[Salesforce Japan Co., Ltd.] Salesforce announces next-generation AI model to power Agentforce
*Salesforce Japan Co., Ltd.*
Press release: September 11, 2024
**
Salesforce unveils next-generation AI model to power Agentforce *Salesforce AI Research delivers xGen-Sales and xLAM AI models that connect data and behavior to people and autonomous agents*
*This document was released in the United States on September 6, 2024. Salesforce Delivers Next-Generation AI Models to
Power Agent force
This is an abridged translation. The official language of this material is English, which shall prevail for its content and interpretation.
Salesforce USA (hereinafter referred to as Salesforce) is Agentforce xGen-Sales, a unique model trained and designed to perform autonomous sales tasks in A large-scale behavioral model group
We announced new AI models including xLAM. Salesforce AI Research
By combining these models developed by Salesforce, Salesforce customers can create autonomous AI agents that take action.
(English) can be quickly configured and deployed to drive unprecedented scale.
Fine-tuned xGen-Sales delivers more accurate and faster responses by increasing the accuracy of relevant industry tasks, generating customer insights, enriching customer contact information, and summarizing calls. , you can automate sales tasks such as tracking your sales pipeline. This model allows Agentforce Sales
Enhanced Agents capabilities allow you to autonomously grow pipelines and coach sales reps more accurately and quickly. According to Salesforce’s own evaluations, xGen-Sales’ capabilities already exceed other, much larger models.
xGen-Sales is a large-scale behavioral model
It is a step towards the next generation language model called LAM (English). large language model
In contrast to LLM (LLM), which frequently requires human involvement and is primarily used for content generation, LAM (Function-calling) performs functions within other systems or applications. We specialize in In other words, LAM can trigger the necessary actions for an AI agent to independently perform a task on behalf of a person.
xGen-Sales plus Salesforce AI
Research offers a new LAM family called xLAM. xLAM models offer lower cost, faster performance, and higher accuracy than many of the larger, more complex models currently available. For example, the xLAM-1B model consists of just 1 billion parameters (variables that the model learns to generate results and insights), yet it
Exceeding (English). xLAM-1B is a non-commercial, open source model dedicated to advancing science with the research community, but Salesforce’s Agentforce uses a more powerful model.
*Why it’s important:*
Organizations need AI agents that can augment the work of their employees and act on their behalf so they can focus on more strategic priorities. These models not only understand the work to be handled, but also their capacity limits, so agents using the models know when to hand off tasks to others to ensure quality or complete the work. Masu. Salesforce recently
LLM Benchmark for CRM
, giving you the opportunity to navigate the many models on the market and compare the best LLM for your CRM use case.
Silvio Savarese, Chief Scientist at Salesforce
Savarese) said: “Building and training your own AI models can be time-consuming, costly, and extremely frustrating. Agentforce gives you right-sized models for your business and allows you to You will be able to achieve results.”
* News background: * Salesforce AI to train xLAM models
Research is a robust, proprietary pipeline for generating high-quality synthetic data, APIGen
I created
（Mixture of
Ideal for industrial applications that require a balanced combination of latency, resource consumption, and performance.
● *Large (xLAM-8x22B)*:
8x22B is a large-scale mixed-expert model that allows organizations with a certain level of computational resources to achieve optimal performance.
* Salesforce Perspective: * MaryAnn Patel, SVP of Product Management at Salesforce
Patel) says: “Salesforce envisions a future where salespeople are empowered by AI to help drive sales efficiency and free up valuable time to focus on customers. It’s purpose-built to help teams build generative AI solutions that augment their work.”
*Salesforce AI Research*
is Salesforce’s artificial intelligence (AI) research lab, developing new technological breakthroughs in this field. Our team of
researchers, engineers, and product managers is committed to shaping the future of AI in business through fundamental research that directly informs product development.
*Availability:*
● A non-commercial open source version of LAM’s xLAM suite is available on Hugging Face for community review and benchmark testing.
Available at A more advanced proprietary version will be included in Agentforce.
● xGen-Sales has completed a pilot and will be generally available soon.
*Detailed information:*
● To learn how small language models outperform LLM, click here . (English)
● Learn how AI agents can revolutionize the AI enterprise here
. (English)
● Click here for an explanation of large-scale behavioral models. . (English)
*The training data for the model announced this time does not include user companies’ Salesforce usage data.
*About Salesforce*
Salesforce helps businesses of all sizes
We help you rebuild your business for the AI era. Agentforce, Salesforce’s trusted platform, enables companies to connect people and agents to drive customer success powered by AI, data, and action. For details
Visit salesforce.com/jp.
Salesforce has opened “News & Insights” which compiles the latest information for Japan. For details,
Visit salesforce.com/jp/news/.
Any undelivered services or features referenced in this or other press releases or publicly available are not currently available and may not be delivered on time or at all. Customers should make purchasing decisions based on currently available features.