AI for Admins blog series post on Einstein Bot Natural Language Processing.

AI for Admins: Increase Customer Understanding with Einstein Bots Natural Language Processing


I love Einstein Bots because they help lighten the load on customer-facing teams. By handling basic tasks, they free up your employees to tackle the tougher questions. The number of digital shoppers increased by 40% in Q1 of 2020, and more customers are relying on digital systems to get information, process returns, and troubleshoot problems. And as a Salesforce Admin, you have everything you need to create a bot that can listen to your customers.

Before we begin…

Here are some terms you should know to understand what’s going on inside Einstein Bots.

Natural Language Processing (NLP) is the process that the bot uses to extract customer text, and Natural Language Understanding (NLU) takes that text and identifies the intent, or what action to take. When the customer types “Hi, please change next Tuesday’s appointment” into the bot, it knows to launch the “Change Appointment” dialog and start to change that appointment. Bots use an intent model that is made up of multiple intents, or every task that the bot can complete.

Each intent contains many utterances, which are many ways to ask for one specific thing. For “Change Appointment”, some examples are: “Change my appointment” and “Can I move my appointment?” The bot compares the customer’s text with all the utterances and then decides which intent to launch.

Why create an intent model?

Do you have a great menu-based bot that you want to take to the next level? Here are three reasons to create an intent model for your Einstein Bot.

1. Prepare your bot for a multi-channel experience.

Not all bot conversations start the same! Customers start by typing directly to the bot on channels such as SMS or WhatsApp. If you’re launching a bot on a human-first channel, your bot must be ready for anything. A robust intent model ensures that your customers stick with the bot instead of calling support.

In addition, one of the highlights when you get a smart device is testing it: asking it to tell you a joke or answer why the sky is blue. Customers expect your bot to have a personality, and by building an intent model, you’re preparing for those customer interactions.

2. Hear directly from your customers.

Do you remember digital pets such as the Tamagotchi? This ’90s toy required constant care — it had to be fed and nurtured to thrive. The best intent models have a lot in common with digital pets: I recommend checking in with the Bot Training tab weekly to view every interaction the bot has with customers to ensure the bot is categorizing inputs correctly.

Bot Training helps you see exactly how customers phrase requests and how the intent model processes them.

This weekly maintenance allows you to check in and hear directly from your customers. When you review what customers ask the bot, you can find out what your customers need — and see the exact words they use to make requests. Many people build multiple bots to handle different sets of customers. For instance, a bot that is trained to answer student questions for technical support is a lot different than a bot that helps seniors. By assigning a bot to each customer base and training separate intent models, your bot is able to better understand specific customers. If these bots were combined into one bot, the intent model would take longer to correctly identify what each customer needs.

3. Create a more robust bot.

Sometimes, automation tools can be seen as static: Build it one time, test it a few times, and revisit it to fix bugs after it’s launched. Einstein Bots that use intent models are unique in the digital automation world because they’re so dynamic. Each customer interaction is a data point on how to make the bot better. For example, your customers may speak to the bot in a different language than intended or ask it for something new. Bot traffic could correlate on important days of your calendar, or your customers may try to input data to the bot that it doesn’t recognize.

By interacting with intent data, you can make the case that your customers are asking for growth in specific areas. Companies with robust bot offerings encourage repeat interactions, which help lighten the load on your customer-facing teams and, in turn, get answers to your customers faster than ever.

The bottom line

A bot that’s prepared for text-based customer engagement helps customers get to their goal faster. Implementing a bot with an intent model helps you hear directly from your customers. The time saved by having the bot handle simple tasks leads to increases in KPIs, such as CSAT or Net Promoter Score.

Need help reporting on bot performance? The Einstein Bot Metrics AppExchange package from Salesforce Labs contains 10 prebuilt reports that cover time-based aspects of bot performance, including frequently visited dialogs, the last dialog before a session ended, or dialogs that encountered errors. If you’re looking to improve your intent model, use the Confusion reports to identify where to improve your model for the biggest impact.

The Einstein Bot Metrics package helps you visualize customer interactions with your bot.

Creating an intent model for the first time sounds scary. A great first goal would be to create an intent model with three to five trained intents, each with more than 20 utterances. When you’re ready to build, we’ve designed Einstein Bots to make the intent model process really easy! To learn more about creating an intent model for the first time, check out our video.

Want to find out how else you can work smarter with Salesforce Einstein? Head over to our AI for Admins blog series. Each post features a different AI product or topic, with real world examples.


Looking for more helpful resources on this topic? Get started here:

How Salesforce Einstein Is Supercharging Mobile Experiences.

How Salesforce Einstein Is Supercharging Mobile Experiences

While its impact is widespread, one of the most exciting aspects of artificial intelligence (AI) is its ability to create conversational interactions that generate personalized experiences, supercharging productivity and efficiency. In this blog post, we’ll explore how the implementation of large language models on mobile devices is reshaping the enterprise mobile landscape and how Salesforce […]

Einstein standing next to text that says, "How to Use Generative AI Tools to Write SOQL Queries."

How to Use Generative AI Tools to Write SOQL Queries

Salesforce Object Query Language (SOQL) is a powerful tool that allows you to retrieve data from Salesforce. You can use SOQL to query any Salesforce object, including custom objects, custom fields, and user permissions like profile and permission set perms. As a Salesforce Admin, I know that writing SOQL queries can be a pain. Not […]

Headshot of Tom Hoffman next to text that says, "AI Prompt Writing for Salesforce Professionals."

AI Prompt Writing for Salesforce Professionals

The rise of the machines Machines and artificial intelligence (AI) have been part of popular discussion since Samuel Butler authored Erewhon (1872), where his satirical utopian society explored the morality of conscious machines as a natural evolution of the Industrial Revolution. One-hundred and fifty years later, OpenAI released GPT-4, introducing the world to AI that […]