Airwire x ibm

Designing a virtual assistant leveraging conversational AI

Overview

Our client Airwire is setting out to introduce an app-based virtual assistant that goes beyond the typical capabilities of voice command tools that exist in the current market. Leveraging emerging conversational AI technology, “J.AI” is a personal concierge that aims to be personalized to each user with the additional capability of being able to alert on a user’s vehicle diagnostics and provide maintenance notifications right from their phone.

The Goal

With an ambitious task ahead, our goal was to deliver an end-to-end app with the prioritized tangible features uncovered through a design thinking workshop including:

Enable biometric security and access features during the onboarding process to build trust with users and ensure their data and privacy safety.
Design an intuitive user experience that quickly gets the user to use J.AI with the goal of ensuring user adoption during the free trial period.

Introduce Smartcar integration giving users the ability to stay up-to-date with their vehicle’s maintenance and provide current vehicle diagnostics

The Solution

MAKING FRIENDS WITH THE ROBOTS

The idea behind J.AI was to grant our users ultimate power over their interactions, allowing them to dictate their journey. We recognise that the exciting, yet unpredictable phase of evolving technologies, like artificial intelligence, can potentially arouse uneasy feelings. With this in mind, it became imperative for us to establish well-defined parameters for J.AI and ensure its safe and transparent implementation.

The Kickoff: A Week Of Hosting A Discovery Design Thinking Workshop

To steer this project into a successful delivery, we started out by leading a design thinking workshop with our internal team and stakeholders to garner a deep understanding of the current state of the application and the intended outcomes and deliverables. Through this design approach, we collected qualitative and quantitative research data and feedback alongside creating measurable success criteria to ensure we were capturing the vision of J.AI.

Understand

Big Idea Vignette

Hopes & Fears

Competitive Audits

Define

Goal & Objective

Impact/Effort Matrix

Journey Map & User Flows

EMPATHIZE

Persona Profiles

Empathy Mapping

User Interviews

DESIGN

Low Fidelity Design

Hi-Fidelity Mockups

Interactive Prototype

VALIDATE

Stakeholder Demos

UAT

The Big Picture

We began the design thinking workshop with a Big Idea Vignette brainstorming session to imagine a variety of possible ideas and functions for the app. We focused on generating ideas as, as opposed to features.

What we discovered:

J.AI learns via clarifying questions and should ask clarifying questions as early as possible in the conversation.
J.AI creates a base of knowledge/profile for the user b asking clarifying questions.
J.AI acts as a personal helper and concierge.
J.AI predicts behaviors and actions and creates a database of behaviors.The conversation style is personal and personalized.
J.AI can “read between the lines” and understand the true context of the user’s question.
J.AI works to transform from unsupervised to supervised learning.

Hopes & Fears in the age of AI

With a better idea of some larger future visions, we moved on to discuss our larger aspirations as well as the fears that may go along with them.

What does our ideal state look like? What is the North Star? With this in mind, we grouped these callouts into similar clusters to begin to visualize tangible items.

The Consensus:

Easy to understand for all age groups, with special consideration for the older demographic.

Ability to carry on a human-like conversation, and be as individual as possible when speaking to the user.
Users quickly reach the “aha” moment in less than 24 hours, which encourages them to continue using the service.
Responds quickly with as small of a delay as possible.
Users have the ability to configure what J.AI can make suggestions about
“Smart suggestions” offer suggestions based on certain trigger points.
The back end should not host personal information so it is less likely to become compromised by a malicious third party.

Uncovering the competitive edge

Through a thorough competitive analysis of similar virtual assistant platforms, we were able to determine the following gaps and opportunities that exist in this emerging technology to better inform our competitive edge:

Gaps:

Limited accessibility to virtual assistant technology that is exclusive to certain devices. Ex: Siri is only available to Apple users and devices

Virtual assistants are limited in customization and are not personalized to the user
Queries lack consistent accuracy and are limited to search engine databases
Lack of security features to provide ease of mind from personal data leaks and privacy concerns
Responses tend to be robotic in nature and lack empathy
Management services are limited to streaming and music platforms-lack
integration to provide vehicle diagnostics and alerts
Some virtual assistants require external hardware

Opportunities:

J.AI is accessible via the app and not limited to devices
Ability to carry on contextual conversations and handle “context switching”
Leveraging emerging conversational AI technology allows J.AI to be more personalized and learn from the user to better provide tailored responses
Security settings allow the user to have more control over their data and privacy
Third-party integration with Smartcar allows the user to view and manage vehicle maintenance all from the app without requiring external hardware.

Personifying the user

After synthesizing the key tangible items, we focused on identifying the specific personas and associated demographics, behaviors, actions, needs, and pain points to uncover gaps in our experience.

