This repository contains the specification for an MVP (Minimum Viable Product) of an AI Chatbot that integrates with website data. The chatbot is designed to be user-friendly and easily trainable, targeting users who may not have a technical background. The MVP includes three different implementations to test user engagement and adoption: Chat, FAQ, and Search.
The primary objective of this MVP is to validate the following hypotheses:
- Existing AI chatbots are not user-friendly enough for non-developers.
- A quick-start engagement approach with easy retraining will enhance user experience.
- Offering multiple implementations (Chat, FAQ, Search) will increase market reach and user adoption.
- Functionality: Normal ChatGPT-like interaction, limited to 3 user messages.
- Use Case: General conversational engagement.
- Functionality: Users ask custom questions, and the AI generates answers based on website data.
- Use Case: Quick access to specific information.
- Functionality: Similar to Chat, but returns a single response per query.
- Use Case: Direct information retrieval.
After using the AI Chatbot (Chat, FAQ, or Search), users are presented with three response options to provide feedback:
- "Nice! It looks good!"
- Action: Encourages integration of our AI tools and prompts the user to contact us.
- "We have concerns..."
- Action: Opens a free text input or predefined response list (check-box group) to capture concerns.
- "Nope, it looks wrong"
- Action: Opens a free text input to detail what went wrong.
Users can edit the AI-generated replies to make them more specific to their organization. These edits are client-side only and used to gauge engagement with the retraining feature. The edits can later be recorded as Q&A data sources.
The MVP will be marketed by sending test URLs to potential clients, such as NGOs. These URLs will link to a demo where the AI bot has pre-crawled their website data. This approach allows clients to immediately see the chatbot in action and test its functionalities.
The MVP test aims to:
- Determine the feasibility and user engagement of the quick-start approach.
- Identify which implementation (Chat, FAQ, Search) has the highest adoption rate.
- Collect feedback on the user experience and potential areas for improvement.