Health tech · AI
Ponder App
A GPT-powered companion people talk to about how they feel. The brief was to grow it without compromising emotional safety.
- Mobile
- AI conversation
- Voice and text
The product
Ponder began as a proof of concept with voice-to-text and GPT-powered chat. As adoption grew, it became a fuller wellness app: personalised recommendations, guided practices, relationship coaching, and voice or text journaling with a conversational AI.
The challenge was to add all of that while keeping the experience comfortable and emotionally safe for the people using it.
What had to be tested
- Conversations with the AI in both voice and text modes
- GPT-powered course and conversation recommendations
- Wellness tools such as Habit Creator, Reframe It and Stop Spiraling, built on structured prompts
- Audio lessons with playback, progress tracking and completion feedback
- Saved sessions and personal plans that users return to
How testing fitted in
Testing ran in iterative cycles as each wellness flow was added, with emotional safety treated as a requirement rather than a nice-to-have.
Each enhancement was checked against established practice in emotional UX, voice-first interaction and mobile wellness design.
The outcome
What the project reported:
- More emotionally relevant recommendations
- A more engaging experience through modular, accessible tools
- Greater user autonomy through plans, practice guidance and session tracking
- Ready for multilingual and cross-platform growth
What we would test today
An AI that talks to people about how they feel is exactly where a wrong answer matters most. Today we would test what Ponder does when the model mishandles someone in distress, drifts into advice it should not give, or follows an instruction hidden in a user’s message, and re-run those checks whenever the model or prompt changes.
Tools and technologies
- Flutter
- OpenAI
- Express JS
- MongoDB
- MySQL
- AWS
- Figma
