Intelligent chatbot
An AI chatbot that knows your catalog and turns visits into business
A chatbot that says "I didn't understand, type 1" drives customers away. A well-built AI chatbot understands the question, queries your system and answers like your best salesperson. OnWeb builds AI chatbots for websites, portals and systems.
- products recommended per answer by Safy, with reasons
- 1 to 3
- products recommended per answer by Safy, with reasons
- invented products: the catalog comes from the system
- 0
- invented products: the catalog comes from the system
- quotes and service on the Master Milhas website
- 24h
- quotes and service on the Master Milhas website
- languages served by the same chatbot
- 3
- languages served by the same chatbot
How a real AI chatbot works
An AI chatbot uses a language model to understand questions in natural language. What separates a useful chatbot from one that makes things up is what happens before the model answers: in every conversation, the system gives the model the right company data (catalog, prices, policies, customer history) and the rules for what it may say.
That way the chatbot answers with real, current information, cites products that exist and leads the person to the next action: the product page, the quote, the booking or a contact with the team.
AI chatbots we have built
Safy, Safetline
receives the live safety footwear catalog and recommends one to three products for the asker's job. See the case.
Davi, Master Milhas
checks the day's price table and records the quote in the CRM when the person decides to sell. See the case.
EDI, Luz no Bolso
reads a photo of the electricity bill and compares local energy companies. See the case.
Alfred, OnWeb
the chat on this website: answers questions about OnWeb and routes the contact to the team.
What your chatbot can do
- Answer from your content: catalog, knowledge base, policies and technical documents.
- Recommend: suggest a product or plan with a reason, based on the asker's profile.
- Quote and simulate: calculate price or savings with the business's real rules.
- Capture and qualify leads: log contact and need directly in the CRM, with the source page.
- Read files: photos, PDFs or spreadsheets sent by the visitor.
- Speak several languages: answers in the visitor's language with the same content.
- Be edited by the team: assistant name, image and texts changed in the panel.
AI chatbot, AI agent or WhatsApp?
An AI chatbot is the starting point for serving and converting on the website or inside a system. When the chatbot starts running whole processes (signing up, checking documents, updating systems), it becomes an AI agent. And when the customer's main channel is WhatsApp, the path is a WhatsApp chatbot. All three share the same base and can coexist.
How we avoid wrong answers
Hallucination is the main risk of an AI chatbot. The defenses we use:
- Source data in every conversation: the model answers with the current catalog and table, not with what it learned from the internet.
- Structured answers: the chatbot returns data in a fixed format (product, reason, next step), checked by the system before display.
- Explicit limits: what it does not know, it says it does not know and routes onward.
- Reviewable history: the team reads conversations and adjusts rules from them.
What you get
Chatbot on your website or system
a light, responsive widget with your brand identity.
Connection to your data
catalog, prices, knowledge base and CRM.
Rules and tone of voice
what the chatbot says, how it says it and when it hands over to the team.
Lead capture
contact and conversation delivered to sales.
Conversation panel
full history for the team to read and improve.
Continuous tuning
monthly calibration based on real conversations.
How we build it
- 01
Real questions
we collect what customers ask today by email, phone and forms.
- 02
Data sources
we connect the chatbot to the catalog, price table and CRM.
- 03
Personality and rules
name, tone, limits and next steps for each type of conversation.
- 04
Team test
the team talks to the chatbot before the public does.
- 05
Publish and calibrate
live, with conversation review and weekly tweaks at first.
Built by OnWeb
Manufacturing · Safety footwear
Safetline
A three-language website with its own management panel, an AI assistant that recommends products, 3D models generated from photos and 627 legacy URLs migrated without losing Google.
Read the case
Travel · Airline miles buying
Master Milhas
A website with instant quotes, a custom CRM with a deal pipeline, a client area and Davi, an AI chatbot that checks the day's price and creates the quote in the CRM.
Read the case
Energy · Community solar comparison
Luz no Bolso
A community solar comparison platform with 75 companies in 26 Brazilian states and EDI, an AI that reads a photo of the electricity bill, compares local providers and completes sign-up in the chat.
Read the caseFrequently asked questions
Which AI do you use in the chatbot?
We use models from Anthropic (Claude), Google (Gemini) and OpenAI, choosing the best for each task, with automatic failover if a provider fails.
Can the chatbot make information up?
The risk exists in generic chatbots. In ours, answers are grounded in company data sent in every conversation, with a format checked by the system and explicit limits.
How much does an AI chatbot for a website cost?
Development depends on integrations and content volume. AI usage typically costs cents per conversation. Talk to us for an estimate.
Does the chatbot work in Portuguese and Spanish?
Yes. It answers in the visitor's language using the same catalog and rules.
Related services
Shall we turn your business on?
Tell us what you need. We reply with a diagnosis of the technical path, timeline and investment, with no strings attached.