Virtual assistant interaction models
Virtual assistant interaction models
Virtual assistant interaction models
Virtual assistant interaction models
Overview
While at the Expedia Global Operations (GCO) group, as part of the innovation pod I worked on the virtual assistant design.
At the time our call center agents had to juggle multiple tools, they needed to remember different policies for different brands and points of sale, they needed to adapt and search for the right information all the time. Their workflows required lots of clicks and memorization!
This project brought an opportunity to present the customer support agents with the best contextual help and simple intuitive interactions as they would service the customers. We wanted tools that could adapt, search for the right information and offer automated workflows that required minimal clicking on the agents' behalf!
While at the Expedia Global Operations (GCO) group, as part of the innovation pod I worked on the virtual assistant design.
At the time our call center agents had to juggle multiple tools, they needed to remember different policies for different brands and points of sale, they needed to adapt and search for the right information all the time. Their workflows required lots of clicks and memorization!
This project brought an opportunity to present the customer support agents with the best contextual help as they would service the customers. We wanted tools that could adapt, search for the right information and offer automated workflows that required minimal clicking on the agents' behalf!
While at the Expedia Global Operations (GCO) group, as part of the innovation pod I worked on the virtual assistant design.
At the time our call center agents had to juggle multiple tools, they needed to remember different policies for different brands and points of sale, they needed to adapt and search for the right information all the time. Their workflows required lots of clicks and memorization!
This project brought an opportunity to present the customer support agents with the best contextual help as they would service the customers. We wanted tools that could adapt, search for the right information and offer automated workflows that required minimal clicking on the agents' behalf!
While at the Expedia Global Operations (GCO) group, as part of the innovation pod I worked on the virtual assistant design.
At the time our call center agents had to juggle multiple tools, they needed to remember different policies for different brands and points of sale, they needed to adapt and search for the right information all the time. Their workflows required lots of clicks and memorization!
This project brought an opportunity to present the customer support agents with the best contextual help as they would service the customers. We wanted tools that could adapt, search for the right information and offer automated workflows that required minimal clicking on the agents' behalf!
While at the Expedia Global Operations (GCO) group, as part of the innovation pod I worked on the virtual assistant design.
At the time our call center agents had to juggle multiple tools, they needed to remember different policies for different brands and points of sale, they needed to adapt and search for the right information all the time. Their workflows required lots of clicks and memorization!
This project brought an opportunity to present the customer support agents with the best contextual help as they would service the customers. We wanted tools that could adapt, search for the right information and offer automated workflows that required minimal clicking on the agents' behalf!
Problem
Problem
We want to provide agents with a virtual assistant (bot) help while servicing customers, so that agents can complete their work fast, efficiently and with the utmost accuracy.
My problem space was to design for complex escalation and transfer cases and to validate the designs through user research. Escalation and transfer functionality allows agents to seek help from their colleagues or higher tier agents and lets the customers be transferred to a supervisor if requested.
We want to provide agents with a virtual assistant (bot) help while servicing customers, so that agents can complete their work fast, efficiently and with the utmost accuracy.
My problem space was to design for complex escalation and transfer cases and to validate the designs through user research. Escalation and transfer functionality allows agents to seek help from their colleagues or higher tier agents and lets the customers to be transferred to a supervisor if requested.
We want to provide agents with a virtual assistant (bot) help while servicing customers, so that agents can complete their work fast, efficiently and with the utmost accuracy.
My problem space was to design for complex escalation and transfer cases and to validate the designs through user research. Escalation and transfer functionality allows agents to seek help from their colleagues or higher tier agents and lets the customers to be transferred to a supervisor if requested.
We want to provide agents with a virtual assistant (bot) help while servicing customers, so that agents can complete their work fast, efficiently and with the utmost accuracy.
My problem space was to design for complex escalation and transfer cases and to validate the designs through user research. Escalation and transfer functionality allows agents to seek help from their colleagues or higher tier agents and lets the customers to be transferred to a supervisor if requested.
We want to provide agents with a virtual assistant (bot) help while servicing customers, so that agents can complete their work fast, efficiently and with the utmost accuracy.
