Salesforce Summaries – Einstein Bots: Build a CRM Powered Chatbot

Share this article...

SalesforceSummaries: a series delivering key insights from Salesforce YouTube videos, to save you time as you keep up to date with the latest technological changes within the Salesforce ecosystem.


In this presentation, you will learn more about Salesforce Einstein powered chatbots and how they can be used to increase productivity.

Details: ‘Einstein Bots: Build a CRM Powered Chatbot

**Presenter: **George Hu, Jonathan Rico

**Details: **20 minutes

Key Terms: Einstein, Chatbots

  1. @1.10 — Einstein Bots is your digital support agent. It’s designed for simple tasks and offers 24/7 support. Einstein Bots are not designed for replacing agents or long conversational engagements.

  1. @2.20 — The design principles of Einstein Bots are:

Don’t pretend to be human

Leverage integrated data to provide user value

Clearly state the capabilities of the bot

  1. Escalate to an agent if the bot is unable to assist.

  1. @3.30 — So how does Einstein Bots work?
  2. @3.50 — There are 2 types of customer interactions.

Menus are 100% accurate but offer limited coverage whereas text input is conversational but requires training.

  1. @5.00 — The first type of interaction is when the customer makes a request. When a customer asks a question, Einstein’s Natural Language Processing (NLP) attempts to map that to an intent. An intent is always linked to a dialog, and the dialog is where you can script the response.
  2. @5.50 — The second type of interaction is when Einstein Bot responds to a request. For example, the bot may ask the customer ‘what is your email address?’. The response from the customer is validated (the data type in the customer’s response is validated).
  3. @8.05 — In the demo, an example of a menu interaction looks like this:

In this example, the bot is being leveraged in Live Agent. Also, in this example, the dialogue is text input (as the customer inputs some text; which has an intent).

  1. We can leverage Twilio integration with Salesforce to understand more about who the customer is; which will help us to provide an improved customer service experience.

Now that the customer has verified his/her phone, using Apex, Salesforce can look up the phone number that was input into the ‘slot’ (data storage) and return the customer detail.

In Salesforce, we can now see that a new Food Order record has been created.

Now that the dialogue has completed, with the customer having ordered pizza, the Food Order record in Salesforce has been updated now to ‘In Progress’.

  1. @11.55 — You can teach Einstein Bot to deliver promo codes to customers. By navigating to Setup > Objects > Einstein Bot, you can see a list of all the bots in your Salesforce org. By clicking on a record, you can see the bot performance information:

By navigating to Dialogs, you can see where all the logic for how the bot should interact with customers is stored:

By navigating to Promo code, you can see all the ways which you can manage the promo code process:

With bots, we only care about the conversation. Once the data is in your Salesforce system from the conversation, that’s when you can apply automation on those records (process builders, invocable methods, apex, flows, workflow rules etc).

  1. @17.05 — Before activating a bot, you can use the helpful ‘preview’ function to test the bot.

Add Comment