AI chatbot for website
An AI chatbot for customer service that answers from your own data
An AI chatbot for your website that reads your help pages, your policies and your order data, and answers customers in your tone while they’re still on the page. When a case needs a person, it hands over with the full conversation, not a ticket number.
- Website chat
- Telegram
Ask it what your customers ask you
The assistant on this page answers from Pazl’s own knowledge base and hands anything else to a person. It behaves exactly like the chatbots we build for clients. Ask about prices, timelines or the process.
It’s live — try it right here, nothing saved.
What the AI chatbot does on your website
Answers from your knowledge base
Help articles, policies, product data, delivery terms, prices. The chatbot answers with what your documents actually say and invents nothing. A question outside the base is handed over, not improvised.
Looks things up in your systems
“Where is my order?”, “Is Saturday free?”, “What’s my booking?” The chatbot reads order status, availability or account data through your systems’ APIs, so the customer gets a real answer, not “please contact support”.
Captures and qualifies enquiries
When a visitor is ready to buy or book, the chatbot collects the details your team needs and sends them to your CRM or calendar. For a chatbot built mainly to qualify and follow up leads, see our AI sales agent.
Hands over with context
When a case needs judgement, a refund above a limit or an unhappy customer, the conversation reaches your team with the history, the customer record and a short summary. Your people start where the chatbot stopped.
Where an AI chatbot on your website pays off
An AI chatbot earns its place when many visitors ask questions that your own content already answers, and the answer decides whether they buy, book or stay.
- Clinics, salons and studios. Prices, preparation, availability and “can I reschedule?” are answered on the page, and the visit gets booked there too.
- Online shops. Delivery times, returns, sizes and order status. These are the questions that otherwise turn into support email.
- Service businesses. “Do you cover my area?”, “How much is it?”, “When can you come?” are answered from your rules. A qualified enquiry reaches your team.
- Software and subscriptions. How-to questions answered from your help centre, with account questions handed to a person.
- Agencies and B2B. First questions answered at any hour, and the brief collected before a manager joins.
It pays off less when every conversation is unique and needs expert judgement. Then the chatbot’s job is to collect the details and route the case, and we’ll scope it that way.
Chatbot for customer service beyond the website
Customers don’t only write on your site. The same chatbot, with one knowledge base and one set of rules, can answer in:
- Email. Routine questions answered, and the rest routed to the right person with a draft.
- WhatsApp. On the official Business API and your own number. See WhatsApp Business API setup.
- Telegram and Instagram. The same answers in the messengers your customers use.
Every conversation is logged in your help desk, CRM or inbox, so your team sees one history per customer, whatever the channel. Need calls answered too? A phone line is a separate project: see AI receptionist.
Grounded, measured, in your hands
The main risk with an AI chatbot is a confident wrong answer. We build against it from the start.
- One source of truth. The chatbot is built on the documents that hold your answers. Where the base has gaps, you get a list of them, not wrong answers.
- Tested before customers see it. We run it against real past conversations from your inbox, including the awkward ones.
- Rules decide what it may do alone. Answer, look up an order, book a slot, or hand over. The rules are yours and change as trust grows.
- You see the numbers. For the first weeks your team reviews transcripts. Resolution rate and hand-off rate are on a dashboard you can read.
- Your accounts, your data. The chatbot runs in your accounts and is connected to your help desk or CRM. Personal data is handled under GDPR, and EU hosting is available.
For where AI helps in customer support and where people should stay in charge, read our guide AI customer support: what to automate and what to leave to people.
AI website chatbot cost
What changes the estimate:
Channels. Website only, or website plus email, WhatsApp and messengers.
Systems the chatbot reads. Orders, bookings, accounts. Each one is an integration.
Knowledge base. Ready help pages, or content that needs to be collected and structured.
Languages. Each supported language is agreed and tested.
AI model usage is billed at cost and listed separately in the proposal. Every project comes with a six-month warranty, and ongoing support is set out in your agreement.
Chatbot plugin or a custom AI chatbot?
A ready-made chatbot plugin is quick to install and fine for general FAQs. A custom AI chatbot pays off when answers depend on your own data and systems.
Knowledge
- Chatbot plugin
- Pages it can crawl
- Custom AI chatbot from Pazl
- Your documents, policies and live system data
Actions
- Chatbot plugin
- Answer and collect an email
- Custom AI chatbot from Pazl
- Look up orders, book slots, update records
Hand-off
- Chatbot plugin
- Ticket or email
- Custom AI chatbot from Pazl
- Your team’s tool, with history and summary
Rules
- Chatbot plugin
- Plugin settings
- Custom AI chatbot from Pazl
- Your rules for what it may do alone
Channels
- Chatbot plugin
- Usually one widget
- Custom AI chatbot from Pazl
- Website, email, WhatsApp, messengers, one base
Data
- Chatbot plugin
- Plugin vendor’s cloud
- Custom AI chatbot from Pazl
- Your accounts; GDPR, EU hosting available
If a plugin covers your case, we’ll say so.
How we launch your AI website chatbot
Send your most common questions
Share the questions your team answers every day, where the answers live and which channels customers use.
Agree the scope and a fixed price
Channels, knowledge sources, systems to read, hand-off rules, price and delivery date, in writing.
Build the knowledge base and rules
We structure your content and write the rules for what the chatbot answers, looks up and hands over.
Test on past conversations
The chatbot answers real questions from your history, and we close the gaps before launch.
Launch on your website
The widget goes live, your team reviews transcripts for the first weeks, and we tune the rules.
Extend
Add channels, systems or new topics once the first set is resolved reliably.
A first chatbot can be live from around three weeks once the knowledge base and a sample of past conversations are ready.
Project examples: an AI agent for an agency, and the assistant on this page
AI agent “Vera” takes over a wedding agency’s daily conversations (text, voice, photos) and brings a manager in only when it matters.
The Pazl Assistant
The chatbot at the top of this page runs on our own site. It answers from Pazl’s knowledge base and hands everything else to a person, the same setup we build for clients. Try it before you talk to us.
The stack we build with.
Fair questions, straight answers.
Send us your ten most common questions
Share the questions your team answers every day, where the answers live and which channels customers use. We’ll scope the chatbot, its rules and its hand-off, and estimate the work.