Intents in Agent Builder

Intents are the core meaning behind the contact's words. They're the need or goal the contact wants to communicate or achieve. Your Mpower Agent analyzes each utterance What a contact says or types. and determines the closest match from the configured intents. The Mpower Agent can then respond to the contact, using the response configured in a story Used to train an Mpower Agent for interaction handling based on intent and context. or rule Used to define an Mpower Agent's response to messages that don't change with context. for the matching intent.

Every utterance a contact sends has an intent. Many messages relate to the reason the contact is reaching out to your organization. The intents for those messages relate to tasks your Mpower Agent A virtual agent created with CXone Mpower Agent Builder that can handle voice or chat interactions. handles, such as resetting passwords or giving directions and business hours. Other messages follow social scripts where the intents are greetings, goodbyes, thanks, small talk, and so on.

Intents are the starting point when configuring a new Mpower Agent. You create an intent and provide examples of messages that contacts use to convey that intent. Next, you configure how you want Mpower Agent to respond to the intent. For example, you can have the Mpower Agent ask follow-up questions, respond with an answer, offer the contact choices, or transfer to a live agent.

Working with intents is an important part of configuring your Mpower Agent during the implementation process. After deployment, it's critical to continue working with your intents as part of the ongoing management of your bot.

Intent Training Examples

Intent The meaning or purpose behind what a contact says/types; what the contact wants to communicate or accomplish. examples train your Mpower Agent on the various ways a contact might express an intent. The more examples of an intent you have, the more variations of that intent your Mpower Agent will be able to correctly identify. Use real-world conversation data to test and train your Mpower Agent. Always opt for quality data over quantity for best results. This helps ensure that the Mpower Agent is well-prepared to handle conversations with real contacts.

During the planning phase of the implementation process, you gather records of real conversations your agents have had with contacts. Use these conversations as a source of intent training examples to add to your Mpower Agent.

Intent Prediction and Confidence

Your Mpower Agent analyzes each message the contact sends. It determines how closely the message matches each of its configured intents and calculates how confident it is in each match. The higher the confidence percentage, the better the match. The Mpower Agent uses the intent with the highest confidence percentage that's over the configured threshold.

If the Mpower Agent's confidence level falls below the threshold of all of the configured intents, the NLU This process expands on Natural Language Processing (NLP) to make decisions or take action based on what it understands. fallback option is triggered. You can configure the threshold that triggers fallback on the NLU > Fallback tab in Agent Builder.

Out-of-Scope Intents

There may be tasks that you haven't configured your Mpower Agent to handle yet, but that you know contacts The person interacting with an agent, IVR, or bot in your contact center. will ask about. Out-of-scope intents allow your Mpower Agent to provide the contact information about how they can accomplish tasks that it cannot help with.

For example, if contacts regularly ask your Mpower Agent to do balance checks, but you haven't configured this task yet, you can create a balance_check intent. Add training data examples to it from the requests made by contacts. Then, create a rule for the intent with Mpower Agent responses such as:

• "I'm sorry, I can't do a balance check yet. Would you like me to transfer you to someone who can help with that?"
• "I'm sorry, I'm not able to do a balance check yet. I can help you with updating your address, resetting your password, and finding branch locations. What would you like to do?"
  

Along with an out-of-scope intent, you also need to create a story to train the Mpower Agent for an out-of-scope path. This teaches the Mpower Agent how to handle various situations that aren't covered by other intents.

Out-of-scope intents are similar to fallback, but they can provide a better customer experience. Having intents for the common questions people ask your Mpower Agent that are out of scope allows you to provide responses customized for each question. This helps your Mpower Agent seem more human-like and conversational.

Intent Confusion

An Mpower Agent may sometimes choose the wrong intent The meaning or purpose behind what a contact says/types; what the contact wants to communicate or accomplish. for the contact's message. This can happen when two intents are similar enough that the Mpower Agent has a hard time telling them apart. This is intent confusion, and it can happen if you use the same training data for more than one intent. It can also happen if the examples are so similar, it's hard to tell them apart, even if the specific examples aren't the same. Two user goals might seem different at first but start to gather similar examples over time.

Keep your intents and their training data distinct. If you want to reuse training examples for more than one intent, it's a sign that you might be able to merge the intents into a single, more general intent instead.