Our research indicated our core users into the following categories:

Busy professionals who need additional support in automating schedule management
Users who are less tech-savvy that need a simple tool to simplify their day-to-day tasks
Users who are more familiar with emerging technology and see the use of having a more robust virtual assistant that adapts to their interest

jOANNA s.

Home maker

Joanna is a stay at home mother to two children with an overflowing schedule that leaves her stressed and overwhelmed trying to be on top of all her daily tasks.

Joanna needs a better way to manage her day-to-day schedule by being able to streamline her tasks such as creating shopping lists, reminders, and knowing ahead of time what the best routes to her destinations are.

Luke B.

Web Designer

Luke is a newly graduated web designer who has relocated to a new city for a career opportunity. Luke struggles to organize and manage his meetings schedule so he can properly allocate time to his clients and work.

Luke needs a tool that will centralize his daily life tasks to help build better time management skills as he adjusts to his big life changes.

Erin D.

Family Therapist

Erin is popular family therapist with her own practice who is often overbooked. Erin needs help managing her schedule but she is not the most tech-savvy user. She is new to using voice-command technology but is easily frustrated when she is not able to get the response back she needs and finds it invaluable at times.

Erin desires an experience where she is able to use this feature tailored to her daily needs.

Empathy Map

Following our persona profiles, we really wanted to understand the visceral experience of our target users using an app like J.AI. We then geared our workshop into exploring this through an empathy mapping exercise to capture these feelings and actions. This critical portion of empathizing allowed us to understand the user’s pain points and needs to ensure a user experience that allowed the user to be in control of their experience and quell concerns about using AI technology.

Key Highlights:

Users are weary of using AI-based tools
66% of users expressed concerning sentiments about their privacy and captured data
With a range of demographics, the tool needs to be intuitive enough to onboard less tech-savvy users, useful enough to simplify the daily life of busy professionals, and entertaining enough to capture the younger demographic raised in the digital era.

Onboarding the user from the get go

Next, we set out to visualize the user experience from the onset by mapping out the customer journey starting at the moment the user first opens the app.

Some of these key topics included:

Discussing whether or not a sign-in would be required immediately.

Deciding when we would ask a user for payment, and how to warn them that they will miss out on crucial features if they do not sign up.
Discussing how J.AI will speak in order to encourage the user to give access to calendars, social accounts, etc. in order to provide the most tailored experience
What does J.AI say when it does not know the answer to a question?

Introducing Conversational AI To The Flow

Along with mapping out the user journey, we created the user flow for the conversational AI with consideration to the overall backend development and deployment of J.AI. This is to ensure how J.AI is to function alongside the customer journey to establish a foundation for a seamless user experience.

Developing The Lo-Fidelity Prototype

The low-fidelity prototype for J.AI was created to conduct usability testing with the initial user flow uncovered during the discovery workshop.

This prototype takes the user from the onset of setting up their user profile, connecting their vehicle with Smartcar integration, and selecting a plan to use the full features of J.AI during the trial period.

Feedback Leads To Design Iterations

A Change In Scope

Through our journey, we unveiled that MVP1 would require a more refined signup process for the user to support the functionality of enabling J.AI to access synced accounts such as calendars, contacts, location services, etc. in order to provide accurate recommendations. This meant, that the user would require an Apple or Google account to start their trial and use J.AI seamlessly.

Refining The Onboarding Process

Initial designs featured a more interactive experience that allowed the user to answer a few questions to better personalize the experience; however, due to user feedback and deadline constraints, we chose to simplify this portion of the user experience into one screen. This eliminated the need for several screens and streamlined the experience for the user.

A More Robust User Experience Emerges In The Hi-fidelity Prototype

Following user and stakeholder feedback, iterations led to an updated user experience that gave users more control over their data and privacy, additional biometric security access options, and a guided onboarding process.

Onboarding, interaction-voice command, speech to text, Smartcar Vehicle Management

Settings, Account Management

SMARTCAR INTEGRATION

Takeaways

What I Learned

Throughout the process of building a virtual AI assistant, we encountered a multitude of unforeseen challenges. Working alongside a conversational designer for the first time added an extra layer of significance to this journey, as it provided an opportunity to delve into a different realm of design I wasn’t as familiar with. The abrupt shift in teams and project scope demanded quick adjustments and on-the-fly design changes to meet tight deadlines. Adaptability became our guiding principle for this project.

Next Steps

After achieving the milestone of MVP1 and establishing a strong and functional foundation for J.AI, the subsequent phases entail crafting a new array of features to elevate J.AI into a more comprehensive and dynamic experience. Furthermore, we will conduct additional user testing and gather feedback to enhance the user experience, which will be incorporated into MVP2. A pivotal upcoming feature involves granting users the ability to personalize the 'personality' of J.AI."

‹ Title Midwest

Hilton ›

Top

Let's Connect

Patricia Benavides