A set of curated design strategies to inspire and empower you.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
.svg)
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
Infer an entity in a new context, even though it’s omitted.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Gaining a head start here will return the greatest benefits to the user.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Anticipate and offer user intent based on If-Then business rules.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Tell the user what they want to know before they ask for it.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Anticipate and offer user intent based on If-Then business rules.
You only get one chance to make a first impression. Minimize user effort by providing an intelligent entry point for the interaction with a yes/no question, a simple menu with underlying NLU, or a predictive example at the NLU.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Allow users to leapfrog and take the next logical step.
Like a good chess player, the system should always be thinking many moves ahead. With next turn anticipation, we can be prepared to keep up with the user if they push forward.
A set of rules dictating when, and how, to handle an intent switch driven by the user.
When a user requests a new intent, sometimes we switch without looking back. Or we can switch and then return. But if we're in the middle of something, we might want to save the new intent for later.
A set of business rules defining dependent intents that may need to be addressed in addition to the requested intent.
We may need to drive some other intent(s) to accomplish the requested one. It all comes down to business rules.
Allow users to navigate through the interaction at their own pace.
Sometimes, you don't have all the information in front of you. Other times, you need a break from interacting with the system. By listening for 'hold on' everywhere, we can accommodate all sorts of interruptions for our users.
A system-initiated outreach that could either deflect an inbound contact, or drive one that the system has been designed to handle.
Instead of waiting for a user to make contact, we might have information we want to communicate proactively to ease the user's experience.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Gaining a head start here will return the greatest benefits to the user.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Anticipate and offer user intent based on If-Then business rules.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
Infer an entity in a new context, even though it’s omitted.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
Give users the flexibility from within a task to access a separate task.
Allow users to make an out-of-context change within a given intent.
Recognize more than one intent in a single utterance and act on it.
Show users we listened by transitioning from the first intent to the second and knowing which task might come first. If we already know the answer to "Is there anything else we can help you with?", don't ask it yet.
Allow users to provide multiple pieces of relevant information in a single turn, in a non-prescribed order.
Infer an intent in a new context, even though it’s omitted.
The easier it is to correct mistakes, the less serious they seem.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Allow users to both reject and correct the last recognition with a single utterance.
Use a word referring to or replacing a word used earlier in an interaction.
A type of anaphora that allows users to refer back to the prompt as part of their response.
It's natural for talkers to refer to bits and pieces of the conversation figuratively. Save users from tedious utterances by accepting phrases like "the first one" instead of the literal option.
A type of anaphora that supports users indirectly referencing something from their profile.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Gaining a head start here will return the greatest benefits to the user.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Anticipate and offer user intent based on If-Then business rules.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Anticipate and offer user intent based on If-Then business rules.
You only get one chance to make a first impression. Minimize user effort by providing an intelligent entry point for the interaction with a yes/no question, a simple menu with underlying NLU, or a predictive example at the NLU.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Tell the user what they want to know before they ask for it.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Prompting that contextualizes user data.
When we’re leveraging data to personalize a user’s experience, it’s important to expose where the data originated and why it was used.
Proactively handle calculations which would be meaningful to the user, reducing cognitive load.
Users might not be in the mood to do mental math when they're interacting with the system. An easy way to reduce the user's cognitive load is to have the system proactively do calculations.
Based on an extensive set of user data, train a prediction model to offer the most likely intent.
When we design we’re always thinking about which next action adds the most value to the user experience. Using data, ideally from multiple channels, we can act on a high confidence prediction with a yes/no question, or on a lower confidence prediction with an example at the NLU.
Proactively state information user is likely to need or ask for.
The beginning of the interaction sets the tone for the entire experience. Gaining a head start here will return the greatest benefits to the user.
Prompting that contexualizes temporal information.
Fill a slot with historical data, instead of making the user repeat it.
Don't let users get déjà vu from a previous interaction. Pull historical data and let users skip a few turns by filling slots behind the scenes. If the user opts for something different, update our understanding going forward.
Anticipate and offer user intent based on If-Then business rules.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
.svg)
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
Identify users by the unique qualities of their voice.
Passwords are so passé. With biometric technology, we can compare a user’s voice to a database and confirm their identity. Voice Bio offloads the tedium of ID/Auth tokens by letting the user simply speak.
Accommodate specific users by customizing the experience.
Most organizations go to great lengths to maintain a database with customer information. If you've got that data, flaunt it - if and when appropriate! Saying a first name or congratulating a user on their company anniversary builds trust.
.png)
.png)
.png)
.png)
.png)