My problem space was to design for complex escalation and transfer cases and to validate the designs through user research. Escalation and transfer functionality allows agents to seek help from their colleagues or higher tier agents and lets the customers to be transferred to a supervisor if requested.
Role
When I joined the team, there were two initial virtual assistant concepts explored. I took those concepts and iterated on them further as well as added my own new explorations as I had to accommodate more complex escalation and transfer scenarios.
For three months I worked closely with two project managers, two engineers, an architect, and a copywriter, and we all met daily. I also collaborated and brainstormed with the designer who designed two original virtual assistant concept. I presented and reviewed my designs with the org's executives bi-weekly.
When I joined the team, there were initial virtual assistant concepts explored. I took those concepts and iterated on them further as well as added my own new explorations as I had to accommodate more complex escalation and transfer scenarios.
For three months I worked closely with two project managers, two engineers, an architect, and a copywriter. I presented and reviewed my designs with the org's executives bi-weekly.
When I joined the team, there were initial virtual assistant concepts explored. I took those concepts and iterated on them further as well as added my own new explorations as I had to accommodate more complex escalation and transfer scenarios.
For three months I worked closely with two project managers, two engineers, an architect, and a copywriter. I presented and reviewed my designs with the org's executives bi-weekly.
When I joined the team, there were initial virtual assistant concepts explored. I took those concepts and iterated on them further as well as added my own new explorations as I had to accommodate more complex escalation and transfer scenarios.
For three months I worked closely with two project managers, two engineers, an architect, and a copywriter. I presented and reviewed my designs with the org's executives bi-weekly.
Process
Process and solution
Process and solution
To kick off the project, I took a close inventory of the transfer process in the current tools
As I worked on the consult and transfer functionality in the new chat platform, I defined the flows and scenarios, sketched, whiteboarded, wireframed and reviewed the designs with the innovation pod on a daily basis
After critiquing the designs with the team, I iterated more and presented my iterations to the team the next day
- I brought new ideas to expand the existing virtual assistant concepts
I conducted a usability study to test initial virtual assistant concepts with the agents
Based on the agents' feedback, I tweaked the initial design concepts
I used Framer to demonstrate micro-interactions
I started using the brand new visual UI Kit desinged by our UI designer and throughout my time on the project we continued collaborating on different design patterns that we needed to make available within escalation / transfer flows
Throughout the project, I remained a strong advocate for adding a human touch to the virtual assistant functionality
- To kick off the project, I took a close inventory of the transfer process in the current tools
- As I worked on the consult and transfer functionality, I defined the flows and scenarios, sketched, whiteboarded, wireframed and reviewed the designs with the innovation pod on a daily basis
- After critiquing the designs with the team, I iterated more and presented my iterations to the team
- I conducted a usability study to test initial virtual assistant concepts with the agents
- I used Framer to demonstrate micro-interactions
- I started using the brand new visual UI Kit for my visual design styles
- Throughout the project, I remained a strong advocate for adding a human touch to the virtual assistant functionality
- To kick off the project, I took a close inventory of the transfer process in the current tools
- As I worked on the consult and transfer functionality, I defined the flows and scenarios, sketched, whiteboarded, wireframed and reviewed the designs with the innovation pod on a daily basis
- After critiquing the designs with the team, I iterated more and presented my iterations the next day to the team
- I conducted a usability study to test initial virtual assistant concepts with the agents
- I used Framer to demonstrate micro-interactions
- I started using the brand new visual UI Kit for my visual design styles
- Throughout the project, I remained a strong advocate for adding a human touch to the virtual assistant functionality
Sketches
Sketches
Original design 1
Original design 1
Original design 1
Before I joined the team, they had explored two ways to engage with an agent. This virtual assistant model assists the agent via a floating bot icon in the bottom right. This was the model I started to integrate escalation / transfer work into.
Before I joined the team, they had explored two ways to engage with an agent. This virtual assistant model assists the agent via a floating bot icon in the bottom right. This was the model I started to integrate escalations/transfer work into.
Before I joined the team, they had explored two ways to engage with an agent. This virtual assistant model below assists the agent via a floating bot icon in the bottom right. This was the model I started to integrate escalations/transfer work into.
