Suivez votre progression
Accueil Trailhead
Accueil Trailhead

Get to Know Einstein Reply Recommendations

Learning Objectives 

After completing this unit, you’ll be able to:

  • Describe Einstein Reply Recommendations.
  • Explain the key benefits of Einstein Reply Recommendations.
  • Prepare your data for Reply Recommendations. 

What’s Einstein Reply Recommendations?

After yet another successful quarter, solar solutions company Ursa Major Solar is powering their customers’ lives like never before. And CEO Sita Nagappan-Alvarez just made a groundbreaking deal with an energy company in Mexico that expands Ursa Major’s customer base two-fold. With all the new customers, Sita wants her support team to be prepared with the best tools to solve cases as efficiently as possible. Luckly, she knows that her trusted admin, Maria Jimenez, has been exploring new ways to support customers using Salesforce Service Cloud. 

Maria shows Sita all the ways Service Cloud helps improve the customer experience. In fact, Salesforce research indicates that 79% of businesses see an increase in overall customer satisfaction scores (CSAT) with the adoption of customer service powered by artificial intelligence (AI). One tool stops Sita in her tracks. 

Einstein Reply Recommendations offers agents a list of suggested replies in real time on messaging and chat channels. The recommended replies are generated using deep learning and natural language processing (NLP) from your historical chat transcripts.

Einstein’s recommended replies shown alongside an agent’s chat window.

Depending on your support team’s needs, Einstein can recommend greetings, follow-up questions, and more. 

Einstein Reply Recommendations and AI 

Sita’s priority is to gain the trust of her customers, so she understands the importance of investing in the technology her teams use every day. She knows that Einstein Reply Recommendations is an artificial intelligence feature, but what exactly does that mean? She asks Maria for details about how it works.

Maria and Sita chatting in the breakroom.

Maria explains that Einstein Reply Recommendations is a really smart tool, but it doesn’t fix problems on its own. Einstein Reply Recommendations gathers data from closed chat transcripts and organizes this data into patterns. This smart technology is called natural language processing (NLP) or natural language understanding (NLU). You can think of these organized patterns as Einstein making a list of problems and another list of solutions that match up to form a given data set. It’s a match made for customer service success. 

First, Einstein analyzes your org’s chat data and generates a list of common replies used by the support team during chats. After you review and fine-tune the replies, you publish them for Einstein to recommend to agents. You can always add new replies and update your existing replies later on. 

When a customer enters a question or message in a chat or messaging session, Einstein gets to work, scanning through your list of message-reply data. Then Einstein suggests one or more of the published replies for the agent to insert directly into the conversation. If needed, the agent can edit the language. These automated replies save support agents from digging through notes or asking their colleagues for advice. 

Benefits of Einstein Reply Recommendations

Sita knows that first impressions matter. To get customers to trust solar energy, customers need fast feedback and excellent service. Solar power is a competitive industry, and Ursa Major Solar wants to do everything possible to retain its loyal customers. 

Luckily, Einstein Reply Recommendations puts just the right response at agents’ fingertips, so your agents are able to:

  • Quicken their response time with the right recommendations at the right time—with all replies in the same place. 
  • Improve the accuracy of their responses.
  • Focus on complex solutions rather than routine problems. 
  • Standardize their communication to customers.

Most importantly, Einstein can help shorten your chat-closing process. No more sticky notes, switching between apps, or consulting higher-ups to find solutions to routine problems. Sita can’t wait to make her agents’ jobs easier with Einstein Reply Recommendations. 

Prepare for Einstein Reply Recommendations

Before jumping into the world of Einstein Reply Recommendations, you need data. 

Why is data key to your success? Well, data is everything when it comes to building powerful AI models. Einstein needs enough data to detect patterns in your team’s communication style. The more data you have for Einstein to learn from, the more accurate its recommendations will be.

To produce a list of common replies, Einstein needs to analyze at least 10,000 English chat transcripts. If you don't have enough closed chats or the chats are too short to provide meaningful data, Einstein lets you know when you start to build your model.

Follow along with Maria to learn how an admin would complete these six steps to get Einstein Reply Recommendations up and running.

  1. Review data requirements.
  2. Turn on Einstein Reply Recommendations.
  3. Build your model.
  4. Review and publish replies.
  5. Give agents access.
  6. Activate your model.

Timeline of the six steps to set up Einstein Reply Recommendations. 

Get ready for some happy customers, courtesy of AI. 

Resources