Freddy
Designing a Human-Centered AI Meeting Assistant
Designing AI for Transparent, Inclusive, and Productive Meetings
ROLE
UX Designer, H-AII Interaction Designer
YEAR
2023
Modern meetings are increasingly remote, complex, and often inefficient. Professionals waste time revisiting discussions, aligning on action items, or just catching up. Freddy, an AI-powered assistant, was created to solve this with
Intelligent meeting summaries
Seamless scheduling
Natural language chat interface
Multilingual transcription
Post-meeting insights
The primary aim of this project was to design an AI meeting assistant that embodies ethical principles in Human-AI interaction. The goal was not just functionality, but a responsible, transparent, and inclusive experience for all users.
Trust by Design
Build user confidence through transparent AI behaviors, explainability, and clear system boundaries.
Ethical Intelligence
Incorporate Microsoft's AI guidelines to ensure the AI respects privacy, reduces bias, and supports fair participation.
Human-Centered Autonomy
Give users the control, allowing them to guide, correct, or override AI actions at any time.
Designing with Ethics
We based Freddy’s design on Microsoft’s 18 Human-AI Interaction Guidelines. Out of the 18, we implemented 11 that were directly relevant to our project. Some of the key principles we followed are
Transparency
Provide clear explanation on Freddy’s capabilities, limitations, and how data is used.
Correctability
Users can review, edit, or override AI-generated summaries and actions at any time.
Inclusivity
Freddy supports multiple languages and avoids jargon.
User Control & Feedback
Freddy can be invoked, dismissed, or corrected mid-meeting, ensuring the user stays in control.
This project wasn’t just about building an AI assistant, it was about designing a responsible, human-centered experience with AI in mind. Through Freddy, we explored how ethical principles can be embedded into everyday tools, making AI more transparent, explanable, and respectful. While testing is the next step, the foundation is now in place for real-world feedback and iteration.