Smartee: AI Chatbot

Timeline

6 months

Role

UX designer - visuals, user testing, research, user flow, prototyping

Team

7 Designers, 1 Engineer, 1 Design Manager, 1 PM

Context


Context

The Joy Education Foundation (JEF) is a virtual reading clinic that leverages tech to connect families with literacy support.

With a primary focus on low-income communities, this program focuses on involving family members and supporting a diverse range of needs, with future plans to expand.

Our team’s role is to investigate user engagement features that will bridge the gap between sustainability and consistent collective action.

What I did


During my time with Joy Education…

I had the opportunity to collaborate with six designers, design manager, and our stakeholders to bring the AI chatbot to life. We scoped out the problem space, conducted in-depth research, and built and intuitive rule-based chatbot that ensures users have the guidance they need to thrive.

I designed how users might interact with the chatbot in different user flows, addressing user needs and implementing user feedback by simplifying the process for users to successfully locate different resources.

What I did

So, here's the problem:

Families are getting lost in shuffle when accessing literacy support they need for their children.

Potential users and current parents looking for learning resources struggle finding resources quickly and experience difficulty navigating JEF's current site. Viewing reading tools, getting involved and getting help isn't as seamless as users need it to be.

✨ Our solution

To enhance access to JEF services, they needed a virtual front door, guiding parents, family members, and community partners from their initial interest to the exact tools and resources that they need.

The solution was a rule-based AI chatbot designed to ensure users reach their destination effortlessly, without getting lost in the process—whether it's finding a digital screener or scheduling a tutoring session.

To enhance access to JEF services, they needed a virtual front door, guiding parents, family members, and community partners from their initial interest to the exact tools and resources that they need.

The solution was a rule-based AI chatbot designed to ensure users reach their destination effortlessly, without getting lost in the process—whether it's finding a digital screener or scheduling a tutoring session.

To enhance access to JEF services, they needed a virtual front door, guiding parents, family members, and community partners from their initial interest to the exact tools and resources that they need.

The solution was a rule-based AI chatbot designed to ensure users reach their destination effortlessly, without getting lost in the process—whether it's finding a digital screener or scheduling a tutoring session.

Initial research

Exploring difficulties with the site

I created some first impressions/notes about my experience on the site before our initial conversation with our clients at JEF.

Difficulties with the current site.


The client shared their aim to create a tech-driven support system that connects families in low-income communities—particularly students of color in areas like Virginia and Detroit—with reading assistance. This program will focus on involving family members and supporting a diverse range of needs, with future plans to expand. Our team’s role is to investigate user engagement features that will bridge the gap between sustainability and consistent collective action.

Difficulties with the current site.


The client shared their aim to create a tech-driven support system that connects families in low-income communities—particularly students of color in areas like Virginia and Detroit—with reading assistance. This program will focus on involving family members and supporting a diverse range of needs, with future plans to expand. Our team’s role is to investigate user engagement features that will bridge the gap between sustainability and consistent collective action.

My first impressions of home, about and login screens

We wanted to understand user perspectives on chatbots and JEF. Through a series of 12 surveys participants, user interviews and market research, we discovered…

📈 75%

of parents prioritize the trustworthiness of the organization when following through with an online learning program.

with

33%

of those parents saying that the reputation and org. mission also influences their decision.

⌛ 73%

of parents agree that a major challenge they face when accessing educational resources is lack of time.

with

46%

of these parents also citing a limited availability of resources to support their child.

🥴 80%

of parents hold mixed concerns about using AI for learning, feeling unsure of what that might look like

with

~50%

of all parents preferred speaking with a live rep over a chatbot.

Here's what our participants said:

"I liked that chatbots are available 24/7."

"I prefer live representatives since AI chatbots are just really expensive and new."

"I prefer chatbots IF it has the answer I need."

"When talking to a live agent, you have the freedom to elaborate more on a personalized issue."

"AI chatbots are quick to answer but lacks emotion."

"I would like the AI chatbot to make my work easier."

"When talking to a live agent, you have the freedom to elaborate more on a personalized issue."

"AI chatbots are quick to answer but lacks emotion."

"I prefer chatbots IF it has the answer I need."

"I would like the AI chatbot to make my work easier."

User insights

Users have mixed perspectives on AI chatbots, but express that it can be useful if certain criteria are met.

We generated user research insights from our user conversations and survey findings and identified the following issues as the major problem statements with chatbots:

Lacks quick and detailed responses

Busy families struggle with finding resources they trust quickly. This is a huge issue for parents who don't have the ability to keep exploring the site to find out on their own.

Lacks emotion & trust-building capabilities

Speaking with an uninviting support bot that doesn't fit your needs creates friction during engagement, leading users to exit conversations.

Lacks prompt technical support

With more complex questions, chatbots are limited with their responses. Users with experience with other bots noted there isn't always a way to receive immediately help promptly.

Establishing user personas

Crafting 3 distinct personas to mimic the core users at Joy Education

We crafted multiple levels of user personas that engage with the JEF site and might require their services so that we may keep these personas as a guide for our user needs/frustrations when designing. These include:

  • Primary user - Current parent

  • Secondary user - Inquiring parent

  • Tertiary user - Inquiring tutor

Mapping out and visualizing touchpoints for primary users: parents/guardians

I created a journey map to visualize interactions, frustrations and opportunities users might have, allowing me to make/test assumptions on how I might proceed with designing the chatbot.

