Track · Product Operations3 min read

Salesforce Caps Employee AI Spend at $1,000 a Month

Salesforce gave every role a monthly AI token budget after some employees spent thousands per month, with technical roles capped at $1,000 and a tiered approval ladder above that.

Salesforce sets AI budgets after employees spent 'thousands of dollars per month on tokens,' memo says
Salesforce sets AI budgets after employees spent 'thousands of dollars per month on tokens,' memo saysAI-generated

Context

  1. Some Salesforce employees spent thousands of dollars per month on tokens last quarter, per an internal memo.

  2. Technical roles in sales get $1,000/month; Customer Success and Professional Services technical roles get $500/month; non-technical roles get $100/month.

  3. Exceptions above 3x the individual limit require approval from Alexa Vignone, Jim Roth, or Lori Steele plus finance, reviewed monthly.

  4. A one-team pilot was operating well under budget by month two with no drop in output.

  5. CEO Marc Benioff said in May Salesforce expected to spend roughly $300 million on Anthropic tokens this year.

Some Salesforce employees spent thousands of dollars per month on AI tokens last quarter, and now every role at the company has a monthly token budget. The shift comes from an internal memo sent by Chief Operating Officer Miguel Milano and viewed by Business Insider.

"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 in the email to employees.

The move lands at a moment when AI cost discipline is becoming a product-operations problem, not just a procurement one. For product managers running AI features internally, the memo is a rare full spec of how a large enterprise turns token spend into a managed, exception-driven budget line.

What are the actual budgets?

The limits vary sharply by org and role type:

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

A Salesforce spokesperson said the company does not have companywide individual token budgets, and that team leaders manage AI usage. "Our focus is always to ensure that our AI investments deliver the greatest impact for our business and customers," the spokesperson said.

How does the exception process work?

The memo builds a tiered approval ladder that scales with the size of the ask:

  • Up to 20% above the individual limit, one-time: the second-line manager approves.
  • More than 20% above, up to 3x the limit: the third-line manager approves, with a required ROI justification; the increase lasts 90 days and gets reviewed on expiry.
  • Above 3x the limit: Vignone, Roth, or Steele approve, together with finance, and these cases are reviewed monthly.

Enforcement runs through Slack. When an individual's spend approaches 100% of budget, both the person and their manager get a notification from the DevBar app, with an option to request approval up to 20% above the limit. Org-unit leaders manage pooled team budgets, similar to travel and expense, so one person's spike can be offset by others. In Works Council countries, leaders see pooled team spending only, not individual usage.

Did the pilot hold up?

Salesforce piloted the budgets on one team, T&P, for a month. The memo states the group "was operating well under budget by month two with no drop in output." Milano's guidance to employees also notes that even heavy users who match the model to the task, starting with Slackbot and Gemini for drafting, summarizing, and research, typically spend under $100 per month.

The memo frames the change as a maturity marker: "With AI confidence now high, we're shifting focus from adoption to impact." Milano asks leaders to review usage monthly, question whether less spend could achieve the same result, and add token consumption to manager one-on-ones.

Who else is doing this?

Salesforce is not alone in capping AI spend. Coinbase has weekly AI spending caps ranging from $500 to $5,000 based on employee role and seniority. JPMorgan Chase introduced a $2,000 monthly limit for some engineers using Anthropic's Claude Code. The scale of the underlying bill is visible in CEO Marc Benioff's May statement that Salesforce expected to spend roughly $300 million on Anthropic tokens this year, largely for coding agents.

The tradeoff is real. Caps that are too tight discourage adoption; caps that are too loose reproduce the "thousands per month" problem the memo opens with. Salesforce's answer is a pooled, exception-driven model with an explicit escape valve: "We never want a budget to get in the way of great work: if you have a high-impact use case that needs more, ask."

Expect token budgets to follow the same path as cloud spend controls before them: dashboards first, then chargebacks, then unit-economics pressure on every AI feature a product team ships.

via feeds.businessinsider.com (Source)

More from Marcus Bennett

Marcus Bennett

Show full bio

Market editor covering regulation, policy and their impact on product management at Roadmap File.

56 articles