Salesforce strengthens AI play with vector database support, enhanced Einstein Copilot

Salesforce is strengthening its AI offering with new support and functionality.

Just ahead of World Tour NYC, the company announced it is adding vector database support and enhancing its Einstein Copilot generative assistant with AI search capabilities.

The features will make their way into the Einstein 1 Platform that brings all the elements together, right from the Data Cloud and Copilot to CRM apps and the Copilot Studio for building AI-powered apps.

This, the company says, will make it easier for teams to take advantage of AI in their workflows, whether they are building a custom generative AI app or calling up the Copilot assistant to get some information for their project.

Vector database support in Salesforce Data Cloud

Salesforce Data Cloud is the company’s in-house data platform that combines data points from different customer touchpoints to host unified customer profiles in real time. To use this data to enable generative AI in their CRM workflows, teams had to fine-tune off-the-shelf large language models (LLMs) on both structured and unstructured data. This took a lot of time and resources. 

With integrated vector database support for Data Cloud, Salesforce is ending the hassle of fine-tuning. Essentially, it converts all unstructured business data – from PDFs and emails to documents and transcripts – into a usable vector format. It unifies it with structured information (like purchase history or support tickets) to enable generative AI and analytics in various Salesforce CRM application workflows. This ultimately provides users with robust data insights and streamlined processes – accessed via relevant prompts.

For instance, when building an email campaign, a marketer can use Marketing Cloud Intelligence to analyze unstructured survey data and transcripts hosted inside Data Cloud and understand the intent of the targeted customer base. Then, using those insights, they can iterate on email templates.

Salesforce is focusing on offering vector database support as most organizations realize the untapped potential of unstructured data (which makes up 90% of total enterprise data). This is especially the case for data science and AI. Major data ecosystem vendors like Snowflake and Databricks also offer support for both structured and unstructured data today.

“By integrating both structured and unstructured data, our new Data Cloud Vector Database transforms all business data…into valuable insights. This advancement in Data Cloud, coupled with the power of LLMs, is a game-changer, fostering a data-driven ecosystem where AI, CRM, automation, Einstein Copilot, and analytics turn data into actionable intelligence and drive innovation,” said Rahul Auradkar, EVP and GM of unified data services at Einstein at Salesforce.

Einstein Copilot will get AI search

Along with CRM apps, such as those targeted by sales and marketing professionals, Salesforce’s own generative AI assistant, Einstein Copilot, will also get the benefit of vector database capabilities in the Data Cloud.

Essentially, the technology will power AI-driven Einstein Copilot Search, a feature that will allow users of the assistant to query all their business data, structured as well as unstructured. It will then pull up precise information in their flow of work. 

Imagine a sales rep preparing for a customer meeting by prompting Einstein Copilot to pull up specific information, like the customer’s financial filings or past email interactions, and ask follow-up questions about that data to get a detailed understanding. Salesforce says users can expect a similar experience across other CRM apps that support Einstein Copilot as a conversational helper for different tasks. 

Additionally, it will provide citations to its source material, making sure that teams can use the information it generates with complete confidence.

Availability in 2024

While the features push Salesforce’s AI agenda ahead, it is important to note that none of these capabilities are available just yet. In February 2024, the company will launch Einstein Copilot for general use.

As the conversational assistant becomes available to customers, the company will pilot vector database support and AI search. This will also happen in the same month, the company has confirmed, although the timeline for general availability remains unclear at this stage.