News

Salesforce Unveils Autonomous AI Customer Service Agent Powered by Einstein

• By John K. Waters
• 07/23/2024

Salesforce has announced the pilot launch of Einstein Service Agent, billing it as the company's first fully autonomous AI agent, it was designed to handle and resolve customer service issues with "human-like proficiency."

Built on the  Einstein 1 Platform, the same "brain" as Salesforce's generative AI assistant, Einstein Copilot, the new agent integrates with all company applications, executing actions autonomously and functioning like an "AI teammate," the company says. Accessible through self-service portals and various messaging channels, including WhatsApp and Apple Messages, the Einstein Service Agent was designed to handle tasks ranging from technical support to upselling products.

"Salesforce is delivering a future where human and digital agents join forces to improve the customer experience," said Service Cloud General Manager Kishan Chetan, in a blog post. "Einstein Service Agent, our first fully autonomous AI agent, will not just complete service jobs on its own; it will augment how human agents work and completely transform how service teams operate, making them far more efficient and productive. We are reimagining customer service for the AI era."

Why has Salesforce developed this new service agent? Even though 61% of people polled by the company say they like the convenience of chatbots for simple issues, 81% continue to prefer human interaction for complex problems. This poll underscores the limitations of traditional systems when it comes to meeting diverse customer needs, the company says.

The Einstein Service Agent leverages large language models to comprehend conversation contexts and customer needs, linking seamlessly with internal company records. According to Salesforce, internal guardrails ensure the agent remains on brand and grounded in company knowledge.

The Einstein Service Agent uses its reasoning engine to interpret and process information to provide answers and solve problems for customers, Chetan explained. "The advanced reasoning engine interacts with LLMs by analyzing the full context of the customer’s input to understand their intent, drawing logical inferences from the data, and connecting various pieces of information to determine the right set of actions to take," he wrote.

The Einstein Service Agent executes those actions and uses GenAI to create responses that align with a company’s brand voice, tone, and guidelines. Chetan offered an example:

"If a customer reaches out to an online shoe store to return a recent purchase, Einstein Service Agent, powered by Salesforce Data Cloud, has all the customer and business data required to autonomously process a return, such as product details, purchase history, customer preferences,  warranty, and inventory information," Chetan wrote. "With that information, Einstein Service Agent can automatically process the return and communicate with the customer from start to finish, even sending a follow-up survey to gauge their satisfaction level."

The Einstein Service Agent was designed to utilize a company’s trusted business data, including Salesforce CRM data, to ground its responses. By leveraging Data Cloud and Unified Knowledge, companies can integrate data and knowledge from third-party systems such as SharePoint, Confluence, Google Drive, and various company websites and files. This integration enables Einstein Service Agent to generate precise and personalized responses tailored to each customer's specific needs and preferences, the company said.

The Einstein Service Agent was also designed to assist customers anytime across self-service portals and messaging channels, such WhatsApp, Apple Messages for Business, Facebook Messenger, and SMS. Because the service agent understands text, images, video, and audio, customers can send photos when their issue is too difficult to explain in words, Chetan explained.

And in scenarios deemed too complex for the AI or upon customer request, the agent can hand off the conversation to a human counterpart, providing a comprehensive transcript and a conversation summary to aid the human agent's response. This feature positions human agents to efficiently continue interactions, with the AI offering further contextual suggestions, the company said.