Salesforce announces AI Cloud to empower enterprises with trusted generative AI

Salesforce announced today the launch of AI Cloud, an enterprise AI solution aimed at boosting productivity across all Salesforce applications. The new open platform integrates various Salesforce technologies like Einstein, Data Cloud, Tableau, Flow and MuleSoft, offering real-time generative AI capabilities that can be easily incorporated into business operations.

“Generative AI…may be the most important technology of any lifetime,” Salesforce co-founder and CEO Marc Benioff said at Salesforce’s AI Cloud announcement event in New York City. Benioff said that Salesforce’s client Gucci, the fashion house, had already been working with an early version for its AI Cloud for more than a year, and had seen tremendous gains in its customer support and response capabilities.

At the core of AI Cloud lies the new Einstein Trust Layer, which the company believes will set a new industry standard for enterprise AI architecture. The company asserts that this layer not only leverages the advantages of generative AI but also guarantees the utmost data privacy and security.

Moreover, customers will have the freedom to choose their preferred large language models (LLMs) for their specific use cases. Options will include models from major tech companies like OpenAI, Anthropic, and Cohere and Salesforce’s proprietary and domain-specific models.

“We created AI Cloud to help customers leverage AI-created content across sales, customer service, marketing, commerce and IT interactions to better connect with their audiences in new, more personalized ways,” Adam Caplan, SVP of emerging technology at Salesforce, told VentureBeat. “Our new offering will empower enterprises to safely turn on the power of generative AI across all of their applications, connect all of their enterprise data, gain new insights about their customers, and automate workflows.”

Salesforce says that the Einstein Trust Layer aims to establish trust in enterprise generative AI by protecting sensitive data within AI applications and workflows. The layer will prevent the integration of proprietary data into public models and address critical concerns related to data privacy, security, residency, and compliance in generative AI.

The company has also developed prompt templates and builders to improve the quality and relevance of AI-generated content. These tools use harmonized data to ground the generated outputs in the unique context of each company. They also strive to eliminate AI hallucinations and optimize workflows.

Empowering safer use of generative AI models

New research conducted by Salesforce indicates that by 2030, AI is expected to drive over $15 trillion in global economic growth and increase GDP by 26%. However, the widespread adoption of AI is contingent upon establishing trust and safeguarding data privacy.

The research findings reveal that 73% of employees perceive generative AI as introducing new security risks, and nearly 60% of those intending to use the technology are uncertain about securing their data.

To tackle this challenge, the AI Cloud incorporates a Trust Layer, prioritizing data privacy and security. The company aims to grant users greater control over enterprise AI and data strategy by providing an open ecosystem.

Salesforce also asserts that integrating generative content powered by secure CRM data will cultivate trust among users.

“With trust at the center, AI Cloud will help customers get the most out of their data using generative AI to create a better customer experience and empower employees to work smarter and faster,” Salesforce’s Caplan told VentureBeat.

Caplan said the new offering enables sales teams to generate personalized emails customized to meet their customers’ needs. Additionally, marketers can use the tool to create audience segments, using natural language prompts and AI-driven recommendations to enhance targeting.

A customizable platform for LLM deployment

The Einstein Trust Layer offers a flexible platform that empowers customers to integrate their preferred large language models (LLMs) into their chosen deployment environment.

Developed in collaboration with OpenAI, the platform enables customers to use external models while ensuring the confidentiality of their prompts and generative responses within Salesforce’s secure environment. To enhance security measures, AI Cloud hosts Salesforce’s proprietary LLMs, as well as LLMs from AWS, Google, Anthropic, Cohere and other domain-specific models. This comprehensive approach allows businesses to integrate diverse large language models with customer data.

The introduction of Einstein GPT serves as a gateway for customers to use their preferred LLMs. Salesforce’s LLMs, including CodeGen, CodeT5+ and CodeTF, offer advanced capabilities like code generation and business process automation assistance. 

Salesforce’s Caplan added that the company has developed a set of guidelines specific to generative AI, based on Trusted AI Principles, a framework to help companies think through how to work thoughtfully with generative AI.

“Our guidelines for responsible generative AI aim to help Salesforce and all users of generative AI mitigate these issues by following responsible best practices when it comes to development,” said Caplan. “Customers who have trained their own models outside of Salesforce can benefit from AI Cloud while storing data on their own infrastructure. These models connect directly to AI Cloud through the Einstein Trust Layer, where all customer data remains within the customers’ trust boundaries.” 

Salesforce’s announcement of its new AI Cloud services comes as competition in the fast-moving generative AI-for-enterprise space continues to escalate, with new startups launching targeting the same customer base, and incumbents like Microsoft and Google also offering enterprise-friendly AI services.