.png)
Overview
Kuri is a personal assistant powered by artificial intelligence. Inspired by OpenAI's ChatGPT, our team delve into the goals that motivate current behaviors when interacting with conversional AI tools and improved their journeys. Kuri users are able to save, search and store chats with a personal assistant that learns from their patterns and behavior to improve their in-app experience.
My role: As Team Lead, I was responsible for ensuring that our team solved our initial problem through project management, facilitating brainstorming and design sessions, making final decision decisions and liaising with our professors. I also was responsible for our chat screen design and features.
The Team
Abby Brams Andrea Romero Diaz Jhordan John (Team Lead) Marconi Douetts Kinsey Still
Skills
Design system Product design Product strategy UI design User research Usability testing
Timeline
Feb - April 2023
.jpg)
Process
Introduction
Laying out the context
Problem Statement: It is difficult for university students to share experiences and connect with other students.
Methodology: We followed Alan Cooper's Goal-Directed Design methodology which emphasizes the user's goals through 5 phases of design: research, modeling, requirements, framework, and refinement.
Solution: An app that allows students to share their university experience, connect with other students, and source relevant university information.
My team designed an iOS mobile app called ComUni. The name itself bridges two our main ideas-- community and university; as we aim to foster and build community life within the university context.
SECONDARY Research
Understanding the domain
To gain an understanding of our potential users and how they may use our product, I led my team through qualitative research, competitor analyses, and user interviews. We began with our secondary research, asking one simple question:
How do students currently find and share information?
The goal of our research was to identify the different methods students currently use to source and share information related to their university. As students ourselves, we had various assumptions about what the findings of existing studies could look like, but our initial research still gave us some surprising results:
- 70% of students preferred to use an app to access school related information while only 22% preferred their student email (Hesseling, 2019).
- 73% of students feel most comfortable going to other students to share their perspectives on topics of importance (Ezarik, 2021)
- 98% of college-aged students use some sort of social media platform to access information on a daily basis (Dossett, 2020)
Analyzing our competitors
We then took time to analyze 4 major products that provide students with campus-related related information. We ultimately found that many web and mobile services did not prioritize verifying users or ensuring content was relevant to the user's university, resulting in a cookie-cutter, surface-level experience.
USER Research
We interviewed five college students
Our next step was to speak directly with our potential users -- college students. We conducted our interviews virtually due to the varying locations of our participants to understand how they found and shared information and how an app like ComUni would align with their goals. I facilitated two of these interviews.
Our participants provided us with key information that gave us much needed insight:
"I usually check my school website but I wish the information was organized differently" - Junior, University of Central Florida
"As an international student, it was hard for me to find information about my university from current students" - Sophomore, New York University
- All participants preferred to receive information about campus life directly from students.
- Participants often visited social media sites such as YouTube, Instagram, and Reddit to get first-hand accounts from students at their current or prospective school
- 80% of our participants found that accessing the school website and going through official channels of communication was complicated and tedious.
- Participants wanted campus-specific information on classes, facilities, dorms, campus life and diversity
After each interview, I led my team through affinity mapping sessions to identify patterns in user behavior and goals. This process helped us to visualize patterns according to topic that can later impact product requirements.
MODELING
Personifying our patterns
To ensure that our persona was both valid and relevant, our team identified the most consistent patterns of behaviors, goals, and motivations and crafted our primary persona, Melody Soto:
REQUIREMENTS
Translating user goals to app requirements
At this stage, our focus was transferring the findings from our research into design solutions that support the goals of our users. From the goals and subsequent pain points that our interviewees expressed, we wrote context scenarios that married their needs to potential solutions that would meet them. These narratives helped us to place our app in the context of the everyday life of our users, encouraging practicality over designer bias.
We then outlined a list of tasks that our users should be able to complete. At this stage, I functioned as a stakeholder to maintain the scope of the project and ensure that our features can have measured outcomes.
- View a dashboard featuring university highlights, selected topics of interests and saved posts
- Locate overall rating of their university
- Create topics for discussion to ask questions or share opinions
- Receive notifications about the most “reliable” posts
- Save posts to view later
- Earn badges by staying active on the app
Frameworks
Ideating and wireframing
Our team was now entering our most intensive stage-- brainstorming the framework for our app with our newly identified app requirements. To do this, we followed Alan Cooper's guidelines for creating keypath and validation scenarios. Our keypath scenario is our main user flow where a user selects a subgroup, views or contributes content, and looks at campus events. Our validation scenarios are any subsequent screens that support our main flow. This processes resulted in our initial lo-fi frames.
Starting with a (design) system
After functioning as a stakeholder to ensure the initial app flow and features met our business outcomes, I guided my team through our high-fidelity design process. Before we began designing, my priority was to design a well-defined design system to ensure our design felt refined and scalable. I played a crucial role in this stage as both a visual designer and the final decision-maker. Our goal was a clean and modern interface that would fit within an educational context.
REFINEMENT
Moving forward with hi-fi designs
The Refinement phase marks the transition from low-fidelity wireframes to a high-fidelity and fully functional prototype. With design guidelines now in place, I assigned my team members different flows to design screens for. We then prototyped our screens and interactions to deliver a hi-fi prototype to be tested.
The Context
There is no doubt that population interest in Artificial Intelligence (AI) and design skyrocketed as OpenAI's ChatGPT set the trend for conversional AI tools. As discussions around how AI would be implemented in different product domains, our team saw this as a challenge to find a problem we identified within existing AI-powered platforms and improve the experience for users were other products fell short.
