AI for Good: Building a Chatbot for FCSN
Non-profit organizations often lack sufficient staff to respond promptly to the large volume of user inquiries they receive every day, from event schedules and volunteer opportunities to resource guides for families. Visitors to their websites frequently find it tedious, time-consuming, and even challenging to locate the information they need across multiple pages and subpages.
Recognizing this challenge, Eric Huang, a youth volunteer at Friends of Children with Special Needs (FCSN), and Catherine Liu, who had learned chatbot development at the CP Mentorship Workshop, teamed up in September 2024 to design and build a customized AI chatbot for FCSN with the support and guidance of CP Mentorship’s Catherine Liu. Their goal was to provide immediate and reliable support to users regardless of time zones or staff availability. This marked the beginning of the Chatbot for FCSN Project.
Catherine of CP Mentorship helped establish the AI for Good Committee and provided guidance and support for the Chatbot for FCSN project throughout its development
Yi is demonstrating how the chatbotresponds to prompts
Under the mentorship of Yi Wang, Eric and Catherine mastered the chatbot development process using the Coze platform. They collaboratively scraped data from over 400 FCSN web pages, ensuring the chatbot would have comprehensive and accurate information.
Catherine focused on developing the chatbot’s image database, leading testing efforts, and identifying FCSN’s needs. She designed and analyzed a survey to gather feedback from staff and families, and led bot testing. She built image databases by organizing and tagging over 300 images and photos showcasing FCSN programs and activities.
Eric, a USA Coding Olympiad Bronze Medalist, took the lead in developing the chatbot’s text and table databases, configuring its technical settings in Coze. He also conducted extensive testing to guarantee the chatbot’s performance and reliability, and provided problem-solving and technical support throughout the development phase and during chatbot integration.
Coze lacks advanced programming capabilities, which limits the precision of data processing, storage, and manipulation. Its API integration is restricted, and data can only be retrieved through Coze's built-in tools. In addition, Coze bots, powered by large language models (LLMs) such as integrated third-party APIs, often struggle with time- and date-sensitive queries because these models are trained on static datasets up to a certain cutoff. Without real-time access, the bot "hallucinates" outdated or generic info, like assuming it's always a certain year or miscalculating current events. This is a common LLM limitation, not unique to Coze.
Despite these challenges, Eric tackled many complex technical issues to ensure the chatbot operated smoothly, optimizing its functionality, resolving integration bugs, and improving system stability throughout the development process. Catherine refined data handling rules and image taggings to improve retrieval accuracy and enhance overall performance.
Eric is configuring the FCSN chatbot databse
Catherine, Eric, and mentor Yi are discussing technical details over Zoom
To ensure long-term sustainability, Eric and Catherine created tutorial videos and presentation slides to train Richard on how to build and update the chatbot. Throughout the chatbot transition and integration process, they provided regular progress updates to the FCSN team and ongoing technical support and troubleshooting to Richard.
Eric and Catherine are testing the FCSN chatbot using the most frequently asked questions
FCSN has been testing the chatbot and plans to integrate it into its website soon. Once live, the chatbot will serve as a 24/7 virtual assistant, providing immediate, accurate responses about FCSN’s programs, services, news, etc., complete with relevant images and direct links. This will free up valuable staff and volunteer time for more personalized, high-touch support.
Chatbot for FCSN Project represents more than a technical achievement. It embodies community engagement and real-world problem-solving. Through this experience, Eric and Catherine gained a deeper understanding of the special needs community and made meaningful contributions by leveraging technology to serve as a bridge, fostering compassion and service.
Eric and Catherine extend their sincere gratitude to:
Yi Wang, for mentorship and technical support
Catherine Liu of CP Mentorship, for encouragement, support, and guidance
FCSN management team and Richard Lo, for their collaboration and trust
AI for Good Committee members -Ethan Chueng and Russell Wang- for assistance with chatbot testing
