How to Add an AI Chatbot to Your Business Website
An AI chatbot on your business website answers customer questions instantly, qualifies leads while you sleep, and ensures no visitor leaves without getting the help they came for. This guide covers everything from choosing the right type of chatbot to building and deploying it on your website.
Choosing the Right Fit
Scripted FAQ chatbot
Pre-defined questions and answers — the chatbot presents menu options and provides pre-written responses. Best for: businesses with a limited, well-defined set of customer questions (booking a service, checking opening hours, understanding pricing tiers). Limitations: cannot handle questions outside the scripted menu, feels robotic when the visitor’s question does not match a menu option. Build time: 1 to 3 days. Cost: $0 to $50/month with tools like Tidio, ManyChat, or GoHighLevel’s built-in bot.
AI knowledge base chatbot
An AI model trained on your business knowledge that can answer any question about your business in natural language. Best for: businesses with diverse customer questions, complex service offerings, or customers who need specific information before making contact. The chatbot from Post 201 — built on Bubble.io and Claude — is this type. Handles any question the knowledge base covers and gracefully deflects questions it cannot answer. Build time: 1 to 2 days. Cost: $29/month Bubble hosting + $10 to $30/month Claude API usage.
AI lead qualification chatbot
An AI that not only answers questions but actively qualifies the visitor — understanding their situation, collecting their requirements, and routing them appropriately (booking a call for qualified prospects, sending a resource for early-stage researchers). Best for: businesses with a consultative sales process where the conversation determines whether the prospect is a good fit. Build time: 1 to 3 weeks depending on the qualification logic. Cost: as above plus GoHighLevel integration for lead routing.
The Practical Build Guide
Write your knowledge base
The quality of an AI chatbot depends entirely on the quality of its knowledge base. Before any technical build: write a comprehensive document covering everything the chatbot should know about your business. Sections to include: what you do (service descriptions, who they are for, what they include), how it works (the process from first contact to delivered outcome), pricing (ranges, what affects price, how to get a specific quote), who you serve (industries, company sizes, geographies), frequently asked questions with honest answers, and what to do next (how to contact you, how to book a consultation, what happens after they enquire). This document is the chatbot’s brain — the better it is, the better the chatbot performs.
Set up the chatbot in Bubble.io
Follow the chatbot build guide from Post 201: (1) create the Bubble.io application with the Conversation and Message data types, (2) build the chat interface (Repeating Group for message history, Input for user message, Send button), (3) set up the API Connector for the Claude API with your API key, (4) build the Send Message workflow (create user message, call Claude API with system prompt + conversation history, create assistant message, refresh the repeating group), (5) handle session management (create a Conversation record on page load, link all messages to it). Test by previewing the page and asking questions about your business.
Write the system prompt
The system prompt is the most important technical element — it defines how the chatbot behaves. A well-structured system prompt: You are the AI assistant for [Business Name], a [brief description]. Your role is to help website visitors understand what we offer and whether we can help them. Guidelines: (1) answer only based on the company information provided below — do not guess or invent information, (2) be friendly, professional, and concise — respond in 2 to 4 sentences maximum unless more detail is genuinely needed, (3) if asked about pricing, provide the range from the information below and offer to connect them with the team for a specific quote, (4) if the visitor seems like a potential client, offer to help them book a consultation using this link: [Calendly link], (5) if you cannot answer a question from the information provided, say so clearly and offer to have a team member follow up. Company information: [paste your knowledge base document].
Embed on your website and test
Embed the Bubble.io chatbot on your website using an iframe or a floating widget built with a small JavaScript snippet. Test with the questions your customers most commonly ask, the questions that previously required a phone call to answer, and questions outside the knowledge base (to verify the graceful deflection works). Share with 3 to 5 colleagues and ask them to try to break it — finding edge cases before customers do. Deploy after addressing the most common issues found in testing.
📌 The single most important chatbot improvement after launch: review the conversation logs every week for the first month. You will find: questions the chatbot answered incorrectly (update the knowledge base), questions the chatbot could not answer (add information), patterns in what visitors are asking (signals about what your website copy is not making clear), and conversations where the visitor was clearly a good prospect but did not book a call (improve the booking CTA trigger logic).
How do I make the chatbot sound like my brand rather than a generic AI?
The brand voice is controlled through the system prompt. Include specific instructions: our tone is [3 adjectives], we always [specific communication habit], we never use [specific phrases or jargon to avoid], and when customers ask about [sensitive topic], we [specific approach]. Also include 3 to 5 example exchanges showing how the chatbot should respond to common questions — few-shot examples in the system prompt produce significantly more brand-consistent responses than abstract instructions alone.
What is the difference between a chatbot on my website and a live chat tool?
A live chat tool (Intercom, Drift, Crisp) connects a visitor with a human team member in real time — or provides a bot that escalates to human. An AI chatbot handles the conversation without a human unless the visitor requests it. The choice depends on your business: if you have a team available during business hours and most enquiries are sales conversations where human relationship-building is important, live chat plus AI for after-hours makes sense. If most enquiries are information-seeking (pricing, service details, how it works) and your team is small, a pure AI chatbot handles most conversations without any human involvement.
Want an AI Chatbot Built for Your Website?
SA Solutions builds AI chatbots on Bubble.io — knowledge base-powered, brand-voiced, and integrated with GoHighLevel for lead capture and booking.
