Salesforce has unveiled two new AI fashions—xGen-Gross sales and xLAM—designed to boost its Agentforce platform, which integrates human and autonomous AI brokers for improved enterprise effectivity.
xGen-Gross sales is a proprietary AI mannequin fine-tuned for sales-related duties, resembling producing buyer insights, summarizing calls, and managing pipelines. It automates routine gross sales actions, permitting gross sales brokers to concentrate on extra strategic duties. This mannequin boosts Agentforce’s capacity to autonomously deal with buyer interactions, nurture pipelines, and help gross sales groups with higher velocity and accuracy.
The xLAM (Massive Motion Mannequin) household introduces AI fashions designed to execute advanced duties and set off actions inside enterprise programs. Not like conventional LLMs (Massive Language Fashions), that are content-driven, xLAM fashions concentrate on function-calling—permitting AI brokers to behave independently, resembling initiating workflows or processing knowledge with out human intervention. The xLAM fashions vary in measurement and capabilities, from smaller, on-device fashions to bigger, extra strong fashions for industrial functions.
Salesforce AI Analysis developed the xLAM fashions utilizing APIGen, a proprietary data-generation pipeline that helped enhance mannequin efficiency. Early variations of xLAM fashions have already outperformed different giant fashions on key benchmarks. For instance, the xLAM-8x22B mannequin ranked No. 1 in function-calling duties on the Berkeley Leaderboards, surpassing even bigger fashions like GPT-4.
These improvements are designed to empower companies to scale AI-driven workflows extra effectively. Organizations utilizing these fashions can now automate advanced duties, enhance gross sales processes, and optimize useful resource allocation.
The non-commercial xLAM fashions can be found on Hugging Face for group evaluate, whereas the proprietary variations will energy Agentforce. xGen-Gross sales has accomplished its pilot section and can quickly be accessible for common use.