Salesforce CEO Marc Benioff says the company could spend close to $300 million on Anthropic tokens in 2026, with most of the spending tied to coding work.
“These coding agents are awesome. Anthropic is awesome,” Benioff said in a podcast, published May 15. “I am going to probably use $300 million of Anthropic this year at Salesforce. Coding. Everything’s going to be cheaper to make.” AI companies typically bill enterprise customers according to how many tokens their usage generates.
Benioff made these comments on the “All-In” podcast. He also mentioned that across service, support, distribution, and marketing that the efficiency improvements driven by AI agents have reached levels he described as “unprecedented.” He also said the pace of product iteration has accelerated sharply, and that he can now ship and sell software in parallel, something he described as impossible before AI agents.
READ: Anthropic nears $1.5 billion joint venture with Wall Street firms (May 4, 2026)
Benioff also hinted at forthcoming development work tied to the workplace messaging app Slack, which Salesforce bought in 2021. “You’re going to see some cool stuff with Slack and code I’m not ready to talk about yet,” Benioff said on the podcast. “But there’s no question that we are in a new moment in coding.”
Benioff also called for what he described as an “intermediary layer” that could route token inputs intelligently — directing complex tasks to a frontier model such as Anthropic’s Claude and simpler ones to smaller, cheaper models. He argued that all employee-generated tokens through a top-tier model is unnecessary.
Salesforce had stopped hiring software engineers in 2025. While making the announcement Benioff pointed to productivity gains of over 30% from Agentforce and other AI tools deployed across engineering teams.
READ: Salesforce unveils Slack updates, centered around AI (April 1, 2026)
Despite the freeze on engineering roles, Salesforce continued expanding in other areas. The company said it planned to bring on between 1,000 and 2,000 salespeople to support AI product adoption among enterprise customers.
Benioff has later clarified that AI has not advanced to the point of replacing engineers entirely. Salesforce currently employs around 15,000 engineers, who now work alongside AI tools including Anthropic models, OpenAI Codex, and Cursor.
In the podcast, Benioff described the shift in how engineers operate, moving into roles that involve overseeing AI agents rather than writing all code manually. He noted that while productivity has risen sharply, the models still cannot function without human oversight.