Mapping out and visualizing touchpoints for primary users: parents/guardians

I created a journey map to visualize interactions, frustrations and opportunities users might have, allowing me to make/test assumptions on how I might proceed with designing the chatbot.

Click to enlarge

🤖 Now that we understood our user perspectives, we initiated establishing our chatbot design by creating a guideline for Smartee's chatbot persona.

This was necessary to save time so that the entire design team understood Smartee at his core.

We wanted to make sure the user was speaking to the JEF brand in a meaningful way since the gateway to our support hub was through Smartee.

Doing this helped save time when creating Smartee's dialogue flows since his purpose and tone were already established.

Improves user experience

Builds trust + engagement

Builds trust + engageme-nt

Follows brand identity

We drafted a framework and developed Smartee's language for how his dialogue flow might look and feel

Before drafting the conversation design, we created a general template that included major types of user queries based on their needs.

Skeleton of our dialogue flow. Click to enlarge.

We designed flow by flow

Focusing on:

  • Not leaving dead ends

  • Using friendly language

  • Utilizing "I" statements

  • Making sure bot responses helped provide specific answers or direct answers to queries

Reiterating our dialogue flow. Click to enlarge.

Lofi designs

Brainstorming options, greetings and dialogue screens

To explore different ways users might interact with the bot, we designed how conversation might appear to users upon opening the chat and selecting options, sending messages.

I experimented with FAQ forum styles, horizontal/vertical suggestions, conversational designs with text bubbles and bottom/top navs.

Although an FAQ/forum style offered a way for users to find their topics/issues from the get-go, it didn't align with our chatbot persona, feeling more technical rather than welcoming and increased choice overload.

Super early explorations. Click to enlarge.

Iterating designs for Smartee greeting screen. Click to enlarge.

Iterating designs for Smartee greeting screen. Click to enlarge.

Iterating designs for parent inquiries. Click to enlarge.

Iterating designs for parent inquiries. Click to enlarge.

Iterating designs for settings menu. Click to enlarge.

Iterating designs for settings menu. Click to enlarge.

Feedback based changes

Participants had a lot to say about finding information on the chatbot.

Despite being limited to one round of user testing during the hifi phase, I was still able to gain a lot of valuable feedback from our 4 participants.

These participants gave a lot of specific insight to how information is found on the chatbot and the look/feel of the chatbot… which ultimately led to these changes…

Users needed it to be easier to reach live agents quickly.

I noticed most of the participants struggled completing tasks related to locating a live support agent, meaning it might have been too convoluted to find within the flow even when they were freely exploring the flow. To improve this, we introduced adding a live agent much sooner in the flow.

Lines marked in 🔴 are new changes, making it much faster to connect with a live agent

Users didn't explore functionalities they weren't familiar with.

Despite wanting to explore the prototype further and adjust Smartee's settings, almost all participants did not interact with our settings menu. They admitted to being unsure as to what that icon was and were thus unsuccessful with task completion.

Inquiring users needed guidance, not just links.

For flows that required more info on how to proceed, participants spent a while reading through the dialogue and wondered questions like 'what might this volunteer role even look like?'. We noticed that inquiring users needed to be walked through when making a big time commitment such as becoming a reading tutor for children. We created a more detailed walkthrough and edited our content to help users reach their end goal more seamlessly.

New vs old tutoring inquiry flow. Click to enlarge.

Project handoff

After polishing our designs, we moved onto preparing documentation and handoff, making note of any specific interaction or alt-text.

Results

Our clients and the families at JEF are all excited for Smartee to go live! When we tested our participants, we measured the participants' notes from a 1-10 rating of the usefulness of Smartee as well as a 1-5 rating on how Smartee's design.

100%

of our users felt Smartee empowered them to find educational support quickly

On average users gave Smartee a usefulness rating of

8/10

On average users gave Smartee's visual design a rating of

4/5

📌 Reflections

I got to meet Demetrius and Zoe and really be involved with the JEF process and mission. I loved being a part of a bright and curious team that helped challenge and push the design forward.

I learned to design for different needs/principles an AI based design requires such as building to increase engagement and trust. Moving forward, I look to continue developing Smartee and be involved with maintaining him after his launch with the JEF team.

Some important learnings:

  • Consistency is so important. Keeping text length short, informative and language similar was a good challenge among the design team

  • Designing for accessibility: making sure all colors and text are visible, and giving users the option to tailor the chatbot towards them is important

I don't own a webcam but thank you to everyone on the Joy Education Foundation Team for the awesome summer! 💖

Thank you for visiting! ☺

Let’s Be Friends!

Made by me with 💗 2023

Thank you for visiting! ☺

Let’s Be Friends!

Email

LinkedIn

Socials

Twitter

Instagram

Medium

Dribbble

Made by me with 💗 2023

Thank you for visiting! ☺

Let’s Be Friends!

Email

LinkedIn

Socials

Twitter

Instagram

Medium

Dribbble

Made by me with 💗 2023