AI Teacher Assistant

AI Teacher Assistant

Designed and launched an AI-powered assistant for teachers to reduce their workload, reaching 200,000+ DAU

Designed and launched an AI-powered assistant for teachers to reduce their workload, reaching 200,000+ DAU

Designed and launched an AI-powered assistant for teachers to reduce their workload, reaching 200,000+ DAU

Timeline

Timeline

3 months, 2024

3 months, 2024

Role

Role

UI/UX Design

UX Research

UI/UX Design

UX Research

Team

Team

Ming Xiao - Product Manager

Qiu Rou - Data Analyst

Chen Yi, Hen Mo, Zha Ban - Developers

Ming Xiao - Product Manager

Qiu Rou - Data Analyst

Chen Yi, Hen Mo, Zha Ban - Developers

User Research β†’ Pain Points

User Research β†’ Pain Points

Uncover teachers daily most time-consuming tasks

Uncover teachers daily most time-consuming tasks

Uncover teachers daily most time-consuming tasks

β‘  Dairy Study πŸ“
β‘  Dairy Study πŸ“

I conducted a 5-day diary study with 8 teachers documenting their workflows on DingTalk. Key findings showed daily time allocation: Lesson Prep (2.2h), Feedback (2h), Teaching (1.5h), Homework/Newsletters (1h), Parent Communication (0.5h).

I conducted a 5-day diary study with 8 teachers documenting their workflows on DingTalk. Key findings showed daily time allocation: Lesson Prep (2.2h), Feedback (2h), Teaching (1.5h), Homework/Newsletters (1h), Parent Communication (0.5h).

β‘‘ User Interview πŸ‘₯
β‘‘ User Interview πŸ‘₯

After identifying teachers’ most burdensome tasks, I interviewed 5 participants from the diary study to uncover key pain points: 1) repetitive lesson prep resources, 2) manual feedback overload, 3) rigid templates, and 4) fragmented workflows across tools.

After identifying teachers’ most burdensome tasks, I interviewed 5 participants from the diary study to uncover key pain points: 1) repetitive lesson prep resources, 2) manual feedback overload, 3) rigid templates, and 4) fragmented workflows across tools.

Digging deeper into their pain points

Digging deeper into their pain points

Digging deeper into their pain points

Design Approach

Design Approach

Mapping pain points to AI-powered solutions that address teachers’ core needs

Mapping pain points to AI-powered solutions that address teachers’ core needs

Mapping pain points to AI-powered solutions that address teachers’ core needs

Solution 1

Solution 1

AIGC Features β€” Tightly integrates AI capabilities within target scenarios, makes it easy to access this feature without disrupting workflows

AIGC Features β€” Tightly integrates AI capabilities within target scenarios, makes it easy to access this feature without disrupting workflows

AIGC Features β€” Tightly integrates AI capabilities within target scenarios, makes it easy to access this feature without disrupting workflows

Generate Homework Content
Generate Homework Content
Generate Homework Content
Write Personal Feedback
Write Personal Feedback
Write Personal Feedback

Solution 1 β†’ Feature 1

Solution 1 β†’ Feature 1

Assessment Auto-Generation β€” Auto-generate flexible, ready-to-use templates tailored to subjects and course materials

Assessment Auto-Generation β€” Auto-generate flexible, ready-to-use templates tailored to subjects and course materials

Assessment Auto-Generation β€” Auto-generate flexible, ready-to-use templates tailored to subjects and course materials

πŸ”„ 1st Iteration - Low Engagement at Entry Point and Friction with Regeneration

πŸ”„ 1st Iteration - Low Engagement at Entry Point and Friction with Regeneration

⚠️ β‘  Friction Point: Low engagement
⚠️ β‘  Friction Point: Low engagement

Click-tracking data shows low engagement - only 64.2% of users clicked the 'Write with AI' button during testing.

Click-tracking data shows low engagement - only 64.2% of users clicked the 'Write with AI' button during testing.

β‘  Design Solution: Place button to a high visibility point
β‘  Design Solution: Place button to a high visibility point

Moved the button to the top-right (a high-priority action zone) to increase discoverability.

Moved the button to the top-right (a high-priority action zone) to increase discoverability.

β‘‘ Design Solution: Added assistant name in button labeling
β‘‘ Design Solution: Added assistant name in button labeling

Renamed the button from β€œAI” to a personalized assistant name, adds personality and user connection.

Renamed the button from β€œAI” to a personalized assistant name, adds personality and user connection.

⚠️ β‘‘ Friction Point: Hard to regenerate results
⚠️ β‘‘ Friction Point: Hard to regenerate results

During user testing, when teachers weren't satisfied with the results, they had to restart the entire workflow.

During user testing, when teachers weren't satisfied with the results, they had to restart the entire workflow.

β‘  Design Solution: Instant regeneration with one click
β‘  Design Solution: Instant regeneration with one click

Added a 'Swap' button to allow instant regeneration of results without re-entering prompts, streamlining iteration and saving extra steps.

Added a 'Swap' button to allow instant regeneration of results without re-entering prompts, streamlining iteration and saving extra steps.

β‘‘ Design Solution: Edit prompts without exiting flow
β‘‘ Design Solution: Edit prompts without exiting flow

Teachers can instantly return to the prompt selection screen (manual input or templates) to refine their prompts without restarting the workflow.

Teachers can instantly return to the prompt selection screen (manual input or templates) to refine their prompts without restarting the workflow.

