Update: Following the publication of the report “Salesforce regrets firing 4000 experienced staff and replacing them with AI agents”, Salesforce has shared a clarification via email addressing the claims around job cuts and its AI strategy. The company stated that it did not lay off 4,000 employees (confusion came from a video interview of CEO Marc Benioff back on August 29, 2025), but instead rebalanced its headcount, redeploying roles from support into sales to increase distribution capacity. Salesforce emphasised that this was a strategic redeployment of roles, not a reduction in its overall workforce.
The company also clarified that it is not pulling back from large language models (LLMs). According to Salesforce, what is currently underway is optimisation rather than retrenchment. As the industry moves from early pilots to full-scale AI deployment, Salesforce said it has become clear that AI must be grounded in accurate data, business logic, and strong governance to deliver trusted and predictable outcomes at scale.
Additionally, the comments made by the senior vice president, product marketing at Salesforce, Sanjna Parulekar, around “trust” were shared in the context of customers transitioning from experimentation to production use. The company reiterated that these remarks do not reflect a loss of confidence in LLMs, but rather highlight the need to augment them with proprietary data, guardrails, and deterministic frameworks to ensure enterprise-grade reliability.
The rest of the story continues below.
Back in September, Salesforce (software company) redeployed around 4,000 experienced staff as part of a broader strategic shift toward integrating agentic AI into core operations. Salesforce CEO, Marc Benioff, mentioned during a podcast that it reduced its support staff from 9000 to 5000 employees (approximately 4,000 jobs) through AI deployment.
Reportedly, the company is now regretting its decision to replace its staff with AI agents. It is pulling back from its heavy reliance on large language models after encountering reliability issues.
Salesforce is said to have encountered several critical technical challenges with large language models during real-world applications, like models beginning to omit directives if given more than 8 instructions. Meanwhile, Home Security company Vivint, which uses Agentforce to handle customer support for 2.5 million customers, experienced these reliability issues firsthand. It was also noticed that AI agents lost their focus on their primary objectives when asked irrelevant questions.
It appears that Salesforce CEO Benioff’s AI ambitions are now colliding with market reality, as the CEO earlier suggested that the company might rebrand itself as “Agentforce.” The company seems to be too confident in the ability of AI systems to fully replace human judgment, particularly in complex customer service scenarios.
According to the Times of India, a Salesforce spokesperson has said in a statement, “While LLMs are amazing, they can’t run your business by themselves. Companies need to connect AI to accurate data, business logic, and governance to turn the raw intelligence that LLMs provide into trusted, predictable outcomes. That’s why we built Agentforce: trusted AI infrastructure that drives real business value. We ground AI in tight guardrails and deterministic frameworks, optimising LLMs to deliver enterprise-grade reliability. Trusted, Reliable, Secure. This is what AI is meant to be.”