Before I joined the team, they had explored two ways to engage with an agent. This virtual assistant model below assists the agent via a floating bot icon in the bottom right. This was the model I started to integrate escalations/transfer work into.
Original design 2
Original design 2
Original design 2
This second model showcases an inline prompt or a template from the virtual assistant that the agents can use to speed up communication with the customer.
This second model showcases an inline prompt or template from the virtual assistant that the agents can use to speed up communication with the customer.
This second model showcases an inline prompt or template from the virtual assistant that the agents can use to speed up communication with the customer.
This second model below showcases an inline prompt or template from the virtual assistant that the agents can use to speed up communication with the customer.
This second model below showcases an inline prompt or template from the virtual assistant that the agents can use to speed up communication with the customer.
Escalation / transfer flows
Escalation / transfer flows
Escalation / transfer flows
Escalation / transfer flows
I put together flow diagrams to illustrate the old and the new escalation and transfer process and journey map. Having diagrams always helps when talking with the stakeholders and cross-functional teams. These diagrams also helped when coming up with the use case scenarios later in the process.
I put together flow diagrams to illustrate the old and the new escalation and transfer process and journey map. Having diagrams always helps when talking with the stakeholders and cross-functional teams. These diagrams also helped when coming up with the use case scenarios later in the process.
I put together flow diagrams to illustrate the old and the new escalation and transfer process and journey map. Having diagrams always helps when talking with the stakeholders and cross-functional teams. These diagrams also helped when coming up with the use case scenarios later in the process.
I put together flow diagrams to illustrate the old and the new escalation and transfer process and journey map. Having diagrams always helps when talking with the stakeholders and cross-functional teams. These diagrams also helped when coming up with the use case scenarios later in the process.
I put together flow diagrams to illustrate the old and the new escalation and transfer process and journey map. Having diagrams always helps when talking with the stakeholders and cross-functional teams. These diagrams also helped when coming up with the use case scenarios later in the process.
Expanding the original inline virtual assistant idea
Expanding the original inline virtual assistant idea
Expanding the original inline virtual assistant idea
I built on the second initial idea of using inline prompts by allowing agents to engage the virtual assistant with @bot tag in the livechat box. I also saw a big value in giving agents a prompt regarding the next step.
I built on the second initial idea of using inline prompts by allowing agents to engage the virtual assistant with @bot tag in the livechat box.
I built on the second initial idea of using inline prompts by allowing agents to engage the virtual assistant with @bot tag in the livechat box.
I built on the second initial idea of using inline prompts by allowing agents to engage the virtual assistant with @bot tag in the livechat box.
I built on the second initial idea of using inline prompts by allowing agents to engage the virtual assistant with @bot tag in the livechat box.
Introducing a new virtual assistant design
Introducing a new virtual assistant design
Introducing a new virtual assistant design
Introducing a new virtual assistant design
I also came up with this design where I used small itinerary snippets and guided step-by-step workflows that are revealed incrementally to the agents in their conversation with a customer.
During my time at Expedia Group, I started realizing that we need to offer agents an opportunity to exercise enough industry and autonomy for them to enjoy their jobs. At the same time, using templated answers feels very impersonal and cold.
This design exploration allows for carefully crafted agent responses while receiving the right amount of information about the booking and the next best step to take.
I also came up with this design where I used small itinerary snippets and guided step-by-step workflows that are revealed incrementally to aid agents in their conversation with a customer.
During my time at Expedia Group, I started realizing that we need to offer agents an opportunity to exercise enough industry and autonomy for them to enjoy their jobs. At the same time, using templated answers feels very impersonal and cold.
This design exploration allows for carefully crafted agent responses while receiving the right amount of information about the booking and the next best step to take.
I also came up with this design where I used small itinerary snippets and guided step-by-step workflows that are revealed incrementally to aid agents in their conversation with a customer.
During my time at Expedia Group, I started realizing that we need to offer agents an opportunity to exercise enough industry and autonomy for them to enjoy their jobs. At the same time, using templated answers feels very impersonal and cold.
