How do you create a chatbot and connect it to Salesforce?

How did the COVID-19 pandemic affect chatbots?

Chatbot usage expanded during the COVID-19 pandemic as gaps in customer service and onboarding began to surface. During the pandemic, the chatbot market exceeded $17 billion -- and over the next five years, it will pass $100 billion, according to Mordor Intelligence.

Beyond customer service, chatbots also appear in employee onboarding to bolster its processes and perform screenings, gather new employee information upfront and conduct policy presentations.

The surge in chatbot popularity over this time also correlates to chatbots' expanding capabilities. For example, Salesforce's Intro Template for Einstein Bots enables developers to generate new chatbots with stock feature sets for order retrieval, lead generation and initiating new support cases, among other capabilities.

Another part of chatbots' increase in popularity is that chatbot technology developed a better command of language. People in different regions express themselves differently, which made it difficult for chatbots to deliver satisfying customer interactions. Salesforce, in particular, has a consolidation of natural language processing libraries to help test new models, which enable more accurate and inclusive communications between customers and chatbots.

During the pandemic, the chatbot market exceeded $17 billion -- and over the next five years, it will pass $100 billion.

Moreover, Salesforce has also launched prebuilt Einstein Bots, a chatbot collection that performs many common functions. These preconfigured bots make the technology uptake faster, as developers don't need to make them productive.

How to create a chatbot

At its simplest, a chatbot is the front end to an API, preconfigured for a chatbot-style UI -- commonly called an agent. The agent works with Salesforce to execute tasks and workflows as it connects to the Salesforce Lightning rules engine.

Follow these three steps to create a chatbot:

1. Select a platform for generating an agent. Users have many from which to choose -- for example, DialogFlow, which Google hosts.
2. Generate the bot per the platform's instructions. Users need to configure it, set up user expressions that the bot can respond to (intent) and identify objects in the expressions (entities).
3. Train the bot. Practice using the expressions to see if the objects work as planned, and add intents and entities as needed. Users must also allow room for errors.

Most platforms can help users do all this without code.

How Salesforce chatbot implementation works

After a Salesforce user creates a chatbot, that person must learn how to connect it to Salesforce. Users have two ways to implement this implementation: with an external platform or internally through Einstein Bot.

Connecting a chatbot through an external platform

If a user creates a chatbot outside of the Salesforce platform, that person must connect the bot to Salesforce Lightning. That's where the API comes in. Sometimes, the chatbot platform directs integration tools -- software development kits with the appropriate classes built in -- but sometimes, it doesn't. If not, some types of middleware can invoke the API. From there, users can set up the app in the Salesforce Lightning app manager, add it to the utility bar and select the chatbot component.

However, users may find it easier to create a chatbot on the Salesforce platform without connecting to it externally. For that scenario, Einstein Bot is available.

Building a chatbot in Salesforce

In Salesforce Bot Builder, the first step is to build the intent model through a manager -- even adding chatbot response pauses so the interactions resemble humans'. The next step is to activate the bot and make it ready to train, as described above. The bot can train on an extensive ready-made body of historical agent training data within the Einstein Bot process.

From this point, the bot can access the full power of Einstein. Users can tie the bot to a broad range of Salesforce objects and processes to access customer profiles and historical data, trigger workflows and store data collected in the user exchange. For highly complex bots, Salesforce offers Einstein Bots for Developers, where developers can do complex custom coding.

Finally, Salesforce makes a dashboard available to monitor the bot's performance once it's deployed.