2 min Applications

Salesforce Data Cloud eliminates the need to fine-tune LLMs with vector databases

Salesforce Data Cloud eliminates the need to fine-tune LLMs with vector databases

Salesforce recently added another lot of AI functionality to its Einstein 1 Platform. Among other things, vector databases are now better supported. In addition, the capabilities of the generative AI assistant Einstein Copilot have also been expanded.

With all the new AI-enabled functionality Salesforce is now bringing to its renamed Einstein 1 Platform, teams of users will soon be able to plug AI into their workflows more efficiently.

In this regard, it does not matter whether they build their AI-based application for the Salesforce platform or use the generative Einstein Copilot assistant to retrieve required data for their (development) projects.

For this, the new functionality should combine the elements from Salesforce Data Cloud and Einstein Copilot to the various CRM applications and Einstein Copilot Studio for building AI-based applications.

Enhanced vector database support

New features include enhanced support for vector databases in Salesforce Data Cloud. This now converts unstructured data into a workable vector format. This data is combined with structured information that should make generative AI tooling and analytics available in workflows for various Salesforce CRM solutions.

This, in turn, should also provide better insights and analytics via prompts.

Ultimately, this vector database support makes it easier for developers to use their combined unstructured and structured data to train various LLMs. The company promises that fine-tuning of these models will be unnecessary from now on.

Update Einstein AI Copilot

The update for the generative Einstein AI Copilot assistant also offers this enhanced vector database functionality. This technology is intended to support the new AI feature Einstein Copilot Search in the generative assistant.

Einstein Copilot Search helps users query all unstructured and structured business data. The exact results of these search queries are then brought up directly in users’ various workflows.

Finally, links are provided to the underlying source material so users can be confident that the data on which the answers are based is reliable.

Een diagram dat het proces toont van het gebruik van een raas met salesforce.


The generative AI assistant Einstein AI Copilot will not become generally available until February 2024, according to Salesforce. At the same time, the company will also test vector database support and enhanced AI Search.

Also read: Salesforce needs AI from AWS to strengthen its offerings