This design exploration allows for carefully crafted agent responses while receiving the right amount of information about the booking and the next best step to take.
I also came up with this design below where I used small itinerary snippets and guided step-by-step workflows that are revealed incrementally to aid agents in their conversation with a customer.
During my time at Expedia Group, I started realizing that we need to offer agents an opportunity to exercise enough industry and autonomy for them to enjoy their jobs. At the same time, using templated answers feels very impersonal and cold.
This design exploration allows for carefully crafted agent responses while receiving the right amount of information about the booking and the next best step to take.
I also came up with this design below where I used small itinerary snippets and guided step-by-step workflows that are revealed incrementally to aid agents in their conversation with a customer.
During my time at Expedia Group, I started realizing that we need to offer agents an opportunity to exercise enough industry and autonomy for them to enjoy their jobs. At the same time, using templated answers feels very impersonal and cold.
This design exploration allows for carefully crafted agent responses while receiving the right amount of information about the booking and the next best step to take.
Three mid-fidelity chatbot interaction models for the usability testing
Three mid-fidelity chatbot interaction models for the usability testing
Three mid-fidelity chatbot interaction models for the usability testing
Three mid-fidelity chatbot interaction models for the usability testing
Three mid-fidelity chatbot interaction models for the usability testing
These are the three interactive Sketch prototypes I put together for the usability study that I conducted with five chat agents from El Salvador. I integrated my new ideas while leveraging the work done before me by the team.
These are the three interactive Sketch prototypes I put together for the usability study that I conducted with five chat agents in El Salvador. I integrated my new ideas while leveraging the work done before me by the team.
These are the three interactive Sketch prototypes I put together for the usability study that I conducted with five chat agents in El Salvador. I integrated my new ideas while leveraging the work done before me by the team.
These are the three interactive Sketch prototypes I put together for the usability study that I conducted with five chat agents in El Salvador. I integrated my new ideas while leveraging the work done before me by the team.
Usability study goals
Usability study goals
The goal of this study was to evaluate three virtual assistant (bot) interaction models and transfer / consult functionality so that I could identify the design that resonated most with the our call center agents.
I also wanted:
to gather user thoughts on giving their own responses vs. using templates
to probe into how can agents make a difference when helping a customer
to discuss their common pain points
The goal of this study was to evaluate three virtual assistant (bot) interaction model prototypes so that I could identify the design that resonated most with the users.
I also wanted:
- to gather user thoughts on giving their own responses vs. using templates
- to probe into how can they make a difference when helping a customer, and how would they like to make a difference
- to discuss what are their common pain points
The goal of this study was to evaluate three virtual assistant (bot) interaction model prototypes so that I could identify the design that resonated most with the users (our call center agents).
I also wanted:
- to gather user thoughts on giving their own responses vs. using templates
- to probe into how can they make a difference when helping a customer, and how would they like to make a difference
- to discuss what are their common pain points
The goal of this study was to evaluate three virtual assistant (bot) interaction model prototypes so that I could identify the design that resonated most with the users (our call center agents).
I also wanted:
- to gather user thoughts on giving their own responses vs. using templates
- to probe into how can they make a difference when helping a customer, and how would they like to make a difference
- to discuss what are their common pain points
The goal of this study was to evaluate three virtual assistant (bot) interaction model prototypes so that I could identify the design that resonated most with the users (our call center agents).
I also wanted:
- to gather user thoughts on giving their own responses vs. using templates
- to probe into how can they make a difference when helping a customer, and how would they like to make a difference
- to discuss what are their common pain points
User research metrics
User research metrics
Discoverability – do agents discover the virtual assistant (bot)? Do they discover bot’s prompts and engage with them easily? Do they discover bot's messages? Do they discover where and how to ask for bot’s help?
Comprehension – do agents understand which messages came from the virtual assistant and which were from the other agents? Do they understand how to receive from and send messages to the virtual assistant? Do agents understand the transfer / consult options?
Usability – can agents complete the virtual assistant to agent and agent to agent workflow while talking to the customer? Is there any confusion or frustration around the experience? Is there a preference for a certain interaction model?
