Farukh Kitchlew | Dec 6, 2023 | 0
Salesforce introduces AI Cloud, aiming to introduce models into the workplace setting
Salesforce is introducing AI Cloud, a multidisciplinary product line combining AI. Expertise with enterprise-ready tools to strengthen its position in the competitive AI market. This extension of the company’s March-launched associative AI initiative aims to integrate AI throughout the entire Salesforce platform. Salesforce SVP Adam Caplan aims to “introduce generative AI in a trusted manner, leveraging its AI history and existing presence in the company’s stack. The company is moving quickly to establish a presence in the AI space.”
Salesforce’s AI Cloud hosts and serves content models from partners like Amazon Web Services, Anthropic, Cohere, and OpenAI. The Salesforce AI research division provides first-party models for code generation and business intelligence automation, while customers can use a specially trained data platform for data collection and analysis.
Amazon’s Bedrock offers trained models, similar to Amazon’s internal and pre-trained models, for entrepreneurs. We’re generally adopting an ecosystem strategy, an accessible and using the best model for the best use scenario, according to Caplan. Salesforce’s core products, including Data Cloud, Tableau, Flow, and MuleSoft, utilize nine Salesforce-created models in AI Cloud. These models include the GPTs for Sales, Services, Marketing, Commerce, Slack, Tableau, Flow, and Apex.
Service GPT generates customer briefings using summaries, production orders, and customer history. Sales GPT creates personalized emails, while Marketing and Commerce GPT creates target markets and offers suggestions for increasing average order value.
Tableau GPT, Flow GPT, Apex GPT, and Slack GPT are specialized frameworks for creating AI-driven no-code frameworks. Slack GPT and Flow GPT enable data insights and natural language prompts, while Tableau GPT generates visual representations. Apex GPT, Salesforce’s proprietary programming language, scans for code vulnerabilities and suggests inline code. Salesforce is promoting its Einstein Trust Layer AI moderation and redaction service, preventing text-generating models from saving customer information, similar to Nvidia’s NeMo Guardrails.
Einstein Trust Layer is designed for businesses with strict regulatory and governance requirements, addressing privacy concerns and limiting generative AI tools like ChatGPT. This is particularly relevant as companies like Amazon, Goldman Sachs, and Verizon have restricted the use of AI. Customer trust and security are top concerns for businesses, and Caplan aims to help them navigate technological advances and the new world securely.
Einstein Trust Layer recognizes sensitive information in prompts and removes it on the backend before approaching the model. The service also screens responses for toxicity, such as sexism, racism, and discrimination, to ensure the model’s integrity. Salesforce offers Einstein Trust Layer for customers connecting designs from external platforms like Vertex AI and SageMaker to AI Cloud. The company also collaborates with OpenAI to provide joint content moderation services using OpenAI’s safety tools and Einstein Trust Layer for OpenAI customers.
Caplan suggests that modifying an AI Cloud generative AI model for a specific use case can save time and money. For example, a customer can create a framework that follows a style guide for email responses or imports customer data from a Salesforce database. Caplan highlights a “fundamental change in email caliber and differentiation between generic and copy-based email content based on customer relationship management data.”
The sales pitch for Salesforce-versed clients effectively targets existing customers.
Caplan highlights “the effectiveness of customer relationship management using AI and data. Enhancing prompts, training models, and providing customers more value are essential steps in this stack.” Salesforce predicts AI Cloud launch in late November, featuring Einstein Trust Layer.