YuanBao, Tencent | 2024.07 - 2024.08
An AI writing feature that supports the full writing journey, driving 30K+ daily active users in its first week.
Last Summer, I joined Tencent's AI Creative Design team as a product design intern. ”Help me write” feature is one of the projects I involved from 0-1, from researching the problem to creating the design for implementation. It helps people quickly finishing their writing tasks in high quality.
Key metrics post-launch week:
Feature Daily Active User
30,000+
Feature Penetration Rate Among MAU
42%
User Satisfaction Rate
4.4 out of 5.0
1-week Long Article Generated
200,000+
Context
What is Tencent YuanBao?
Tencent YuanBao is a conversational AI assistant, similar to ChatGPT, designed to help users complete a wide range of tasks through natural language interaction. We launched the product on both web and mobile platforms in June 2024, and quickly reached 600K registered users and 120K DAU in the first week. Our first big goal is is to grow DAU to 1 million before 2025 Q2.
Opportunity
Improving writing experience as a potential growth driver
Writing-related tasks emerged as one of the most common use cases in YuanBao, yet our product lacked any experience optimized for it. Users relied on one-question-one-answer interactions to complete complex writing tasks, resulting in friction, inefficiency, and drop-off.
Data from the research team’s survey responses:
50%+
write on digital devices more than 5 times per week.
40%+
have already started using AI tools to assist their writing.
Research results & insights
What we learned from our users
Through 687 survey responses and 18 in-depth focus group interviews, our research team uncovered several key insights:
HMW Statement
How might we support users across the entire writing journey—from early ideation to final editing—while adapting to different writing styles and needs?
User Journey
Mapping the writing workflow with YuanBao
I created a customer journey map to align stakeholders and highlight key pain points in how users leverage YuanBao to assist with writing tasks.
Scoping
Finding where we can deliver the most impact
Based on the journey map and user research insights, we scoped the AI Writing feature around the two most critical stages in the writing flow:
Draft Stage
Refine Stage
But wait a second, why are we deciding not to include the Discovery & Research stage when users are also facing some challenges there?
From Product Perspective
Brainstorming and info searching aren’t limited to writing—they appear across many scenarios. We’ll build another feature to better support these needs.
From Tech Perspective
Current limitations in our model make it difficult to fully support Discover and Research phase tasks. Accuracy and timeliness need gradual model upgrades.
Solution #1 - Prompt Library
Enhancing First-Try Output Quality
One of the key challenges users faced was that AI-generated content often didn’t match their intent, especially when prompts were vague or unclear. To solve this, we designed a Prompt Library with curated, scenario-specific prompts—covering all high frequent tasks, such as emails, essays, reports, and social media posts.
Solution #2 - Outline-first
Supporting Structure Before Content
Long-form writing posed a unique problem: users wanted more control over structure and flow, but AI often generated content that felt scattered or missed key points. We introduce outline-first option, allowing users to generate a outline first. They can view & edit the outline before YuanBao generate the entire content.
Solution #3 - Inline editor
A seamless space for writing & refining
Instead of relying on back-and-forth Q&A to revise AI-generated content, we introduced a built-in rich text editor, similar to tools like Notion, designed for direct editing and quick adjustments.
Besides manual editing, we add a lot of quick AI editing tools to help users quickly fine-tune the details of the content.
Solution #4 - Mobile Version
Simplified Mobile experience: only what matters
Given tech constraints and mobile usage patterns, we chose to keep just the Prompt Library and redesigned the interactions to better fit mobile.
Reflection
What I learned from this project
This project gave me hands-on experience designing for AI under real product constraints. I learned how to align cross-functional teams, move fast without losing clarity, and adapt quickly when things don’t go as planned.
✅ What went well:
Created a journey map that aligned the team and resolved product scope debates
Managed time and priorities effectively, ensuring on-time delivery of the feature
⚠️ What could be improved:
Overestimated AI capabilities — a title suggestion feature led to high drop-off and was removed.
Forgot to consider new feature onboarding. This part is later designed by my mentor, HuoHuo : )