gensai.aiChatbots and answer engines
Conversational assistants that answer correctly, from your own content
Gensai designs AI chatbots for businesses and institutions: conversational assistants connected to your documents, catalogues and business databases through RAG, that answer in natural language, cite their sources and know how to say they do not know. On your website, on WhatsApp or internally, the same answer engine serves your customers, your visitors and your teams.
At the heart of an AI chatbot sits an answer engine: it does not return a list of links, it answers. The user asks the question as they would ask a colleague and gets a precise, written answer grounded in your content. Semantic search finds the relevant passages even when the words differ, reranking refines relevance, and the model generates the answer while citing its sources.
Each assistant is built on your data: internal documentation, catalogues, regulatory databases, FAQs, history. The RAG architecture anchors the answers in this verified content, which removes most hallucinations: the chatbot only answers about what it knows, and knows how to say it does not know. Test sets measure reliability before any move to production.
The same foundation runs on every channel: a chat window on your website, WhatsApp Business, intranet, business application. And confidentiality guides the architecture: depending on your requirements, everything is hosted on a sovereign cloud, on your servers or on site, with open models. Your documents are never used to train third-party models, and your users' access rights are enforced on every query.
Use cases for conversational AI
Automated customer support
An intelligent FAQ that really answers: fewer tickets, self-sufficient customers, a support team relieved of recurring questions.
Internal knowledge base
Your staff query HR, IT or quality procedures in natural language and get the right answer immediately.
Visitor information
Opening hours, prices, access, programme: a tourist office, a museum or a leisure park answers in every language, from its official content.
Event companion on WhatsApp
Trade shows, congresses, seminars: a WhatsApp chatbot accompanies participants in real time, from the schedule to access maps.
Regulatory monitoring
Query legal and normative databases by sector and company size, with answers sourced article by article.
E-commerce product search
The customer describes their need, the engine understands the implicit criteria and suggests the matching products from the catalogue.
Quoting and sales assistance
A virtual expert that knows your products, your prices and your history: faster, better qualified quotes.
Members and donors relations
A non-profit informs, answers frequent questions and simplifies registrations, without mobilising its volunteers.
Our references
Frequently asked questions
What is an AI chatbot?
An AI chatbot is a conversational assistant that understands a question asked in natural language and answers it with a language model. Connected to your data through RAG, it becomes a reliable business chatbot: it answers about your products, your procedures or your opening hours, not about generalities.
What is the difference between an AI chatbot and an answer engine?
The answer engine is the component that retrieves the information and writes the sourced answer. The chatbot is the dialogue interface that exposes it to users, with the conversation thread, rephrasing and channels (website, WhatsApp, intranet). The two go together, and the same engine can also be integrated without a chatbot, in an internal search engine for example.
How do you avoid hallucinations?
Answers are anchored in your content by RAG: the model only answers from the retrieved passages, cites its sources and knows how to say it does not know. Reranking steps and evaluation sets measure reliability before going live.
Can a chatbot be deployed on WhatsApp?
Yes, through the WhatsApp Business API. It is a very effective channel when your users are on the move, especially at events: an assistant centralises all the useful information there and answers instantly, in the thread everyone already knows.
Can a chatbot work without the Internet?
Yes. A chatbot connected to your data can run entirely on site, on a simple Mac Mini: language model, RAG and interface, without any connection. Gensai did it for a complete voice installation in a busy aquarium, and the same architecture applies to text.
Does my data stay confidential?
Yes. Depending on your requirements, the assistant is deployed on a certified sovereign cloud, on your servers (on premise) or even on site. Your documents are never used to train third-party models.
How do you measure the quality of the answers?
Before going live, a set of real questions serves as a test: rate of correct, sourced answers, unanswered questions, ambiguous wording. Once online, conversations are analysed to enrich the knowledge base wherever users get stuck.
How much does an AI chatbot cost?
As a guide, a first use case on real data costs between €3,000 and €5,000 excluding VAT, and a first operational version between €5,000 and €10,000 excluding VAT, hosting and model usage not included. The cost mainly depends on the number of sources to connect and the channels to cover.
A chatbot or answer engine project?
Tell us about your data and your users: Gensai proposes a concrete approach, from the first use case to production.