Salesforce introduces new AI assistant, Einstein Copilot, for all its CRM apps

Customer relationship management software (CRM) leader Salesforce is announcing a big update today to its artificial intelligence (AI) suite, Einstein, at its annual Dreamforce 2023 conference in San Francisco.

Einstein Copilot is a new generative AI conversational assistant that Salesforce is bringing natively into its CRM and all supported apps, allowing it to help with a wide range of application-specific tasks.

Salesforce’s enterprise customer administrators can also customize how Einstein Copilot works and what data of theirs it can access and reference, as well as harness third-party LLMs such as OpenAI’s GPT-3.5, using a new Einstein Copilot Studio.

“We truly believe every industry CEO of any level needs to adopt this,” said Muralidhar Krishnaprasad, Salesforce’s executive vice president of software engineering, in an exclusive video call interview with VentureBeat.

What Einstein Copilot offers

Salesforce was early to the enterprise AI game, launching Einstein in 2016 to aid with customer outreach, search, segmentation, and product recommendations.

All throughout this year, Salesforce has been eager to show how rapidly and robustly it’s adopting generative AI — AI that can generate new content based on prompts and simple user inputs. It launched a conversational CRM tool, Einstein GPT, in partnership with OpenAI, in May.

Now it is going even further with Einstein Copilot, which Salesforce says offers multiple AI agents that can complete a range of CRM and application-specific tasks on their own.

• Sales: Automatic account updates, meeting preparation, and even auto-generating sales emails to fit the customer context.
• Service: Streamlines customer service by providing agents with timely, relevant responses, making use of real-time customer data.
• Marketing: Generates email copy and even creates website landing pages based on consumer preferences.
• Commerce: Assists in setting up digital storefronts and automating complex tasks like catalog management.
• Developers: Converts natural language prompts into Apex code and suggests coding improvements.
• Industry-Specific: Whether it’s a financial advisor creating tailored plans or a healthcare administrator reducing appointment no-shows, Einstein Copilot can be customized by every Salesforce customer administrator to suit their business’s needs.

Specifically, Einstein Copilot Studio “will also provide configurability to make Einstein Copilot available for use across other consumer-facing channels like websites to power real-time chat, Slack, WhatsApp, or SMS,” according to a Salesforce press release.

In other words, a Salesforce customer can use Einstein Copilot Studio to build chatbots for its employees and deploy it in the employee Slack channels (Slack being owned by Salesforce, of course), or build a chatbot for customers to help answer their questions about products and services and deploy it on their company website or texting number, or any wide variation of implementations.

Using generative AI often requires some degree “trial and error,” entering multiple variations of prompts to coax an GenAI model to deliver the exact (or close enough) response you want. Einstein Copilot Studio also contains a “Prompt Builder” that lets nontechnical users simply describe what they want to do, then the Prompt Builder turns the plain English into the right prompt to produce their desired result.

“For example, a marketer could ask Prompt Builder to generate a personalized message and discount for a new product based on the customer’s purchase history and location. Einstein Copilot will then auto-generate personalized messages that align with individual customer preferences, reference past purchases, and demographic information,” Salesforce’s press release states.

Security and trust are paramount

A lot of that customization power comes down to the data that Salesforce’s enterprise customer admins provide to Einstein Copilot — which is kept secure on Salesforce’s Sales Cloud, not sent through the public internet, nor to third-party LLMs that the customer may want to tap.

“We make sure that whatever LLMs we use underneath the covers has zero retention, that it cannot store data beyond just the answer it’s going to give,” Krishnaprasad told VentureBeat.

That’s why underlying Einstein GPT and Einstein Copilot is Salesforce’s Einstein Trust Layer, a “secure AI architecture” that is constantly running and screening every AI response.

The layer also includes warnings about biased generative AI responses, and records every AI interaction for record-keeping, compliance, and auditing.

Einstein Copilot supports multiple data formats gathered from many disparate locations, from sales objects in the CRM to emails to product information to Google Insights, Apache Parquet and JSON. You can bring data in from Databricks and even run Einstein Copilot using a private Einstein instance available on Amazon Sagemaker.

All the data is then analyzed and sorted into a visual relationship graph showing lines connecting products, contacts, and more. And of course, made accessible to Einstein Copilot.

(Editor note: To help enterprise executives learn more about how to manage their data to prepare for generative AI applications, VentureBeat is hosting its Data Summit 2023 on November 15. The event will feature networking opportunities and roundtable discussions among executives around strategy. Pre-registration for a 50% discount is open now.)

A receptive market?

Demand for AI in business is growing, across sectors. According to a Gartner survey cited by Salesforce, 45% of executives are ramping up their AI investments, though a separate VentureBeat survey conducted ahead of the VB Transform 2023 conference found that only 18.2% of respondents thought their companies were willing to spend more on the tech.

Salesforce customer Shohreh Abedi, EVP at AAA, is optimistic about the new Einstein Copilot capabilities.

“We see a ton of value in implementing Salesforce’s conversational AI assistants to drive greater customer engagement,” she said.

Heathrow Airport is another organization already experiencing the benefits of Einstein Copilot, using it to create personalized interactions based on real-time data of millions of passengers.

As a result, Einstein Copilot seems poised to see widespread adoption — though it is still early days for Salesforce’s new AI assistant.