Salesforce Advanced AI Models

Salesforce has introduced two advanced AI models—xGen-Sales and xLAM—designed to enhance its Agentforce platform, which seamlessly integrates human agents with autonomous AI for greater business efficiency.

xGen-Sales, a proprietary AI model, is tailored for sales tasks such as generating customer insights, summarizing calls, and managing pipelines. By automating routine sales activities, it enables sales teams to focus on strategic priorities. This model enhances Agentforce’s capacity to autonomously handle customer interactions, nurture leads, and support sales teams with increased speed and precision.

The xLAM (Large Action Model) family introduces AI models designed to perform complex tasks and trigger actions within business systems. Unlike traditional Large Language Models (LLMs), which focus on content generation, xLAM models excel in function-calling, enabling AI agents to autonomously execute tasks like initiating workflows or processing data without human input. These models vary in size and capability, from smaller, on-device versions to large-scale models suitable for industrial applications.

Salesforce AI Research developed the xLAM models using APIGen, a proprietary data-generation pipeline that significantly improves model performance. Early xLAM models have already outperformed other large models in key benchmarks. For example, the xLAM-8x22B model ranked first in function-calling tasks on the Berkeley Leaderboards, surpassing even larger models like GPT-4.

These AI innovations are designed to help businesses scale AI-driven workflows efficiently. Organizations adopting these models can automate complex tasks, improve sales operations, and optimize resource allocation.

The non-commercial xLAM models are available for community review on Hugging Face, while proprietary versions will power Agentforce. xGen-Sales has completed its pilot phase and will soon be available for general use.