The Problem
Current AI tools have focused primarily on providing conversational responses to users that fail to meet a user's specific preferences, parameters and unique lifestyle .
The Challenges
Unlike many of my other projects that existed in very developed domains, this project was heavily dependent on software and information that was still being studied and developed, Many of our initial assumptions were based off of products that simply didn't exist. Being new to AI and machine learning ourselves, I encourage my team to go the extra mile and do additional research that could benefit our product. The novelty of our product called for true reliance on research and design thinking and allowed us to tap into speculative design.
The Solution
A more customizable and elevated experience to a language model AI to create a personalized assistant.
Our team’s solution is a web app called Kuri. This Japanese name (pronounced coo-ree) stems from local folklore which alludes to wisdom. Our team found this to be the perfect name for our AI personal assistant whose goal is to know you better than you know yourself.
Kickstarting with our assumptions
For this project, I acted as both a stakeholder and designer as I posed a list of questions to my team that we brainstormed together. I used this opportunity to provide my team with a general framework for our project and then we discussed our assumptions regarding context, potential users and limitations, and other expectations.
Some off our kickoff questions and our initial assumptions:
.png)
Sprint 1: Exploring our existing context through research
To better understand the domain we dedicated a full sprint to research, starting with secondary research. The goal for our secondary research was to better understand tools people go to for answers, the problems they aimed to address and the solution they offered to address them.
Literature Review and Competitive Analysis
At the time of this project, companies like Microsoft and Google had just released plans for BingAI and Bard, their respective AI tools, so solutions that we could actually interact with were limited. But, this did not stop our research. I assigned my team members to 9 different AI tools and search engines and encouraged them to use any material to better understand the product -- from beta products to press release notes.
Ask the Experts!
Our team met with 2 SMEs to understand the technical side of AI and machine learning. As a team leader, I saw the benefit of understanding the backend of what we aspired to design so we could understand the limitations of existing technology and learn more about safety, best practices and other considerations in this space.
User Interviews
Based on assumptions made in our kickoff meeting, we met with 6 potential users after placing them into 4 quadrants. new vs experienced with AI and using AI for work vs personal. This helped us to get a range of different types of users who may be interested in our product. Our goal was to understand their current experience when using tools that provide them with information or assist them with completing tasks, the main functionalities of a personal assistant.
Key questions we asked our participant:
- What are some challenges you have encountered when using search engines, and how have you addressed them?
- Is there a certain conversation tone you prefer to read when provided with information?
- How do you feel about search engines, web browsers, and websites tracking your behavior?
- Based on your current knowledge of AI Chatbots, are there any additional features you wish they had?
- Is being able to “fact-check” important to you?
- Tasks to see how they interacted with their preferred search engine vs ChatGPT
.png)
Affinity Maps
After each interview, I lead my team through affinity mapping sessions where we common insights based on feedback from our users. These sessions allowed our team to discuss key moments after each interview and compare them to other patterns we identified with other participants. We used out affinities to identify different behaviors and goals that would be used to construct our personas.
Sprint 2: Turning our results into requirements
Visual Continuum Matrix
After summarizing each user interview, our team plot each participant across various spectrums to visualize all the patterns we had identified. Each matrix was based on questions we had asked in our interviews. By doing this, our team was able to quickly see clusters that symbolized how similar or different our 6 participants were. These clusters helped us ot identify the leading goal or behavior related to each question asked in our interviews.
Personas
We based our personas on the clusters that were formed on our matrices. To do this, we took note of how often two or more participants were clustered together on a spectrum. We found that majority of our participants wanted to use a personal assistant to complete work-related tasks as efficiently as possible. Other participants were interested in a personal assistant to creatively explore their personal interests.
Sprint 3: Laying out the frameworks for Kuri
In Goal-Directed Design, this phase is known as the Frameworks phase. As the name suggests, our team engaged in discussions to identify our key user flow, main and supporting features and Kuri's brand. We wanted an app that looked futuristic and modern but still felt trustworthy as Kuri heavily relied on user's sharing specific information. We also wanted a two panel layout to allow users to navigate previous chats while still prioritizing their current chat.
Sprint 4: Bringing Kuri to life
As we advanced from frameworks to a more refined design, we established our design system, branding and voice for Kuri.
Sprint 5: Improving Kuri through testing
Kuri was tested by 8 participants, some from our initial user interviews and some first-timers. This combination allowed us to seek feedback from those who may have thought-through expectations of our app and those who are new to our concept entirely. Helping us to improve the necessary complexity Kuri requires while still making it intuitive and easy to use.
Reflection
Outcome: I led my team through to Goal-Directed Design process to deliver an extensive research report, design files, and a final stakeholder presentation. I learned so much wearing many hats throughout this 12 week project:
- A Project of Many 'Firsts'
This was my first large-scale design project as well as my first time leading a team of designers. I learned that I enjoy taking the lead to figure out a problem I have never faced before. - Managing Constraints and Limitations
I learned how to guide my team through the many limitations that comes along with designing within an academic context such as time, access to resources and software. - Wearing Many Hats
Due to the nature of this project, I had to function in many capacities. From stakeholder, to researcher, to designer, I developed so many skills and learned how to manage all my responsibilities to support my team as a design lead (another hat!).