What is Sprinklr Conversational AI?

Sprinklr’s Conversational AI is an automated AI-powered virtual assistant that can engage with customers. It deciphers the intention of your customer's query, and after identifying the reason they are reaching out, it takes suitable actions as configured by the brand.

💡 Customers want to resolve issues with as little effort as possible – and as quickly as possible. Sprinklr AI bots provide customers with immediate help, without forcing them to navigate your IVR or wait on hold.

Conversational AI can be used to automate routine tasks such as checking an account balance or tracking an order. It can troubleshoot your customers' questions or provide answers to FAQs on your support or sales site or can be used to triage your customer messages and issues to increase agent productivity while enhancing the customer experience.

💡 Sprinklr’s Conversational AI is Omni-channel and supported in multiple languages. You can build the application once and deploy it across multiple channels, such as Live Chat on the web, Voice, Facebook Messenger, WhatsApp, Twitter, or SMS, to name a few.

At the same time, the platform can also automatically convert bot responses to match rich templates supported by these channels. It allows you to run the same bot on channels supporting interactive content as well as the ones supporting only text like SMS

Powered by Sprinklr AI

The bots are powered by natural language processing and AI that leverages Sprinklr’s in house AI micro-services. Sprinklr has built extensive in-house NLU capabilities for the last 7 years. This not only includes Intent and entity models but also other AI capabilities to augment the bot functionality like - text classification models, language detection, sentiment, CSAT prediction, optical character recognition, image classification etc. These models are built on large unstructured datasets, operating at a scale where Sprinklr’s AI is making more than 10 Billion predictions per day for our clients.

Bots powered by Sprinklr Conversational AI have more human like conversations as the bot has ability of:

1. Handling of multiple intents in a single conversation

2. Handling of context switching & language switching

3. Handling of information already provided to skip steps accordingly

Basic Concepts of Conversational AI

In order to serve customers in a human manner, bots need to go beyond reacting only to a preselected set of options and leverage Natural Language Processing capabilities to understand customer requests.

We have three key components to understand here:

​

1. Understanding the user message using NLP

💡 Intent: This is the intention of the user. This is typically the verb in the sentence, the “why” behind the message.

💡 Entities: These are the nouns that the customers use like a product name. However, entities can also be a value like an email address, serial number, or the last four digits of a credit card.

In a conversation, not all intentions are treated similarly. Issue type intents deal with intents which describe an issue of the customer like ‘battery replacement,’ ‘screen repair,’ ‘cancel booking,’ or ‘change flight’. Dialogue intents are concerned with generic dialogue such as “yes, that’s correct”, ‘I don’t understand the question’ or “I’m fine thanks”. It’s a part of the conversation that’s not directly related to the issue.

Read more about intents & entities - Intents

Read more about how to create intents - Discovery

​​2. Defining what needs to be done next using contextual understanding

Next building block of conversational AI is a dialogue tree which allows users to configure the input & action associated to the respective input. It is a no code configurator which has ability to add/ edit nodes.

Dialogue tree

​​3. Taking the appropriate actions

Once the system has understood the user request (the intent & entity), it then needs to be able to take the appropriate action on it. Actions can be:

• Asking a clarifying question

• Providing a response to the request

• Taking action by making changes in other systems using API calls

• Route to user to the right agent or bot who will be able to better solve the issue.

To know more about each step & other product capabilities on conversational AI -

​

How to build a conversational AI bot?

Now that we know the basic concepts of Conversational AI,

There are 5 key steps of creating a conversational AI bot:

1. Discover: Leverage unsupervised AI to analyze historical data and discover the main reasons customers reached out to the brand to create customer-centric based Bots instead of intuition-based bots.

2. Build: Manage your intents and entities and build your dialogue trees

3. Test: Test the AI and provide feedback in order to improve the accuracy of intents continuously.

4. Deploy: Define deployment strategies around how different bots get deployed in the same application.

5. Measure: Measure and report on the performance of the application to improve the bot flows

You can learn more about conversational AI in the product documentation section for Conversational AI