Salesforce executives acknowledged decreased confidence in the reliability of large language models (LLMs). The company is shifting its Agentforce product strategy away from generative AI. This move favors more predictable, 'deterministic' automation to eliminate the inherent randomness of AI models.
Muralidhar Krishnaprasad, CTO for Agentforce, detailed a key technical issue: models omit instructions when given more than eight, posing a significant problem for business-critical tasks. This strategic pivot follows the deployment of AI agents that replaced approximately 4,000 support staff roles.
Sanjna Parulekar, SVP of Product Marketing, stated that executives "were more confident about large language models a year ago." Financially, the company's stock has declined about 34% from its peak in December 2024. The direct market reaction to this specific announcement remains unclear.