OVERVIEW

Helping students navigate real-life conversations, understand corrections, and explore new ways to express themselves.

Helping students navigate real-life conversations, understand corrections, and explore new ways to express themselves.

My contribution

I worked across Edsy’s AI features and learning activities, partnering with product to understand business requirements and user feedback, then collaborating closely with the design team to shape each experience.

My contribution covered interaction flows, interface design, copy, and communication. I helped explore design alternatives and guided revisions as we learned from usability testing.

This case study focuses on three AI experiences: two-way conversation, grammar check, and paraphrasing.

The challenge

Making AI useful in a learning moment

Edsy uses AI to support English practice from playing a character in a conversation to evaluating a student’s answer.

Each capability introduced a different interaction challenge. In conversation, students needed to understand their role and know when to respond. In grammar check, they needed to connect corrections to their own answer. In paraphrasing, they needed to compare alternative expressions and decide what to apply.

The design work connected these capabilities to clear actions students could follow.

01 — Two-way conversation

Preparing students to speak with an AI partner

Preparing students to speak with an AI partner

The product team wanted a new activity where students could practice listening and responding in everyday situations, such as greeting a visitor or visiting a dentist.

AI served two roles: a character participating in the scenario and a system evaluating answers and providing feedback.

Working with the design team, I explored layouts and interaction flows for introducing the scenario, distinguishing the student from the AI character, and guiding the exchange.

The product team wanted a new activity where students could practice listening and responding in everyday situations, such as greeting a visitor or visiting a dentist.

AI served two roles: a character participating in the scenario and a system evaluating answers and providing feedback.

Working with the design team, I explored layouts and interaction flows for introducing the scenario, distinguishing the student from the AI character, and guiding the exchange.

The product team wanted a new activity where students could practice listening and responding in everyday situations, such as greeting a visitor or visiting a dentist.

AI served two roles: a character participating in the scenario and a system evaluating answers and providing feedback.

Working with the design team, I explored layouts and interaction flows for introducing the scenario, distinguishing the student from the AI character, and guiding the exchange.

Testing exposed gaps in guidance

Testing exposed gaps in guidance

We tested a revised prototype with users and gathered feedback on the experience.

The findings highlighted several areas to address:

We tested a revised prototype with users and gathered feedback on the experience.

The findings highlighted several areas to address:

We tested a revised prototype with users and gathered feedback on the experience. The findings highlighted several areas to address:

The findings highlighted several areas to address:

The findings highlighted several areas to address:

  • Role clarity: Participants needed more preparation to understand their part in the scenario.

  • Instructions: What to do was confusing, particularly in the first mission.

  • Feedback: Responses felt too generic and needed more detail.

  • Vocabulary support: The helper was useful but could feel like memorization pressure.

  • These findings shifted attention toward the guidance surrounding each interaction: how students prepared, understood their task, and interpreted the response.

  • Role clarity: Participants needed more preparation to understand their part in the scenario.

  • Instructions: What to do was confusing, particularly in the first mission.

  • Feedback: Responses felt too generic and needed more detail.

  • Vocabulary support: The helper was useful but could feel like memorization pressure.


    These findings shifted attention toward the guidance surrounding each interaction: how students prepared, understood their task, and interpreted the response.


  • Role clarity: Participants needed more preparation to understand their part in the scenario.

  • Instructions: What to do was confusing, particularly in the first mission.

  • Feedback: Responses felt too generic and needed more detail.

  • Vocabulary support: The helper was useful but could feel like memorization pressure.

  • These findings shifted attention toward the guidance surrounding each interaction: how students prepared, understood their task, and interpreted the response.

What this example demonstrates

What this example demonstrates

The conversation feature required coordination between scenario setup, turn-taking, speaking, and evaluation. My work with the team addressed how those parts fit together as a student experience.

02 — Grammar check

Making AI corrections visible in the student’s own answer

Making AI corrections visible in the student’s own answer

The grammar-check experience uses AI to identify grammatical errors and suggest corrections to a student’s answer. The interface keeps the original response visible, highlights identified errors, and guides students through the corrections one at a time.

An answer could contain several grammatical errors. The design needed to make those errors visible while helping students understand each proposed change.

I proposed a sequential review flow: highlight the identified errors in the answer, then walk students through corrections one at a time. This preserved the context of the full response while giving each correction a clear focus.

The approach prioritizes guided review over speed. It introduces more steps, but was intended to make multiple corrections easier to process. Backward and forward navigation lets students revisit a change before continuing.

Working with the design team, I guided how text highlights, original-and-replacement wording, and correction panels connected the AI’s suggestions to the student’s answer.

Rather than presenting only a corrected answer, the flow exposes what the AI proposes changing. This gives students an opportunity to review the differences as they move through their response.

03 — Paraphrasing

Supporting exploration without taking over the student’s answer

Supporting exploration without taking over the student’s answer

We introduced paraphrasing as guided support for exploring different ways to express an idea. The design needed to make AI-generated alternatives accessible while keeping the student involved in choosing what to change.

The flow connects a highlighted passage to suggested alternatives. Students can compare wording, access audio and translation, and explicitly apply a selection. This creates an opportunity to review the AI’s suggestion before it changes their answer.

I guided the experience toward familiar interaction patterns, reusing components from other learning activities. Text highlights, panels, and action buttons provided continuity as we introduced a new capability. The intent was to let students focus on the language rather than learn another interface.

This provides a foundation for learning scaffolding. Further validation would be needed to understand whether students can use what they encounter here to paraphrase independently.

Reflections

The two-way conversation tests showed how much guidance students needed around an AI interaction. Role explanations, mission prompts, and feedback wording were central parts of the learning experience.

Across the work, my focus was on helping the team turn AI capabilities into understandable interactions: what students do, how the system responds, and what students can do next.

The two-way conversation tests showed how much guidance students needed around an AI interaction. Role explanations, mission prompts, and feedback wording were central parts of the learning experience.

Across the work, my focus was on helping the team turn AI capabilities into understandable interactions: what students do, how the system responds, and what students can do next.

EDSY ✺ SHIPPED 2026 ✺ Web App

Designing AI-powered English practice for confident speakers


FOCUS

AI Integration, Interaction design, UI, copy, design direction, usability-driven iteration

YEAR

2024 - Present

TEAM

1 Junior designer, 1 PM, 3 Developers

ROLE

Design Consultant

FOCUS

AI Integration, Interaction design, UI, copy, design direction, usability-driven iteration

YEAR

2024 - Present

TEAM

1 Junior designer, 1 PM, 3 Developers

ROLE

Design Consultant