πŸ”„ 2nd Iteration - Action-driven Labeling Drives Click Rate Up to 84.7%

πŸ”„ 2nd Iteration - Action-driven Labeling Drives Click Rate Up to 84.7%

⚠️ Friction Point: High drop-off rate for sending prompt
⚠️ Friction Point: High drop-off rate for sending prompt

The CTR for the generate button is only 24%, indicating a high drop-off rate.

The CTR for the generate button is only 24%, indicating a high drop-off rate.

Design Solution: Visual optimization with action-driven labeling
Design Solution: Visual optimization with action-driven labeling

Visually optimized the 'send' button by adding a label that says 'Instant Generate' to make it more prominent and noticeable.

Visually optimized the 'send' button by adding a label that says 'Instant Generate' to make it more prominent and noticeable.

βœ… Final Design

βœ… Final Design

πŸ“ Before - 60min
πŸ“ Before - 60min

Browse template β†’ Manually adjust layout β†’ Add content β†’ Repeat revisions β†’ Finalize.

Browse template β†’ Manually adjust layout β†’ Add content β†’ Repeat revisions β†’ Finalize.

πŸ€–οΈ After - 8min
πŸ€–οΈ After - 8min

AI auto-generate assessments with unlimited regeneration.

AI auto-generate assessments with unlimited regeneration.

Solution 1 β†’ Feature 2

Solution 1 β†’ Feature 2

Feedback Auto-Generation β€” Automatically generate personalized feedback based on subject and students' grade

Feedback Auto-Generation β€” Automatically generate personalized feedback based on subject and students' grade

Feedback Auto-Generation β€” Automatically generate personalized feedback based on subject and students' grade

πŸ“ Before - 120min
πŸ“ Before - 120min

Select student β†’ Manually write comments (1.5min) β†’ Jump to next β†’ Repeat for 80 students.

Select student β†’ Manually write comments (1.5min) β†’ Jump to next β†’ Repeat for 80 students.

πŸ€–οΈ After - 40min
πŸ€–οΈ After - 40min

AI auto-generate personalized feedbacks for 80 students.

AI auto-generate personalized feedbacks for 80 students.

Solution 2

Solution 2

AI Agent β€” AI agents autonomously handle repetitive operations via automated workflows, eliminating repetitive workloads

AI Agent β€” AI agents autonomously handle repetitive operations via automated workflows, eliminating repetitive workloads

AI Agent β€” AI agents autonomously handle repetitive operations via automated workflows, eliminating repetitive workloads

Lesson Prep
Lesson Prep
Issue Homework and Newsletter
Issue Homework and Newsletter

Solution 2 β€” Feature 1

Solution 2 β€” Feature 1

Lesson Preparation β€” Rapidly generate lesson plans and materials tailored to course contents and curriculum goals

Lesson Preparation β€” Rapidly generate lesson plans and materials tailored to course contents and curriculum goals

Lesson Preparation β€” Rapidly generate lesson plans and materials tailored to course contents and curriculum goals

πŸ“ Before - 120min
πŸ“ Before - 120min

Review textbook β†’ Review practice questions β†’ Browse template β†’ Look for course images β†’ Add content β†’ Adjust slides layout β†’ Revise β†’ Finalize.

Review textbook β†’ Review practice questions β†’ Browse template β†’ Look for course images β†’ Add content β†’ Adjust slides layout β†’ Revise β†’ Finalize.

πŸ€–οΈ After - 10min
πŸ€–οΈ After - 10min

AI auto-generate lesson plans with slides and images.

AI auto-generate lesson plans with slides and images.

Solution 2 β€” Feature 2

Solution 2 β€” Feature 2

Streamling Workflow β€” Streamline multi-step tasks and accomplish it with single prompt

Streamling Workflow β€” Streamline multi-step tasks and accomplish it with single prompt

Streamling Workflow β€” Streamline multi-step tasks and accomplish it with single prompt

Design Details

Design Details

Progress Visualization During Result Generation β€” Visualize real-time generation progress, and alleviate impatience during AI processing

Progress Visualization During Result Generation β€” Visualize real-time generation progress, and alleviate impatience during AI processing

Progress Visualization During Result Generation β€” Visualize real-time generation progress, and alleviate impatience during AI processing

⚠️ Friction Point: 34.6% drop-out rate
⚠️ Friction Point: 34.6% drop-out rate

Due to technical constraint, the average loading is 30s, resulting in 34.6% drop-out rate during loading phase.

Due to technical constraint, the average loading is 30s, resulting in 34.6% drop-out rate during loading phase.

Design Solution: Reduce drop-out rate to 18.2%
Design Solution: Reduce drop-out rate to 18.2%

To ease user impatience and boost retention, I visualized decision-making with percentage-based progress cuesβ€”reducing drop-out rate to 18.2%.

To ease user impatience and boost retention, I visualized decision-making with percentage-based progress cuesβ€”reducing drop-out rate to 18.2%.

Character Customization β€” Customize AI Assistant's character for a personalized experience and foster emotional engagement between users and AI

Character Customization β€” Customize AI Assistant's character for a personalized experience and foster emotional engagement between users and AI

Character Customization β€” Customize AI Assistant's character for a personalized experience and foster emotional engagement between users and AI

Transform AI into a relatable companion: design human-like characters and IM-driven chats that feel authentic, not robotic.
Transform AI into a relatable companion: design human-like characters and IM-driven chats that feel authentic, not robotic.