, several employers bragged about how many tokens their workers consumed. Visa said it doubled monthly token use from 1 trillion in February to nearly 2 trillion in March, and handed prizes to its heaviest AI users. Nvidia CEO Jensen Huang said in March the company aimed to spend up to $2 billion a year on tokens for engineers — and that he would be "deeply alarmed" if a $500,000 engineer used less than $250,000 worth.

Dashboards amplified the push. Disney's showed its biggest AI users making hundreds of requests a day and burning tens of millions of tokens. KPMG US launched a dashboard in May for its advisory unit's roughly 10,000 workers to encourage more "frequent and sophisticated" use, then discontinued it once AI became routine, a spokesperson said.

The economics have shifted. A Menlo Ventures study found enterprise AI deployment spending hit $177 billion by September, up from $33 billion in all of 2025. Yet only 13% of about 2,000 tech executives polled by Gartner from May through July said their organizations had gained significant value from AI.

Alex Sobol, cofounder of the C-suite community Millennium Alliance, said workers have little incentive to curb usage. Without limits, "they're going to use it in every possible way they can," he said.

Is security driving the crackdown too?

Partly. A 2025 KPMG survey of more than 1,000 US workers found 44% used AI at work without authorization or in inappropriate ways, and 46% had uploaded sensitive company information or intellectual property to public AI platforms.

Workforce-software startup Rippling set monthly caps in August after projecting token spending could reach 40% of its R&D head count budget. The caps plus smarter model routing cut that projection to between 10% and 15%, while token use stayed near its peak.

"The point isn't to cut AI costs," Chief Product Officer Matt MacInnis said in a statement. "We're actually fine with cost growth, provided it's productive."

That line may define the next phase of enterprise AI: not less usage, but metered, routed and audited usage — with the frontier models reserved for the jobs that genuinely need them.