Post-test virtual assistant models rating
- Discoverability – do agents discover the virtual assistant (bot)? Do they discover bot’s prompts and engage with them easily? Do they discover bot's messages? Do they discover where and how to ask for bot’s help?
- Comprehension – do they understand which messages came from the virtual assistant and which were from the other agents? Do they understand how to receive from and send messages to the virtual assistant? Do agents understand the transfer/consult options?
- Usability – can agents complete the virtual assistant to agent and agent to agent workflow while talking to the customer? Is there any confusion or frustration around the experience? Is there a preference for a certain interaction model?
- Post-test virtual assistant models rating
- Discoverability – do agents discover the virtual assistant (bot)? Do they discover bot’s prompts and engage with them easily? Do they discover bot's messages? Do they discover where and how to ask for bot’s help?
- Comprehension – do they understand which messages came from the virtual assistant and which were from the other agents? Do they understand how to receive from and send messages to the virtual assistant? Do agents understand the transfer/consult options?
- Usability – can agents complete the virtual assistant to agent and agent to agent workflow while talking to the customer? Is there any confusion or frustration around the experience? Is there a preference for a certain interaction model?
- Post-test virtual assistant models rating
- Discoverability – do agents discover the virtual assistant (bot)? Do they discover bot’s prompts and engage with them easily? Do they discover bot's messages? Do they discover where and how to ask for bot’s help?
- Comprehension – do they understand which messages came from the virtual assistant and which were from the other agents? Do they understand how to receive from and send messages to the virtual assistant? Do agents understand the transfer/consult options?
- Usability – can agents complete the virtual assistant to agent and agent to agent workflow while talking to the customer? Is there any confusion or frustration around the experience? Is there a preference for a certain interaction model?
- Post-test virtual assistant models rating
- Discoverability – do agents discover the virtual assistant (bot)? Do they discover bot’s prompts and engage with them easily? Do they discover bot's messages? Do they discover where and how to ask for bot’s help?
- Comprehension – do agents understand which messages came from the virtual assistant and which were from the other agents? Do they understand how to receive from and send messages to the virtual assistant? Do agents understand the transfer/consult options?
- Usability – can agents complete the virtual assistant to agent and agent to agent workflow while talking to the customer? Is there any confusion or frustration around the experience? Is there a preference for a certain interaction model?
- Post-test virtual assistant models rating
Main user research findings
Main user research findings
Participants were able to successfully engage with the floating and the inline virtual assistant interaction models
Participants found @bot communication from the livechat box to ask virtual assistant a question potentially error-prone
Participants liked getting templates
Participants liked the option to personalize the presented message templates
Participants preferred the floating virtual assistant to the inline option
Participants liked the itinerary snippets model – it helped them to quickly gain context into customer itinerary details
Participants were able to successfully engage with the floating and the inline virtual assistant interaction models
Participants found @bot communication from the livechat box to ask virtual assistant a question potentially error-prone
Participants liked getting templates
Participants liked the option to personalize the presented message templates
Participants preferred the floating virtual assistant to the inline option
Participants liked the itinerary snippets model – it helped them to quickly gain context into customer itinerary details
- Participants were able to successfully engage with the floating and the inline virtual assistant interaction models
- Participants found @bot communication from the livechat box to ask virtual assistant a question potentially error-prone
- Participants liked getting templates
- Participants liked the option to personalize the presented message templates
- Participants preferred the floating virtual assistant to the inline option
- Participants liked the itinerary snippets model – it helped them to quickly gain context into customer itinerary details
Participants were able to successfully engage with the floating and the inline virtual assistant interaction models
Participants found @bot communication from the livechat box to ask virtual assistant a question potentially error-prone
Participants liked getting templates
Participants liked the option to personalize the presented message templates
Participants preferred the floating virtual assistant to the inline option
Participants liked the itinerary snippets model – it helped them to quickly gain context into customer itinerary details
- Participants were able to successfully engage with the floating and the inline virtual assistant interaction models
- Participants found @bot communication from the livechat box to ask virtual assistant a question potentially error-prone
- Participants liked getting templates
- Participants liked the option to personalize the presented message templates
- Participants preferred the floating virtual assistant to the inline option
- Participants liked the itinerary snippets model – it helped them to quickly gain context into customer itinerary details
Escalation / transfer options explored
Escalation / transfer options explored
Escalation / transfer options explored
Escalation / transfer options explored
Escalation / transfer options explored
Below are different early variants of the escalation / transfer design explorations
Below are different early variants of the escalation/transfer design explorations
Below are different early variants of the escalation/transfer design explorations
Below are different early variants of the escalation/transfer design explorations
High fidelity Consult option 1
High fidelity consult option 1
High fidelity Consult option 2
High fidelity consult option 2
Proposed final design
Proposed final design
Proposed final design
Proposed final design
My proposed final design includes a floating virtual assistant in the right-hand bottom corner, an itinerary snippet sent to the customer on the virtual assistant's prompt, and an agent to agent chat on the right-hand side that got triggered when the agent chose 'transfer' option.
