Salesforce launches Einstein Copilot for general availability

At its World Tour NYC today, Salesforce announced the general availability of Salesforce Einstein Copilot, its conversational AI assistant for CRM. The company first announced Einstein Copilot in September last year, and released Einstein 1 Studio in March as part of its bid to become the platform of choice for building AI assistants for business.

Salesforce also unveiled new enhancements to the platform, including “Copilot Actions,” a library of preprogrammed capabilities to help sellers benefit from conversational AI in Sales Cloud.

“This is built for Salesforce users no matter what they’re using,” said Sanjna Parulekar, VP of Einstein AI product marketing at Salesforce. “The way our Salesforce admins and developers — or Trailblazers — are going to add enterprise value to conversational AI is through these actions.”

Parulekar explained that Copilot Actions are workflows that can be built via the Einstein 1 Studio platform. Business process automation (BPA) can be an action, for instance, as can anything that updates a record, populates information, and pulls data. New preprogrammed Copilot Actions include:

• Close Plans. Using this action, sales reps and managers can ask Einstein Copilot to create a personalized close plan for an opportunity, including step-by-step tactics based on historical opportunity and account data, as well as recommended dates of action.
• Forecast Guidance. This action enables sales managers to ask questions, such as which deals are at risk, and receive a summarized list of deals that could explain why a rep might not hit their quota, as well as top reasons why those deals were chosen, including close data or lack of recent activities.
• Call Explorer. Using retrieval augmented generation (RAG), this action enables sales reps to query prior call transcripts captured in Einstein Conversation Insights. It also allows sales reps to ask questions about customer sentiment.
• Follow-up Emails. This action gives sales reps the power to ask Einstein Copilot to write customized follow-up emails based on prior calls.

“When an end-user asks Copilot, ‘Help me identify my top opportunities,’ Copilot can answer by reasoning across the actions that are available to it,” Parulekar said.

She noted that Salesforce plans to release new packs of actions for its various cloud offerings on a regular cadence.

“Einstein has always been predicated on this notion that we’re going to deploy AI in an out-of-the-box way for use cases we know our customers are looking to us to solve, but always provide a customization option,” she said.

Greater ease of use

High-level users can leverage Copilot Builder in Einstein 1 Studio to build their own actions, but the beauty of the preprogrammed actions, Parulekar said, is that users can leverage them without having to train or fine-tune a large language model (LLM). For CRM use cases such as writing an email or providing forecast guidance that draws on well-defined customer data and workflows, users can simply enable those actions and go, she said.

“For our more sophisticated customers that might have data scientists, and maybe have built their own LLM [and] have the money and resources to train on their own corpus of data, they can absolutely do that,” Parulekar says. “But that’s a small fraction of our customer base. It’s all about flexibility and making sure that AI is not cost-prohibitive for our customers.”

Salesforce has also introduced new platform enhancements, including:

• Copilot Analytics. Parulekar said a preconfigured analytics dashboard can help admins visualize key Einstein Copilot metrics such as actions used, average interactions per user, and success rates.
• Recommended Actions. The Copilot can now recommend single-click standard actions specific to the page a user is on. Examples include “summarize opportunity” on the opportunity page, or “draft an email” on the contact page.
• Mobile app access. Einstein Copilot is now live on the Salesforce mobile app, and voice-to-text functionality gives mobile users the ability to talk with their data and workflows.

“With Copilot Analytics, folks can measure the efficacy, the usage they are getting out of Copilot — who’s using it, and what they’re asking,” Parulekar said. “Copilot Analytics is really a trust feature because you need to be able to measure something and understand how to improve it in order to trust it’s doing the thing you hope it is.”

Einstein Copilot, Copilot Sales Actions, and Copilot Analytics are all now generally available. Salesforce notes that customers looking to scale Einstein Copilot can invest in its new AI Implementation Bundle, which includes discounted sandbox environments, and reduced-cost licenses to customize and deploy conversational AI.