Hybrid Chat and Human Agent Pipeline
I developed a chat platform that combines chatbots and real people. We call this "Artificial AI"
Messages coming in from users go through this pipeline:
1. Logic Layer
The RiBot scripting runtime manages state variables, allows conditional branching and script logic. There is also a simple regex type parser.
2. NAP Natural Language Processor
If the RiBot layer cannot interpret users input we pass through to our NAP layer, to extract intents and entities. This can be thought of as a 'fuzzy matcher' that then returns JSON results back to the Logic layer.
3. Human Agent / LiveChat
If neither of these layers find a result, a flag is triggered to call a human agent to respond. Responses can be added to the database so the system gets smarter and next time the bot will answer. This is what I call the "Learning Loop" and is critical for any AI/ML business model so the company gains value over time.
The RIKAI LiveChat dashboard has all the tools for an agent to manage the conversation.