The agent can get help from the virtual assistant at any time in the process!
My propoesed final design includes a floating virtual assistant in the right-hand bottom corner, an itinerary snippet sent to the customer on the virtual assistant's prompt, and an agent to agent chat on the right-hand side that got triggered when the agent chose 'transfer' option.
The agent can get help from the virtual assistant at any time in the process!
My propoesed final design includes a floating virtual assistant in the right-hand bottom corner, an itinerary snippet sent to the customer on the virtual assistant's prompt, and an agent to agent chat on the right-hand side that got triggered when the agent chose 'transfer' option.
The agent can get help from the virtual assistant at any time in the process!
My propoesed final design includes a floating virtual assistant in the right-hand bottom corner, an itinerary snippet sent to the customer on the virtual assistant's prompt, and an agent to agent chat on the right-hand side that got triggered when the agent chose 'transfer' option.
The agent can get help from the virtual assistant at any time in the process!
My propoesed final design includes a floating virtual assistant in the right-hand bottom corner, an itinerary snippet sent to the customer on the virtual assistant's prompt, and an agent to agent chat on the right-hand side that got triggered when the agent chose 'transfer' option.
The agent can get help from the virtual assistant at any time in the process!
Micro-interactions design in Framer
Micro-interactions design in Framer
Micro-interactions design in Framer
Reflections and final thoughts
This project was one of the best projects I got to work on while at Expedia! It was scrappy and fast!
It was truly team-driven, agile and collaborative. It was amazing to have developers be part of my design reviews on a daily basis, whiteboard together as a team and design scenarios together.
This project shows how design and research can be utilized successfully to gauge initial user feedback before investing in an expensive full-on development cycle and a product that might not resonate with the user!
This project was one of the best projects I got to work on while at Expedia! It was scrappy and fast!
It was truly team-driven, agile and collaborative. It was amazing to have developers be part of my design reviews on a daily basis, whiteboard together as a team and design scenarios together.
This project shows how design and research can be utilized successfully to gauge initial user feedback before investing in an expensive full-on development cycle and a product that might not resonate with the user!
This project was one of the best projects I got to work on while at Expedia! It was scrappy and fast!
It was truly team-driven, agile and collaborative. It was amazing to have developers be part of my design reviews on a daily basis, whiteboard together as a team and design scenarios together.
This project shows how design and research can be utilized successfully to gauge initial user feedback before investing in an expensive full-on development cycle and a product that might not resonate with the user!
This project was one of the best projects I got to work on while at Expedia! It was scrappy and fast!
It was truly team-driven, agile and collaborative. It was amazing to have developers be part of my design reviews on a daily basis, whiteboard together as a team and design scenarios together.
This project shows how design and research can be utilized successfully to gauge initial user feedback before investing in an expensive full-on development cycle and a product that might not resonate with the user!
This project was one of the best projects I got to work on while at Expedia! It was scrappy and fast!
It was truly team-driven, agile and collaborative. It was amazing to have developers be part of my design reviews on a daily basis, whiteboard together as a team and design scenarios together.
This project shows how design and research can be utilized successfully to gauge initial user feedback before investing in an expensive full-on development cycle and a product that might not resonate with the user!