How might we boost engagement and retention for Le Chat?
Project Overview
Bootcamp Project
B2C Mobile App
2 Weeks Sprint
Team of 3 PrD
My Role
UI, UX Design
Prototyping
User Testing


Context
Mistral is France’s highest-valued unicorn at $6.2B. In 2024, they launched “Le Chat”, a chatbot similar to ChatGPT that runs on Mistral’s proprietary models.
Despite having a high-quality AI model, Le Chat struggles to compete and retain users against competitors like ChatGPT or Claude.
# of Le Chat users who remain daily active users after one month
Problem Statement
How might we boost engagement and retention for Le Chat?
Benchmarking
We benchmarked other apps to make sure our solutions conform to industry standard
Ideation
We used ideation techniques to brainstorm ways to improve stickiness of the app

V1 Solution
Our hypothesis: Mistral can differentiate by offering specialized, expert models rather than a single general-purpose AI
We brought “Le Chat” to life by creating a family of specialized cat assistants, each an expert in their domain


Each cat assistant suggests relevant prompts as you chat, guiding the conversation and helping you get better answers
Every conversation is saved in Chat History, while Artifacts preserve important takeaways and resources for easy access

User Testing
We tested our prototype with users to validate these objectives:
🎯 Onboarding
Users complete interest selection and select the right cat
💬 Smart Prompts
Users notice, use, and value contextual prompts
📚 Saved Content
Users easily navigate Chat History and access Artifacts
✨ Value Prop
Users prefer specialized cats over general-purpose AI
Profiles
4 Women, 2 Men
Ages 28 - 39
Prior LLM Experience
Parameters
Qualitative
Semi-structured interviews
45 min via Google Meet
Iteration
Based on user feedback, we prioritized two key areas of improvement:
1. Reduce number of categories and improve grouping
2. Use visual icons to strengthen connection between interest and cat experts


Changed auto-send to pre-fill behavior. Users can now edit suggested prompts before sending
V2 Solution