There are two common scenarios where intent confusion happens:

• Providing information: Your Mpower Agent might ask the contact to provide information at various points in a conversation, so you could end up with intents such as provide_email and provide_name. However, training data examples for these variations are too similar:
  • "My name is Sherry Khan."
  • "My email is sherry.khan@email.com."
  • "It is Nathoo Mannkind."
  • "It's nathoo@otheremail.com."
• Asking for information: A contact might ask the Mpower Agent for information, such as an account balance. If there are several types of information a contact could ask about, you might have an intent for each type. For example, for an account balance request, you could have intents for balance_checking and balance_savings. However, training data examples for these variations are too similar:
  • "Could I have my checking account balance?"
  • "What's my savings balance?"
  • "Can you tell me my checking account balance?"

The only real difference in the training data examples for each of the preceding scenarios is the entity Keyword or phrase defined in your company profile in Interaction Analytics. Related to an entity type. Can include variants. that's provided.

Avoid Intent Confusion

The general rule to follow to fix intent confusion—or to avoid it—is to merge intents and train on entities. It's better to have a single, general intent that you apply all your training examples to, then use stories to train your Mpower Agent to recognize the entities. When the Mpower Agent can recognize the entities in contact utterances, it can take the appropriate path.

The following table shows the common scenarios from the preceding section. It demonstrates the original intents, the new general intent, and example training stories to use.

Scenario	Original Intents	New, General Intent	Train with Stories
Providing information	provide_address provide_email provide_name	provide_info	Create several stories that start with contact message examples that contain the various types of information the Mpower Agent can collect:  "My name is Sherry Khan." "My email is nathoo@email.com." "The new address is 123 Oat Street, Sometown, Ohio 43210" All stories trigger the same intent, provide_info. Each story uses a different entity, such as accountAddress, accountEmail, and accountName. The last example, "I need to update my info," doesn't state a specific type of information, so the Mpower Agent might need to ask a follow-up question.
Asking for information	balance_checking balance_savings balance_giftcard	balance_checks	Create several stories that start with contact message examples such as:  "What's my checking account balance?" "How do I check my savings account balance?" "I want to check my gift card balance." "What's my account balance?" All stories trigger the same intent, balance_checks. Each story uses a different entity, such as balanceChecking, balanceSavings, balanceGiftCard. The last example, "What's my account balance," doesn't specify an account type, so the Mpower Agent might need to ask a follow-up question.

Multi-Intents

Mpower Agents can identify one or more intents The meaning or purpose behind what a contact says/types; what the contact wants to communicate or accomplish. in a contact's message Anything a contact says in a bot interaction, whether question or statement, written or spoken.. Identifying more than one intent in a message allows your Mpower Agent to correctly handle requests such as "Please show me my balance and transfer $200 to my savings account."

If you create a multi-intent from two existing single intents, you don't have to provide many examples. Your Mpower Agent can recognize the single intents in the message and handle them as a multi-intent.

Only use multi-intents when they are really necessary to the natural flow of conversation. Too many multi-intents can make your Mpower Agent too complicated to manage easily.

Intent Review and Management

Each intent on the Intents tab in the NLU  section has a number next to it. The number indicates how many training examples that intent has. The numbers are color-coded to indicate if the training examples are sufficient:

• Red: Intent is used up to seven times. Add more training examples.

• Orange: Intent is used between eight and fourteen times. Add more training examples if possible.

• Green: Intent is used more than fifteen times. No more training examples are required.

Some intents may have a yellow triangle next to the number of training examples. This indicates that there are no dialogues Mpower Agent stories, rules, and flows in Agent Builder. that include this intent. If you have intents with this symbol, add them to stories or rules as appropriate.

On the Intents tab, you can also perform the following intent management tasks:

• View the dialogues (stories and rules) an intent appears in.
• See recent contact messages and the intent predicted for each one.
• Hide an intent from training.

Best Practices for Intents

Follow these best practices when creating intents:

• Intents aren’t always clear-cut. Two user goals might seem different at first but start to gather similar examples over time. Keep your intents and their training data distinct. If you want to reuse training examples for more than one intent, it's a sign that you might be able to merge the intents into a single, more general intent instead.  This helps you avoid intent confusion.
• Always include an out-of-scope intent. Out-of-scope intents allow your Mpower Agent to respond to contact requests that are outside the tasks it is trained to do.They allow you to recover the conversation and often result in improved performance.
• Test and train. Test your Mpower Agent to uncover problem intents. For example it might provide incorrect responses or choose an out of scope intent at times when you think it should have chosen a different intent.
• Use multi-intents sparingly. Only use multi-intents when they are really necessary to the natural flow of conversation. Too many multi-intents can make your Mpower Agent too complicated to manage easily.