Suggest Replies with Einstein Service Replies and Reply Recommendations

Service Replies uses Einstein generative AI to draft responses for your support team, without requiring separate model training or regular Reply Recommendations setup. In a chat, Einstein quickly drafts fluent and courteous replies based on conversation context. From the case, Einstein drafts emails based on recommended Knowledge articles. . Add grounding to Service Replies with Service AI Grounding or Agentforce Data Library to get recommendations based on your Knowledge base, your uploaded files, or the case object.

Reply Recommendations without generative AI drafts replies based on your org’s closed chat transcripts. No need to manually compose or search for replies. Service reps can select the most relevant replies from a short list to post in open chats or messaging sessions. Reply recommendations are extra helpful to new reps or to reps handling multiple chats simultaneously.

Note To keep things clear in the age of AI agents, we’re now referring to the humans who use Service Cloud as “service reps” rather than “agents”. We wish we could snap our fingers to update this language everywhere, but you can expect to see the original use of “agent” in a few places until we replace it, including in the app itself.

• Einstein Service Replies
  Increase customer satisfaction and reduce service rep handling time with AI-generated replies. Einstein Service Replies for Chat follows a Messaging Session or Live Chat conversation as it unfolds and then drafts and recommends fluent, courteous, and relevant replies for your service reps to review, edit, and post. Or, they can draft and send emails based on recommended knowledge articles, directly from the Case page.
• Einstein Reply Recommendations
  Build a model that recommends relevant replies to support reps in the console during chat and messaging sessions. Replies are based on your Salesforce org’s closed chat transcripts, and Einstein recommends replies only if you’ve reviewed and published them.