You can easily install an AI chat function on your company website.
Even if all the necessary information is available on the website or in the manual, customers don’t have the time to look it up every time. Therefore, they’re more likely to call and ask a question. It might just be to confirm something very simple. However, if the person who answers the phone can’t provide a satisfactory answer, the customer’s anger will explode. Easy access to information greatly contributes to improving service quality.
Therefore, many companies are now actively promoting customer support through AI chat. AIs like ChatGPT learn from all kinds of information on the internet and are very knowledgeable, but they cannot cover the latest information even for products and services of large and well-known companies, let alone those of ordinary companies. This is where a technology called RAG (Search Augmentation) comes in. In other words, information related to the user’s question is added and provided to the AI, allowing the AI to generate an answer.
The first technique to emerge was vector search, which searches for documents with similar meanings that have a similar vector to the question (input text). However, the crucial problem was the very basic fact that “the text of the question and the text of the answer are not necessarily similar.” Next came graph RAG, which uses a knowledge graph to build a model that is compatible with AI from text information in advance and finds related information through “inference.” While theoretically this method seems perfect, building the model is not easy, and it is virtually impossible to model all the meanings in a text in the first place.
ThinkNavi proposes a third method: a self-organizing conceptual structure network model. Roughly speaking, it’s a model that groups similar texts and connects each group in a network-like structure. In other words, it’s a method somewhere between vector search and graph RAG. In vector search, each document is in a disorganized state and unorganized, and graph RAG attempts to model the relationships between entities in text comprehensively and in detail, but ThinkNavi organizes text information in an efficient and economical way.
ThinkNavi’s greatest strength is that it operates using only a lightweight conceptual structure network model, without requiring large-scale systems such as vector databases or graph databases.
Users simply enter the URL of their existing website into ThinkNavi or upload a PDF or Word file or MD, Text file, receive a short code generated by ThinkNavi, and paste it onto their website. In as little as a few clicks, your company’s personalized chatbot (concierge) will be ready.
The only costs are a base fee of 6.4 USD/month and 0.032 USD/Chat.