2026年9月10日

After laying off 4,000 employees, Salesforce admits AI is not a panacea.

Salesforce had aggressively pushed AI automation, cutting its customer support team from 9,000 to 5,...

Salesforce had aggressively pushed AI automation, cutting its customer support team from 9,000 to 5,000 and promoting large language models (LLMs) as transformative for business processes. However, reality proved challenging. Recently, Salesforce has reduced reliance on LLMs, turning to “deterministic automation” to ensure consistent execution of critical workflows.
Agentforce’s CTO noted that LLMs can miss steps with complex instructions, while deterministic automation is more reliable and cost-effective. For example, Vivint’s AI system once failed to send customer surveys consistently, resolved only by deterministic triggers. CEO Marc Benioff also acknowledged that AI models without robust data backing can produce “hallucinations,” underscoring the need for caution.
The move signals to enterprises: AI is powerful, but reliability, governance, and data infrastructure matter more.

接著讀