Salesforce CEO Benioff Projects Hundreds of Millions in AI Token Spending

Salesforce stopped hiring software engineers last year, yet CEO Marc Benioff expects the company to spend hundreds of millions on Anthropic tokens this year. The projection highlights how the software company is shifting investment from expanding engineering teams to increasing spending on AI tools that assist with coding and other tasks.

Benioff, who said the company wouldn't hire software engineers in 2025 because of productivity gains from AI, now says coding agents and AI models are changing how engineers work rather than eliminating them. That approach is in line with a broader trend in the tech industry: relying more and more on AI systems to improve efficiency and still involving humans.

Speaking on the All-In podcast, Benioff said Salesforce could spend around $300 million on Anthropic tokens in 2026, with most of that spending linked to coding-related work. He described AI coding agents as “awesome” and said the investment would lower software development costs while accelerating output.

How AI productivity gains led to hiring freeze by Salesforce CEO Marc Benioff

In 2024, Benioff first announced that Salesforce would stop hiring software engineers in 2025, citing productivity gains of more than 30% from Agentforce and other AI technologies.

“We’re not adding any more software engineers next year because we have increased the productivity this year with Agentforce and with other AI technology that we’re using for engineering teams by more than 30%,” Benioff said at the time. He added that engineering teams had achieved higher velocity due to AI integration.

Despite freezing engineering hiring, Salesforce increased hiring in other areas. Benioff previously said the company planned to add between 1,000 and 2,000 salespeople to explain AI products and their business value to customers.

Benioff later clarified that AI has not reached a stage where it can replace engineers entirely. He said Salesforce’s approximately 15,000 engineers increasingly work alongside AI tools, including Anthropic models, OpenAI Codex and Cursor.

“When they start to use these models, they’re now working not only with the AI, but agents to help them code – and they can even become somewhat supervisory over these agents. But still, those engineers are needed. The model still cannot operate autonomously,” he said.

Benioff added, “We’re not at that level yet of AI,” while repeating that Salesforce’s engineering organisation had become around 30% more productive.

The projected token spending comes as Salesforce expands its AI offerings across products, including Slack and Agentforce. Benioff also suggested the company is developing systems to route AI requests between larger and smaller models depending on complexity, potentially reducing operational costs.

Salesforce has invested more than $300 million in Anthropic and holds an estimated stake in the AI company, according to reports. At the same time, Agentforce, Salesforce’s AI-focused business, has reached $800 million in annual recurring revenue, while Slack has added new AI capabilities powered by Anthropic’s Claude models.

Benioff has described the ongoing shift as a “digital labour revolution,” with AI reportedly accounting for 30%-50% of Salesforce’s overall workload.