Fluxion

Inspiration

Organisations worldwide struggle with implementing technological change. More than 50% of change initiatives fail due to poor communication and inadequate planning.

What it does

Fluxion is an AI-powered change management assistant that transforms how organisations implement digital transformation initiatives. It leverages ProSci methodologies and ADKAR frameworks through a specialised knowledge base, combining RAG technology with real-time web search capabilities via LangChain and SerpAPI. Users can create projects, upload supporting documentation, and receive tailored strategies, communication plans, and measurement frameworks—all informed by industry best practices and current research.

How we built it

We architected Fluxion using TypeJS and Vercel for the frontend, with a robust AI pipeline built on LangChain orchestration. Our system ingests and processes change management frameworks from ProSci and research papers into vector embeddings for our RAG system. We implemented a fallback mechanism that seamlessly transitions to SerpAPI when the knowledge base requires supplementation. The Gemini model powers our response generation, while the AI SDK handles the integration between components.

Challenges we ran into

The development journey was not without its obstacles. Setting up Supabase proved particularly challenging as we encountered authentication issues and database schema complexities when storing conversation histories and project data. Vector database integration for our RAG system required significant optimisation to maintain performance while scaling our knowledge base. Additionally, orchestrating the LangChain components to smoothly transition between RAG and SerpAPI required careful prompt engineering and extensive testing to ensure consistent, high-quality responses.

Accomplishments that we're proud of

We are honestly just happy that we managed to piece everything together

What we learned

Building Fluxion challenged us to deeply understand both the technical and human aspects of change management. We discovered that effective AI solutions in this space require not just technical sophistication but domain-specific knowledge integration. The process of designing our LangChain architecture taught us valuable lessons about prompt engineering and context windows.

What's next for Fluxion

Future iterations will include more sophisticated document generation features with the capability to create customized training materials, impact assessments, and stakeholder-specific communication plans.

Built With

• aisdk
• gemini
• huggingface
• langchain
• typejs
• vercel