Einstein Personalization and Copilots

Salesforce launched a suite of new generative AI products at Connections in Chicago, including new Einstein Copilots for marketers and merchants, and Einstein Personalization. Einstein Personalization and Copilots

To gain insights into these products and Salesforce’s evolving architecture, Bobby Jania, CMO of Marketing Cloud was interviewed.

Salesforce’s Evolving Architecture

Salesforce has a knack for introducing new names for its platforms and products, sometimes causing confusion about whether something is entirely new or simply rebranded. Reporters sought clarification on the Einstein 1 platform and its relationship to Salesforce Data Cloud.

“Data Cloud is built on the Einstein 1 platform,” Jania explained. “Einstein 1 encompasses the entire Salesforce platform, including products like Sales Cloud and Service Cloud, continuing the original multi-tenant cloud concept.”

Data Cloud, developed natively on Einstein 1, was the first product built on Hyperforce, Salesforce’s new cloud infrastructure. “From the start, Data Cloud has been able to connect to and read anything within Sales Cloud, Service Cloud, etc. Additionally, it can now handle both structured and unstructured data.”

This marks significant progress from a few years ago when Salesforce’s platform comprised various acquisitions (like ExactTarget) that didn’t seamlessly integrate. Previously, data had to be moved between products, often resulting in duplicates. Now, Data Cloud serves as the central repository, with applications like Tableau, Commerce Cloud, Service Cloud, and Marketing Cloud all accessing the same operational customer profile without duplicating data.

Salesforce customers can also import their own datasets into Data Cloud. “We wanted a federated data model,” Jania said. “If you’re using Snowflake, for example, we virtually sit on your data lake, providing value by forming comprehensive operational customer profiles.”

Understanding Einstein Copilot

“Copilot means having an assistant within the tool you’re using, contextually aware of your tasks and assisting you at every step,” Jania said.

For marketers, this could start with a campaign brief created with Copilot’s help, identifying an audience, and developing content. “Einstein Studio is exciting because customers can create actions for Copilot that we hadn’t even envisioned.”

Contrary to previous reports, there is only one Copilot, Einstein Copilot, with various use cases like marketing, merchants, and shoppers. “We use these names for clarity, but there’s just one Copilot. You can build your own use cases in addition to the ones we provide.”

Marketers will need time to adapt to Copilot. “Adoption takes time,” Jania acknowledged. “This Connections event offers extensive hands-on training to help people use Data Cloud and these tools, beyond just demonstrations.”

What’s New with Einstein Personalization

Einstein Personalization is a real-time decision engine designed to choose the next best action or offer for customers. “What’s new is that it now runs natively on Data Cloud,” Jania explained. While many decision engines require a separate dataset, Einstein Personalization evaluates a customer holistically and recommends actions directly within Service Cloud, Sales Cloud, or Marketing Cloud.

Ensuring Trust

Connections presentations emphasized that while public LLMs like ChatGPT can be applied to customer data, none of this data is retained by the LLMs. This isn’t just a matter of agreements; it involves the Einstein Trust Layer.

“All data passing through an LLM runs through our gateway. Personally identifiable information, such as credit card numbers or email addresses, is stripped out. The LLMs do not store the output; Salesforce retains it for auditing. Any output that returns through our gateway is logged, checked for toxicity, and only then is PII reinserted into the response. These measures ensure data safety beyond mere handshakes,” Jania said.