Salesforce Announces AI Cloud with Einstein Trust Layer Filter
Salesforce, widely known for its eponymous CRM system, has announced the release of its new AI Cloud service package. The corporate bundle contains tools that can be used to create and deploy turnkey AI work models for businesses. AI models from a number of the company's partners, including OpenAI, Amazon Web Services, Cohere, Anthropic, and others, are housed and maintained in the AI Cloud infrastructure. If necessary, its clients can host their specially trained AI models there, keeping their data separately, in private infrastructure.
Salesforce's own models in the AI Cloud are capable of generating program code, automating business processes, and more. They significantly expand the capabilities of the developer company's advanced products:
- Sales GPT (automatically generates emails that solve personalized marketing tasks).
- Service GPT (based on customer data, makes briefings to improve their service quality, forms case and order summaries).
- Commerce GPT and Marketing GPT (based on client data, generate audience segments for maximum personalization of product descriptions and effective targeting, and also give advice – for example, how to increase the average check).
- Flow GPT and Slack GPT (allow the creation of workflows with AI involvement without code).
- Tableau GPT (generates visualizations based on hints and data analysis results).
- Apex GPT (scans program code, checks it for vulnerabilities, and offers built-in code for Apex, a language created by Salesforce).
Several models from this list are already available for use today (Sales GPT, Service GPT, Commerce GPT, and Slack GPT). The company plans to launch the remaining ones, except for Flow GPT, which will be released in October, by the end of June.
A distinctive feature of AI Cloud is the moderating service Einstein Trust Layer. It prevents AI generative models from storing confidential information (for example, a client's phone number and the content of their order). Einstein acts as a kind of firewall between any service or application and the generative AI model. It checks the transmitted data for the presence of confidential information and automatically deletes such information before it reaches the AI. In addition, the Einstein Trust Layer filters out toxic statements, combating discrimination (racism, sexism, and other forms of it) not only in a human-generated prompt but also in text generator responses.
According to Salesforce representatives, access to the Einstein Trust Layer service will be available to everyone by the end of June, and the AI Cloud package will be released by the end of 2023.