Visitor questions
Guide people through approved product, service, location, policy or help content and point them to relevant source pages.
Website AI Chatbot
AI chat support for websites and WordPress businesses.
A website AI chatbot helps customers get quick answers from your own business content while giving your team a cleaner path to capture leads and handle escalations.
Nanovise works with Artbits Solutions Ltd. for many standard SmartChat-style deployments: self-hosted website chat, knowledge-base answers, live agent handover, lead CRM workflows, and WhatsApp follow-up options.
Answers from your site and documents
Use pages, FAQs, PDFs, policies, product notes, and onboarding documents to ground replies.
Live agent handover
When the customer asks for a person or the question is sensitive, the assistant can route the conversation to your team.
Lead capture and follow-up
Capture visitor details, summarize intent, and prepare next steps for sales or support follow-up.
Good fit for
Product websites, hotel and restaurant sites, PM-WANI information pages, support portals, SaaS websites, and service businesses that want faster first response without losing human escalation.
Website AI chatbot and RAG knowledge assistant
Help people find a grounded answer—and a person when they need one
SmartChat can be scoped as a website assistant or an internal knowledge interface that retrieves from approved content before answering. Nanovise designs the source pipeline, conversation experience, integrations, uncertainty behaviour and human handoff around the intended users.
Chat and knowledge use cases
One interface, different audiences and controls
A public website visitor, a customer with an account question and an employee searching internal policy do not need the same sources or permissions. SmartChat is scoped around the audience and the decision the answer supports.
Qualification
Answer common pre-sales questions, collect the minimum useful details and route a structured enquiry to sales when consent and integration are in place.
Self-service triage
Retrieve approved troubleshooting guidance, clarify the issue and create a better support handoff when the answer is not in scope.
Knowledge assistant
Help authorised staff navigate policies, product material or operating documents, with access controls and source visibility suited to the content.
How a RAG knowledge assistant forms a grounded response
Retrieval-augmented generation adds a controlled knowledge step before the model composes its answer.
- Question A visitor or authorised team member asks in natural language.
- Retrieve The system searches the content collection permitted for that user.
- Rank context Relevant passages are selected within configured limits.
- Answer & source The model responds from that context and can expose references.
- Clarify or hand off Missing, risky or uncertain requests follow a safer route.
RAG reduces guesswork; it does not remove the need for evaluation
Retrieval can make answers more specific to an organisation and easier to trace, but the assistant can still retrieve the wrong passage, misread context or receive incomplete content. Source quality and testing remain essential.
The design can require source links, decline unsupported questions, ask for clarification and route sensitive categories to a person.
Content ownership is part of the product
The project identifies which website pages, articles, product records, manuals, policies or internal documents are approved; who can change them; how often the index refreshes; and how obsolete material is removed.
Different collections can be separated for public, customer or employee audiences when the chosen architecture supports the required access controls.
Implementation scope
A useful chatbot includes the work behind the chat window
The interface is only one layer. A production assistant also needs a maintained content pipeline, clear boundaries, evaluation scenarios, analytics appropriate to the use case and an operating process for unanswered questions.
- Audience, intents and exclusions
- Approved sources and source owners
- Document parsing and update approach
- Retrieval and answer instructions
- Citations or source-link behaviour
- Access and data-retention controls
- Lead or ticket integrations
- Human handoff and feedback review
From answer to action
Add integrations only when they improve the journey
A website assistant may be assessed for creating a lead, preparing a ticket, looking up permitted data, scheduling a callback or routing the conversation into an existing inbox. Each integration depends on suitable APIs, permissions and a clear owner for exceptions.
High-impact writes can remain draft-only or require explicit confirmation. The interface should tell the user when an action succeeds, fails or has been handed to a person.
Evaluation beyond “it answered”
Representative testing can review retrieval relevance, factual consistency with sources, citation correctness, unsupported-answer refusal, access boundaries, latency, handoff quality and user feedback. Acceptance thresholds are set for the actual audience and content risk.
Questions to resolve before a pilot
AI chatbot and RAG assistant FAQ
What is a RAG knowledge assistant?
A retrieval-augmented generation, or RAG, assistant searches approved content for relevant context before composing an answer. This can make responses more specific to the organisation and can support source links or citations when the content and interface are configured for them.
How is a RAG assistant different from a general chatbot?
A general chatbot may rely mainly on broad model knowledge and instructions. A RAG assistant retrieves from a controlled business knowledge collection at answer time, which improves traceability and makes content maintenance part of the operating process.
Can SmartChat answer from website pages and internal documents?
Potential sources can include approved website content, help articles, policies, product information or internal documents. The usable sources depend on ownership, format, access controls, sensitivity and the intended audience.
Can the chatbot collect leads or create support tickets?
Yes, when the required form fields, consent language and destination system are agreed and a suitable integration is available. High-impact write actions should be constrained, validated and logged.
What happens when the assistant does not know the answer?
The preferred behaviour is to acknowledge uncertainty, ask a clarifying question, show relevant sources or route the conversation to a person. The assistant should not be instructed to invent an answer when approved context is missing.
Bring the questions people ask and the sources they should trust
Nanovise can help assess content readiness, audience access, retrieval design, integrations, evaluation and handoff before a SmartChat or RAG assistant moves into production.
Nanovise.com © 2026. All rights reserved.