Inspiration
The Cornell Team and Leadership Center (CTLC) serves the Cornell and greater New York communities through programs centered on teambuilding and professional development. We believe that effective facilitation skills shouldn't be kept behind closed doors - everyone should have access to the tools and knowledge needed to build stronger, more collaborative communities.
As a group of colleagues from CTLC, we saw an opportunity to leverage generative AI to democratize access to our expertise. Whether they're a new or experienced facilitator on our team or someone outside CTLC looking to run a teambuilding session, CTLChat is designed to help anyone create meaningful programs that bring people together.
What it does
CTLChat is an intelligent program design assistant powered by our collection of 200+ vetted teambuilding activities. Users simply describe their needs - group size, goals, time constraints, experience level - and the chatbot generates a customized program schedule complete with activity recommendations tailored to those specific parameters.
How we built it
We began by identifying the key variables our experienced facilitators consider when designing programs. We then created a comprehensive database of our activities, documenting the skills each one develops (collaboration, communication, trust-building, etc.) along with facilitation guidelines. Using Claude's API, we built a RAG (Retrieval-Augmented Generation) chatbot that intelligently matches activities to user requirements, then wrapped it in an interactive frontend interface.
Challenges we ran into
- Version control learning curve: Half our team was new to Git, which slowed our initial workflow as we learned to collaborate effectively through version control.
- Document digitization: We wanted to include real program logs from past years as training examples, but these handwritten documents proved difficult for Claude's vision capabilities to parse accurately, requiring significant manual cleanup.
- Development environment inconsistencies: We encountered platform-specific issues—one team member faced storage limitations while another had to adapt our Mac-based codebase for Windows compatibility.
Accomplishments that we're proud of
- Mastered Git workflows as a team and established smooth collaboration practices
- Transformed 200+ activities into clean, well-structured Markdown optimized for vector search
- Successfully implemented a RAG pipeline that provides contextually relevant recommendations
- Designed and deployed an intuitive, user-friendly frontend
- Strengthened our team dynamics through collaborative problem-solving
- Built a solid foundation that's ready for future expansion and refinement
What we learned
- Generative AI dramatically accelerates development in unfamiliar technical domains
- Data curation is a time-intensive but critical component of RAG systems
- Thoughtful database structure directly impacts the quality of AI-generated recommendations
What's next for CTLChat
We're planning several enhancements to make CTLChat even more valuable:
- Refined complexity ratings: A more nuanced difficulty assessment system to better match activities to group experience levels
- Enhanced facilitation guides: Detailed tips, common pitfalls, and troubleshooting advice for each activity
- Expert insights integration: Wisdom and best practices from CTLC's veteran facilitators
- Broader accessibility: Expanding the platform to serve organizations and communities beyond Cornell
Our ultimate vision is to make high-quality teambuilding facilitation accessible through CTLChat to anyone looking to strengthen their community through collaborative experiences.
Log in or sign up for Devpost to join the conversation.