Salesforce caps employees' AI spending after some ran up thousands of dollars a month in tokens

Tech Business

Salesforce Caps Employee AI Token Budgets at $100 to $1,000 a Month

Salesforce set monthly AI token budgets from $100 to $1,000 per role after some staff spent thousands of dollars a month, with an approval ladder managed through DevBar alerts.

By
Grace Kim
Filed
Channel
Tech Business
Read
3 min read

Salesforce has imposed monthly AI token budgets of $100 to $1,000 per employee after some staff ran up "thousands of dollars per month on tokens" last quarter, according to an internal memo viewed by Business Insider.

Chief Operating Officer Miguel Milano announced the policy in the memo. "Starting this month, every role has a monthly token budget, so we can focus our AI spend on the work that matters most for our customers and business," Milano wrote. He added that "with AI confidence now high, we're shifting focus from adoption to impact."

A Salesforce spokesperson framed the move differently. The spokesperson told Business Insider the company does not have companywide individual token budgets and leaves AI usage management to team leaders. "Our focus is always to ensure that our AI investments deliver the greatest impact for our business and customers," the spokesperson said.

How do the caps break down?

The limits vary sharply by organization and role, according to the memo:

  • Sales, under President and Chief Revenue Officer Alexa Vignone: $1,000 a month for technical roles, $100 for non-technical roles such as sellers.
  • Customer Success, under Jim Roth: $500 for technical roles, $100 for non-technical ones.
  • Global Professional Services and Forward Deployed Engineering, under Lori Steele: $500 for technical roles, $100 for non-technical ones.

Organizational leaders manage the budgets as a pool, "similar to T&E," so one person's spike can be offset by lower usage elsewhere on the team.

What happens when someone hits the limit?

The enforcement mechanism runs through an internal application called DevBar. When an employee approaches 100% of the budget, both the employee and their manager receive a Slack alert.

The approval ladder then escalates by magnitude:

  • A one-time overrun of up to 20% above the limit requires a second-line manager's approval.
  • Spending up to three times the limit requires a third-line manager plus a return-on-investment justification, valid for 90 days.
  • Anything above three times the limit goes to Vignone, Roth or Steele together with the finance team, with monthly reviews.

The memo also directs staff to match the model to the task. It recommends starting with Slackbot and Gemini for drafting, summarizing and research, and notes that "even heavy users typically spend <$100 per month."

Salesforce piloted the budgets on one team before rolling them out. That team "was operating well under budget by month two with no drop in output," the memo said. In countries with works councils, leaders will see only pooled team spending rather than individual usage data, a concession to European labor rules.

Is Salesforce alone in capping AI spend?

No. Business Insider previously reported that Coinbase sets weekly AI spending caps of $500 to $5,000 depending on role and seniority, and that JPMorgan Chase introduced a $2,000 monthly limit for some engineers using Anthropic's Claude Code.

The Anthropic connection runs deep at Salesforce. Chief Executive Marc Benioff said in May that the company expected to spend roughly $300 million on Anthropic tokens this year, mostly for coding agents, according to the report.

The news landed on a day when Salesforce shares led a broad software sell-off, one day before Anthropic prices its IPO.

The pilot result suggests the caps will hold output steady while cutting runaway spend, but the real test comes as the $300 million token bill grows and engineering teams push against the three-times ceiling.

Source: mlq.ai

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,099 articles

Related articles

« Previous article